Why buy solid state drives? [Archive]

MasterNetra

March 25th, 2011, 02:47 AM

So it seems to me that technology is slowly moving away from HDD's and towards SSD's. When I first realized this, I thought it was terrific! After all, solid state drives can be accessed much quicker and do not have any moving parts to break. But then I realized something... don't SSD's use flash? And don't flash devices eventually lose the ability to be written to? Why buy something if you know it will DEFINITELY break in the future, where as with HDD's, it can be hit and miss (although I believe some HDD's in my house have lasted over 10 years). Thoughts?

I've been using a 4GB Flash Drive for 3 years now, still going strong. At any rate most Hard Drives don't make it that long. Also reading from Flash Drives doesn't take away from their life span just writing to it. Of course A: Life Span varies between brand/device/usage. B: There most likely are companies who may have it purposely design to fail after a period of time (usually past warranty) so that you will buy another. After all if someone produced a computer that lasted generations people wouldn't need to go back purchase another, the company would eventually go out of business or have to raise its price to match its duration most likely making it unaffordable. But thats how it goes, monetary economy doesn't allow for most efficient and longest lasting most of the time.

I bought a SSD last fall just to see if it was what everyone was raging about, but sadly, I will have to say that it's not. Yeah I notice a better startup and shut down time, but that's really about it. Games load a little faster and that's about it. It just sucks making sure you aren't running out of disk space all the time. :(

I have 80GB Hard Drive... as 120GB SSD would be a considerable improvement. But meh 10-20 years or so (if not sooner) SSD probably be obsolete or on the way... In favor of something probably dealing with NanoTech, meanwhile SSD's will improve and cheapen so meh.

racie

March 25th, 2011, 02:55 AM

As for lifespan, my OCZ Vertex drive is rated for 1.5 million hours of use. Correct my math if I'm wrong, but this should last 171 years. How that was determined may not be totally reliable though :P
Woah! What in the world?! I'm thinking something must be off here, but you never know. It's a relatively new technology.

I've been using a 4GB Flash Drive for 3 years now, still going strong. At any rate most Hard Drives don't make it that long.

Stop right there! This can't possibly be correct. I'd say most HDD's definitely surpass three years of usage.

One another note... I've heard that it's better to leave your computer in standby mode so you hard disks can remain spinning rather than starting and stopping (when turning off and shutting down the computer). I have no idea if this is true or not.

NMFTM

March 25th, 2011, 03:13 AM

Technically speaking, don't HDD and SDDs have the same number of moving parts?Think how the actual information is stored.

On a hard drive it's pieces of copper oxide (rust) on a magnetized platter that are in one of two positions to create the binary equivilent of 1's and 0's.

On a solid state drive there are billions of transistors that have a tiny little "gate" or w/e in them that's either in one of two positions as well.

Both of these are parts and both of them have to move to store data. Therefore, they're moving parts.

Khakilang

March 25th, 2011, 03:14 AM

I still using a 10GB hard disk which is about 10 years old. I install with Puppy Lucid Linux. The performance is quite good. Go for a reliable brand like Seagate.

NMFTM

March 25th, 2011, 03:16 AM

I still using a 10GB hard disk which is about 10 years old. I install with Puppy Lucid Linux. The performance is quite good. Go for a reliable brand like Seagate.
Oh God, here comes the inevitable WD vs Seagate debate. :popcorn:

Cracklepop

March 25th, 2011, 03:20 AM

HDDs will *definitely* break at some point too...

SSD are quieter, faster, cooler, consume less energy for the same amount of data written/read, and have no moving parts to break.

Khakilang

March 25th, 2011, 03:27 AM

Oh God, here comes the inevitable WD vs Seagate debate. :popcorn:

I don't hate Western Digital but just prefer Seagate. My 80GB WD died on me after 2 years. Luckily it is still under warranty. Got it claim and now using it for about 3 years now. I guess it could be the power fluctuation that kills it rather than a natural death.
:popcorn:

racie

March 25th, 2011, 03:34 AM

HDDs will *definitely* break at some point too...

SSD are quieter, faster, cooler, consume less energy for the same amount of data written/read, and have no moving parts to break.

Yes, but you KNOW that SDD's have a limited amount of writes to them, while HDD's do not. I dunno, I guess I just can't get over that.

But I guess if solid state drives really last 171 years, there's nothing to worry about. ;)

disabledaccount

March 25th, 2011, 03:42 AM

1. SSDs are fast - in general, but there are many exceptions.
Only latest medium to high-end SSDs with built in cache are faster than HDDs, especially in writing. Older and low-end SSD are slower or sometimes even deadly slow in comparison to new HDDs.

2. SSDs can last for 1.5 million of hours - FALSE: marketing or just lie (pick one)
Nand Flash data retention guaranteed by most manufacturers reaches up to 10 years for first written data -> every next write is shortening data retension, and first write errors can appear even after several thounsands of writes <to single flash page>
Thanks to speciall agorithms that are spreading writes to different flash pages (by dynamically remapping LBA) this weakness of Flash could be reduced to considerably low level. However, there is a catch: If the "disk" is nearly full, the spreading of writes becomes ineffective - and in some heavy write load conditions it's easy to wear several pages in several hours. The problem is, that SSD firmware cannot remap used pages, beacause it would need full implementation of file system core. Instead, semi-solution was invented: the TRIM command. This command is used to inform SSD firmware which pages can be remapped (because FS knows which pages (LBAs) are free).

3. SSD manufacturers are still experimenting with algorithms - there was many FW failures in not so far past - even OCZ and Intel have had problems.

SSDs are very good solution for netbooks/laptops (extending on-battery time) and for desktops (faster loading of games, etc). For Servers SSD technology is still unreliable, but there are experiments with using SSDs as cache for Databases.

1clue

March 25th, 2011, 03:55 AM

IMO, if you have a laptop and use it on battery, then you can DEFINITELY use a SSD. if you need a lot of storage or you need storage for files which are frequently written to, then consider getting a mechanical drive too.

FWIW the multi-write situation can happen to a mechanical drive too, but considering that at this point huge SSDs are financially unfeasible for many people, it's more likely on an SSD.

SSDs (at least some of them) have an algorithm that attempts to spread writes out across the whole drive evenly so as to prevent "wearing a hole" (my phrase) in the drive. That can be negated though if you have very little room on the drive and you keep writing temp files or whatever to it.

Say you have a small amount of RAM and you continually hit SWAP. Say you have a largely full drive and a small to medium sized database that is frequently written to. You have a YouTube addiction which wears a hole wherever the temp files go. The SWAP is going to wear a hole in the swap file, and the database is probably going to wear a hole where the records go.

In those cases I'm simply saying that the writes can be confined to a small area, and in some cases you can make a LOT of writes to that small area.

So if you're going to use an SSD, make sure it has plenty of free space, and make sure you have a lot of RAM. Enough that you only ever use half of it or so, I've found that Linux automatically caches disk data in otherwise unused RAM.

Now, to help you dispel some of the myths about SSDs.

First, what people who build systems for a living do: Lots of laptops use SSDs. It makes huge sense even if you just look at battery life. Pick a big one.

However, I have yet to see a single piece of Enterprise server hardware that has an SSD at all. You would not expect that if you buy the massive speed increase. Something is making the guys who normally don't pay much attention to price tags say "no" to SSDs.

Next, let's look at limits of the systems we use.

Keep in mind that this is still going through the same disk interface that a mechanical drive would use.

Top speed is going to be that interface speed.
There are drives commonly available that push the sort of interface you find on a normal motherboard.
If you have multiple faster mechanical drives on a single on-the-motherboard controller, then chances are your interface is being saturated.
Better mechanical drives have caches which are large enough to be very similar to an SSD on the second read, and also cache writes. On these even one drive can saturate the disk interface for most reads in a best case scenario.
If you use tmpfs intelligently and have lots of RAM, the only real advantage of an SSD is boot time. If you leave the system up but sleeping, you have effectively negated the benefits of SSD for almost every scenario.

Now, keep in mind some other facts:

If you have lots of RAM, that RAM is used to cache disk access.
RAM is going to be much, much, much faster than an SSD because the SSD is using a disk interface.
Putting tmpfs in strategic places can speed up ANY system significantly.
The same tmpfs can prevent "wearing a hole" in a drive.
While an SSD is extremely rugged, dropping your laptop down the steps usually causes a failure in something like the monitor, not the drive.

I am NOT anti-SSD. I'm anti-stupid-marketing-ploy. A laptop, you definitely want a huge SSD if you can swing it. Anything else you need to really think it through IMO. I've wanted one since I first heard of them, and I CAN afford one, but I have not yet considered the advantages worth the money.

Copper Bezel

March 25th, 2011, 04:17 AM

Stop right there! This can't possibly be correct. I'd say most HDD's definitely surpass three years of usage.

Depends on whether it's a desktop or a laptop. SSDs are far more dependable if they're going to be moved around; a laptop hard drive will be very lucky to survive for two years. Magnetic drives are fine for desktops (but still painfully slow at loading apps.)

Yes, but you KNOW that SDD's have a limited amount of writes to them, while HDD's do not.

Magnetic drives don't have a defined limit. There's still a practical limit, but it's unpredictable.

LowSky

March 25th, 2011, 07:02 AM

I can buy a 7200RPM Sata II 2TB hard drive for under $100, or Maybe get a 60GB SSD for almost $100. Boot time isn't really that bad. If SSD was around 10 years ago it might have made a bigger dent and be more popular than it is now in its short life.

I am looking forward to Hybrid drives. 30-60GB for the OS, and maybe 1TB for data storage would be an amazing data drive in a small package.

Cracklepop

March 25th, 2011, 07:17 AM

However, I have yet to see a single piece of Enterprise server hardware that has an SSD at all. You would not expect that if you buy the massive speed increase. Something is making the guys who normally don't pay much attention to price tags say "no" to SSDs.
There are two reasons for that: the bottleneck is the connection, not the disk, and SSDs cost far more than HDD. Believe it or not, even big companies do care about how they spend their money ;)
If you have multiple faster mechanical drives on a single on-the-motherboard controller, then chances are your interface is being saturated. But for the rest of us, we're waiting on the disk. And you're going to need a lot of fast HDDs to use 100% of sata3's 6Gb/s.
Better mechanical drives have caches which are large enough to be very similar to an SSD on the second read, and also cache writes. ...but not as good.
... the only real advantage of an SSD is boot time. The large majority of people will get a noticeable speed increase, less noise, less heat, lower energy consumption, less susceptibility to physical damage.
While an SSD is extremely rugged, dropping your laptop down the steps usually causes a failure in something like the monitor, not the drive. You are kidding, surely? I personally have killed the HDDs of several laptops with rough handling/drops, but *never* seen a screen fail.
I've wanted one since I first heard of them, and I CAN afford one, but I have not yet considered the advantages worth the money.SSDs have one very large disadvantage: they cost much more than HDDs.

disabledaccount

March 25th, 2011, 07:32 AM

Depends on whether it's a desktop or a laptop. SSDs are far more dependable if they're going to be moved around; a laptop hard drive will be very lucky to survive for two years. Magnetic drives are fine for desktops (but still painfully slow at loading apps.)HDDs are failing due to manufacturing-time defects, Firmware bugs, PSU failures and user actions, such as dropping, kicking (yes kicking - PC that crashed :) ), extreme temperatures and similar. Assuming that drive has no factory or firmware defects and is properly maintained it can work for many, many years. There is no write limit for magnetic surface - unless head becomes damaged.

Copper Bezel

March 25th, 2011, 07:50 AM

Every device still has a limit on a long enough time scale, even if it's impossible to reach for other reasons. = ) But yes, I understand that magnetic drives generally die because they've been bashed about, which is precisely what I said makes them impractical for portables and precisely the reason I can't understand why manufacturers go on putting them in laptops and especially netbooks, where they're doomed to be the first component to fail. (Of course, ideally, a netbook should be using integrated flash memory, anyway.)

I actually have a brother who killed one by kicking. = ) For me, my last (ever) magnetic drive died after a year of good service, being, after all, toted around like a security blanket.

Boot time isn't really that bad.

It's also launch time for applications. I could care less about boot time with a little effort, but responsive launching is a necessity to me.

If SSD was around 10 years ago it might have made a bigger dent and be more popular than it is now in its short life.

Flash memory was around 10 years ago, but it was uselessly limited and has improved dramatically in recent years. The only new game in town that seems likely to replace it is integrated flash memory, as in the current-gen Macbook Airs, or hybrid flash/magnetic drives as you mentioned yourself. Since an SSD is just flash memory in a hard drive slot, there's no point in making a distinction. You're right that SSD is a transitional standard, but flash memory itself is likely to become ubiquitous.

johntaylor1887

March 25th, 2011, 07:53 AM

I put a 60gb SSD in my netbook. Apps open insanely fast and battery life is a bit better.

wizard10000

March 25th, 2011, 09:00 AM

Keep in mind that this is still going through the same disk interface that a mechanical drive would use.

- Top speed is going to be that interface speed.
- There are drives commonly available that push the sort of interface you find on a normal motherboard.

Maximum read speed (once the onboard cache is empty) will be sectors per track * platter rpm * 512 bytes for a sequential read. SATA 2.0 is twice as fast as any current mechanical hard drive can sustain a data transfer. After factoring in encoding losses 3.0Gbps SATA should be able to sustain a transfer rate of 300mb/s. That's more than twice as fast as the best modern hard drive.

- If you have multiple faster mechanical drives on a single on-the-motherboard controller, then chances are your interface is being saturated.

It's a serial interface - it can only talk to one drive at a time ;)

This is why SATA supports command queueing. Only way I can see this happening with a mechanical drive is with a RAID array with more than two drives - the fastest SSDs can get close to that limit with a single drive.

- Better mechanical drives have caches which are large enough to be very similar to an SSD on the second read, and also cache writes. On these even one drive can saturate the disk interface for most reads in a best case scenario.

Yes, but with my 1TB drive with a 32mb cache that perfect cache hit would saturate the interface for about a tenth of a second ;)

- If you use tmpfs intelligently and have lots of RAM, the only real advantage of an SSD is boot time. If you leave the system up but sleeping, you have effectively negated the benefits of SSD for almost every scenario.

IMO in Linux preload would probably serve you better than tmpfs, at least for application load time.

Both of these assume a cache hit considerably more frequently than is possible in real life.

Yup. Great minds think alike :D

I'm anti-stupid-marketing-ploy.

Best comment in the post. I liked that one ;)

My desktop system is a Core i7-920 with about a 20% overclock and a fast 1TB drive - 7200 rpm, dense platters and a 32mb cache. Sustained transfers are about 100mb/sec and cache reads are about 100 times that fast. I installed a reasonably fast SSD (64GB Corsair Extreme X64) about a year ago. It did cut boot time in half but here are some hard numbers. /dev/sda is the SSD and /dev/sdb is the 1TB mechanical drive -

wizard@wizard-desktop:~$ sudo hdparm -tT /dev/sda
[sudo] password for wizard:

/dev/sda:
Timing cached reads: 19744 MB in 2.00 seconds = 9881.78 MB/sec
Timing buffered disk reads: 430 MB in 3.01 seconds = 143.06 MB/sec
wizard@wizard-desktop:~$ sudo hdparm -tT /dev/sdb

/dev/sdb:
Timing cached reads: 21276 MB in 2.00 seconds = 10649.57 MB/sec
Timing buffered disk reads: 296 MB in 3.01 seconds = 98.49 MB/sec
wizard@wizard-desktop:~$

Now let's do some write tests. First results are the SSD, second are the mechanical drive.

wizard@wizard-desktop:~$ sudo dd if=/dev/zero of=/tmp/output.img bs=8k count=256k && sudo rm /tmp/output.img
262144+0 records in
262144+0 records out
2147483648 bytes (2.1 GB) copied, 54.9821 s, 39.1 MB/s
wizard@wizard-desktop:~$ sudo dd if=/dev/zero of=/home/wizard/temp/output.img bs=8k count=256k && sudo rm /home/wizard/temp/output.img
262144+0 records in
262144+0 records out
2147483648 bytes (2.1 GB) copied, 23.5096 s, 91.3 MB/s
wizard@wizard-desktop:~$

My SSD is advertised 220mb reads and I've never been able to achieve that in the real world. Real-world speed is around 2/3 of that.

Anyway, interesting topic. Just thought I'd share ;)

piquat

March 25th, 2011, 09:03 AM

Just a few years ago, you might have something to worry about.

When I was reasearching my purchase (90G OCZ VertexII) I quickly learned that you can't really trust anything you read that's over about 1 to 1.5 years old. It's changed THAT MUCH in that time.

Biggest problems for SSD's today is price and the bad rep they've earned. Because of those two things, adoption is slow.

disabledaccount

March 25th, 2011, 09:29 AM

It's a serial interface - it can only talk to one drive at a time ;)

This is why SATA supports command queueing. Only way I can see this happening with a mechanical drive is with a RAID array with more than two drives - the fastest SSDs can get close to that limit with a single drive.Not quite accurate - every chipset and many controllers uses internal SATA splitters - You have 2 (or more) SATA ports for each single SATA controller. Due to many limitations single controller usually can't exceed about 300-400MB/s (I've seen newest Intel chipsets maxing at 450, but it also depends on MB, RAM speed, etc.) - so integrated controllers are unable to fully utilize SATA2 troughoutput (edit: with 2 or more drives per physical channel)

NCQ has nothing to do with SATA. NCQ is possible on every Interface that supports set of NCQ commands. Prallel SCSI was the first interface with NCQ support - real support, not marketing - like in case of first SATA2 drives.

hdparm is testing sequential read - and in such test SSD wont show its power. Use seeker to measure average seek time - this is the key for random access performance.
Sequential transfers almost never happens in real life.

When I was reasearching my purchase (90G OCZ VertexII) I quickly learned that you can't really trust anything you read that's over about 1 to 1.5 years old. It's changed THAT MUCH in that time.

Biggest problems for SSD's today is price and the bad rep they've earned. Namely, what exactly changed THAT much?

Biggest problem was (IS) lack of support for TRIM command in many SSDs, mostly those with "internal raid0 architacture" (marketing again). Secondary problem was (IS) implementation of support for TRIM in many filesystems/OSes. But there is also secret true about TRIM and SSD - there are already filesystems that are Flash-friendly, the only problem for manufacturers is that winblows is not capable of using them ... (they dont have to implement TRIM - in fact hardware could be much cheaper and FW less buggy)

Copper Bezel

March 25th, 2011, 09:54 AM

Namely, what exactly changed THAT much?

Write limits, that is, lifespan, as I understand it (although the rest of this is way above my reading level.)

Grenage

March 25th, 2011, 10:06 AM

disabledaccount

March 25th, 2011, 10:20 AM

Write limits, that is, lifespan, as I understand it (although the rest of this is way above my reading level.)Nope, write number limits and data retension haven't icreased - still 100K writes, 10years first write retension at 20deg. C - higher chip temperature lowers both limits. That's silently omitted information - data retension will render device unusable much earlier than number of succefull writes.

Cracklepop

March 25th, 2011, 10:39 AM

lz1dsb

March 25th, 2011, 10:46 AM

ikt

March 25th, 2011, 10:48 AM

don't SSD's use flash? And don't flash devices eventually lose the ability to be written to? Why buy something if you know it will DEFINITELY break in the future, where as with HDD's, it can be hit and miss (although I believe some HDD's in my house have lasted over 10 years). Thoughts?

FUD/myth: http://www.storagesearch.com/ssdmyths-endurance.html

Johnsie

March 25th, 2011, 10:58 AM

I think many people assume that the data life on SSD's is short. This is a very common misconception that I am yet to see proved in the real world. It takes an awfully long time to lose data on a flash drive. I've had some usb sticks since 2005 and they are still going strong. I had an eeepc that I got in 2007 and apparently that is still going strong too (I gave it to someone when I got an atom powered device). From my experiences with ssd data I am confident that it is a reliable medium, as long as you buy the right products.

Cracklepop

March 25th, 2011, 11:08 AM

Nope, write number limits and data retension haven't icreased - still 100K writes, 10years first write retension at 20deg. C - higher chip temperature lowers both limits. That's silently omitted information - data retension will render device unusable much earlier than number of succefull writes.

I think you need to read JESD22-A117 and JESD22-A108.
Retention time should be roughly inversely proportional to cycles, so degrading retention time is *not* going to kill an SSD before write cycles do.

Also an interesting paper here (https://docs.google.com/viewer?a=v&q=cache:STvAxbREp1sJ:www.imation.com/PageFiles/83/SSD-Reliability-Lifetime-White-Paper.pdf+&hl=en&pid=bl&srcid=ADGEESh5WaOsHZa-R3kO2QPCr9UcNScyBZwv8_VE1BWrS38KoJR80WndUVsAw25w42 SNzV2ZV93v137BOizCku2Wrfl_FP8BVgH3qGpG3B6Ws1rqhiYP UmhNK0wC_s1Po1jpMCco4YCR&sig=AHIEtbRB0VLQ8MsEq4fnXksCM7WCtQE_0A).

Here's the table from the end of it, showing expected lifetime in years of a 32GB SSD rated for 100k cycle depending on the GB of sequential data written per day:

GB/day Years of life
3.2 2366
10 757
32 237
100 76
320 24 Looks like even with a tiny 32GB SSD I can easily expect it to outlast my computer, and probably even my own life! ;)

Paqman

March 25th, 2011, 11:11 AM

Once the new Intel batch come out, I'll be buying a few. SSDs are worlds apart from traditional hard drives, and the newer models are great.

They aren't viable for mass data storage, but you don't normally need speed for archived data.

Exactly. The sweet spot is to put your OS on a smallish SSD then do your bulk storage on magnetic drives. I switched a couple of machines to an internal SSD and have all my data on a RAID1 NAS. I love it. The idea of struggling along with a 5400rpm hard drive on a netbook fills me with dread.

NightwishFan

March 25th, 2011, 11:31 AM

I think in the future this is the way I will go. To have a trim Ubuntu system on an SSD and keep my bulk data on a large external (or another internal for desktops).

wizard10000

March 25th, 2011, 11:43 AM

Not quite accurate - every chipset and many controllers uses internal SATA splitters - You have 2 (or more) SATA ports for each single SATA controller. Due to many limitations single controller usually can't exceed about 300-400MB/s (I've seen newest Intel chipsets maxing at 450, but it also depends on MB, RAM speed, etc.) - so integrated controllers are unable to fully utilize SATA2 troughoutput (edit: with 2 or more drives per physical channel)

I think this is a more technical explanation of what I just said :D

NCQ has nothing to do with SATA. NCQ is possible on every Interface that supports set of NCQ commands. Prallel SCSI was the first interface with NCQ support - real support, not marketing - like in case of first SATA2 drives.

I worded some of that rather poorly. I didn't mean to infer that NCQ was a SATA-only thing.

hdparm is testing sequential read - and in such test SSD wont show its power. Use seeker to measure average seek time - this is the key for random access performance.
Sequential transfers almost never happens in real life.

True dat. I'm at work now and can't test the same drive but let's see what happens with the 500gb drive in my netbook - it's got fairly dense platters.

Drive info: SAMSUNG Spinpoint MP4 HM500JJ 500GB 7200 RPM 16MB Cache 2.5" SATA 3.0Gb/s Internal Notebook Hard Drive

wizard@wizard-netbook:~$ sudo seeker /dev/sda
[sudo] password for wizard:
Seeker v2.0, 2007-01-15, http://www.linuxinsight.com/how_fast_is_your_disk.html
Benchmarking /dev/sda [476940MB], wait 30 seconds..............................
Results: 63 seeks/second, 15.80 ms random access time
wizard@wizard-netbook:~$

Samsung's drive specs -

http://www.samsung.com/global/business/hdd/productmodel.do?group=72&type=94&subtype=99&model_cd=517&tab=fea&ppmi=1087

Samsung says average seek is 11ms when seeker says it's closer to 16ms. I'd tend to believe seeker ;)

Samsung also says max disk to buffer transfer rate is 189mb/s to which I say *coughbullshitcough*. Still trying to find sectors per track information but I'd say that transfer rate is impossible given the rotational speed and sector density of a 500gb, 7200rpm two-platter drive.

Biggest problem was (IS) lack of support for TRIM command in many SSDs, mostly those with "internal raid0 architacture" (marketing again).

I'd never heard that one. That's great :D

I googled it and found this - it is interesting, and talks about using a pair of JMicron SSD controllers, each addressing half the DRAM in the drive.

http://benchmarkreviews.com/index.php?option=com_content&task=view&id=290&Itemid=60&limit=1&limitstart=3

The OCZ Apex SSD uses two of these SSD control chips, which combine efforts and connect into the JMB390 RAID controller.

Interesting.

disabledaccount

March 25th, 2011, 01:35 PM

I think you need to read JESD22-A117 and JESD22-A108.
Retention time should be roughly inversely proportional to cycles, so degrading retention time is *not* going to kill an SSD before write cycles do.
Google article is explaining this a bit: every write cycle damages gate insulation a bit - so its ability to hold the electrical charge vanishes over the time. With such big number of cells this is not linear, but rather discrete process.
Samsung Note:

NOTE :
1. The device may include initial invalid blocks when first shipped. Additional invalid blocks may develop while being used. The number of valid blocks is
presented with both cases of invalid blocks considered. Invalid blocks are defined as blocks that contain one or more bad bits. Do not erase or pro-
gram factory-marked bad blocks. Refer to the attached technical notes for appropriate management of initial invalid blocks.
2. The 1st block, which is placed on 00h block address, is guaranteed to be a valid block at the time of shipment.
3. The number of valid block is on the basis of single plane operations, and this may be decreased with two plane operations.
Data retention time in relation to one cell is completely different thing from from data retention of whole page or chip. Cells are dying randomly during normal usage period - this, plus random (natural) errors generated during normal operation lead to the point where ECC algorthms are unable to recover the data. In other words, page becomes unusable when number of randomly damaged cells + number of random read errors exceeds recovery possibilities, not when all cells becomes damaged.

edit: maybe i should say it clearly: I don't want to start some HDD vs SSD war - I'm using SSDs - just saying that this technology is still far from beeing perfect.

Cracklepop

March 25th, 2011, 02:08 PM

Google article is explaining this a bit: every write cycle damages gate insulation a bit - so its ability to hold the electrical charge vanishes over the time. With such big number of cells this is not linear, but rather discrete process.
Sorry, I wrote that badly. It won't occur linearly, but the limit of what is acceptable is linear.
For most of it's life an SSD's UBER will be far below the linear limit, until near end of life when it increases exponentially and approaches the linear limit.

The simple fact is that an SSD can very easily outlive the useful life of a computer (and quite possibly it's owner).

disabledaccount

March 25th, 2011, 03:26 PM

CF, pendrives and first SSDs are prooved to die in relatively short time, comparable to or less than classic HDD drive. Newest SSDs are different things - theoretical assumptions and quick tests often are far from real-life resulsts. We'll see.

Lightstar

March 25th, 2011, 04:12 PM

My netbook is ssd. One of the first asus eee pc that came out. It's practical since it moves around alot.
But I wouldn't bother paying $200 for a desktop ssd to get 120gb when i can have a 2TB for $100 hdd.

My life is not in such a rush that I need to save 30 seconds on boot time for so much money.

Lucradia

March 25th, 2011, 06:20 PM

It's too bad that high-capacity SSDs are so expensive :<

http://www.newegg.com/Product/Product.aspx?Item=N82E16820227515

Then again, I barely get past 100 GB on my own windows 7 install (857 GB free of 917 GB) on a non-SSD. I'd probably get one of these: http://www.newegg.com/Product/Product.aspx?Item=N82E16820139110

luceerose

March 25th, 2011, 06:44 PM

Because I don't have 25 seconds worth of patience to wait for the operating system to load.

1clue

March 25th, 2011, 07:30 PM

There are two reasons for that: the bottleneck is the connection, not the disk, and SSDs cost far more than HDD. Believe it or not, even big companies do care about how they spend their money ;)

Bottleneck is the connection is my whole point. However, if spending an extra $1000 on a system that already costs $12,000 is going to provide some significant improvement then big companies won't bat an eye at the price. Or take it further, if you're talking about something really big then $1000 is a drop in the bucket. Which means that the numbers the starry-eyed SSD proponents are quoting can't be that much better, or there is some extenuating circumstance.

But for the rest of us, we're waiting on the disk. And you're going to need a lot of fast HDDs to use 100% of sata3's 6Gb/s. ...but not as good. The large majority of people will get a noticeable speed increase, less noise, less heat, lower energy consumption, less susceptibility to physical damage.

If you're getting good drives then the cache built into the drive is large enough to absorb most of any speed increase, and once they have it cached then it's in RAM and there's no SSD that can come close to competing with that. So the speed increase is irrelevant because it will only ever get used once. Less noise, true. Less power, true. Less heat, probably. Physical damage, that's true but IMO irrelevant because other things will break first.

You are kidding, surely? I personally have killed the HDDs of several laptops with rough handling/drops, but *never* seen a screen fail. SSDs have one very large disadvantage: they cost much more than HDDs.

We have a drawer with a bunch of dead laptops. I think there was one dead HDD (from a system that sat on a desk for years) and the rest are screens. Either the hinge breaks first and then causes the case to bend around and fracture the screen or the screen cracks from an impact right on the flat part. If the screen went, we plug the drive into another system and pull the data off.

What do your guys do, get the drive accessing and then throw them around like a frisbee?

ssam

March 25th, 2011, 07:49 PM

might get one when bcache becomes a bit more stable.

then you can assign the SSD to cache your HDD, so you get the speed of an SSD for you commonly used files, and the space of a HDD.
http://bcache.evilpiepirate.org/

jennybrew

March 25th, 2011, 08:38 PM

Because I don't have 25 seconds worth of patience to wait for the operating system to load.

Just love that :D

:lolflag:

andrewabc

March 25th, 2011, 10:03 PM

Lots of misinformation in this thread. I guess SSD are new to some.

Once the new Intel batch come out, I'll be buying a few. SSDs are worlds apart from traditional hard drives, and the newer models are great.

They aren't viable for mass data storage, but you don't normally need speed for archived data.

New Intel are already out (http://www.newegg.ca/Product/Product.aspx?Item=N82E16820167042) (although they are really marvell). You can get 120gb for ~$300. New OCZ Vertex will be out in a week with low inventory.

If you have desktop, then yes SSD are really great. This is because you can still have HDD installed as well for data storage. So the whole "not enough storage room on SSD" argument not valid on desktops.

I bought first 60 OCZ Vertex1 October 2009 for $200. I bought another of exact same model ~June 2010 for $140. They make huge improvements in speed.

I bought new custom built computer in November 2010. I have my 2x60gb vertex in it, both working fine (not in RAID, one has OS/apps, the other games). I have 1tb caviar black and 1tb caviar green for storage needs (+another 1tb green in external enclosure).
These old SSD work great, big difference in OS load speed, app startup and responsiveness (no more hdd thrashing while multitasking), games load faster (load saves/areas faster).

Use SSD for OS/apps, and HDD for storage.
If on laptop you can still get SSD and have 2.5" external HDD enclosure to store everything. What SSD to get for laptop depends on whether it can do SATAII or not. Size also depends on what you use laptop for. If laptop is your secondary computer, in the sense that you have a desktop at home for storage, then SSD is good. If you only have a laptop, then I'd stick with default HDD.
When it comes to laptops and say HDD dies and you need replacement, good 500gb 7200rpm HDD costs ~$80. For twice the price you can get 90gb or so (in couple months hopefully 120gb for under $200.) Now if you can fit all your OS/apps/important data on 80gb, and the rest somewhere else (home desktop, external enclosure), and laptop has SATAII, SSD is a no brainer, as the speed will be 2-3 times faster. I've seen so many laptops with decent specs (for basically web browsing and simple tasks), yet are very slow due to ****** HDD (80-320gb 2.5") that only get 40-80mb/s and hdd is constantly thrashing.

Now with SATAIII and SSD getting 500mb/s read, I don't see why anyone on desktop with SATAIII would want a HDD that gets at most 150mb/s read as primary OS/app device.

Yes SSD are expensive, but if you are at all wanting speed, you need it.

I'll be upgrading to SATAIII SSD at some point (end of year maybe?), and I'll put the 2009 60gb vertexes in other devices (nettop, laptops, older web browsing desktops), as they bring new life to these devices.

Benchmark comparisons I made.
On my eeebox 160gb hdd vs 60gb vertex limited to SATAI. The SSD is clearly minimum twice as fast as hdd.
http://img442.imageshack.us/img442/6176/screenshot160gbharddisk.th.png (http://img442.imageshack.us/i/screenshot160gbharddisk.png/)

http://img175.imageshack.us/img175/4657/screenshot64gbsolidstat.th.png (http://img175.imageshack.us/i/screenshot64gbsolidstat.png/)

On old desktop of mine (SATAII):
3.5" 320gb 7200rpm hdd vs 60gb vertex.
My 320gb 7200RPM gets
min 37.2
max 78.5
avg 63.9
access time 13.6 ms

http://img190.imageshack.us/img190/4657/screenshot64gbsolidstat.th.png (http://img190.imageshack.us/i/screenshot64gbsolidstat.png/)

SSD is 3 times faster in sequential read.

My 1tb 5400rpm WD green
http://img101.imageshack.us/img101/7830/screenshot10tbharddiska.th.png (http://img101.imageshack.us/i/screenshot10tbharddiska.png/)

SSD twice as fast.

And this is with an SSD from May 2009. They've since come out with faster vertex2 and brand new vertex3.

In summary:
1. Don't worry about nand write failure, unless you are writing >5gb every single day (you'd still be safe on 10gb/day or so depending on ssd size). Obviously should not be used for those types of work (non stop torrents etc). SSD have 3 year warranty, more likely to fail for reasons other than write fail.

2. DO RESEARCH. Make sure you know everything about SSD you are looking at. Read reviews. Anandtech has written many articles about how SSD work, and benchmarks of most stuff. In benchmarks HDD come nowhere near SSD in any benchmark. 2-3 years ago they were close/better in sequential write, but now SATAIII has eliminated that.

3. Hybrid drives are currently no good. If you have laptop, do yourself a favor and either keep HDD that comes with laptop, or get SSD. Don't get a hybrid drive as you will not get big benefit of speed.

4. SSD can fail, just like HDD can fail. Who knows the reasons all the time. I always see reviews of people getting SSD or HDD and they come dead. SSD is no miracle in that there are never failures.

5. Laptop HDD fails? You only use <100gb of important stuff (OS/apps/data used on daily basis)? Then get SSD. Don't bother with HDD replacement. If you find HDD a bit slow and want some more speed. Why spend $x on 2.5" HDD when you can spend twice as much on SSD and get maybe 3 times speed increase? As long as laptop not primary computer, and no access to storage (online/external enclosure/desktop).
Example is if you use laptop just for portability and only need OS/apps/web browsing/ then even a 40gb SSD for $100 should make a nice difference as the read speeds are quite fast. Of course mostly depends on if you have SATAI(120mb/s) or SATAII(250mb/s) on laptop. If SATAII, then definitely get SSD. 2.5" HDD barely saturate SATAI. Slow 5400rpm 2.5" make multitasking difficult, and you get to listen to hdd thrash constantly.

I've been doing hobby research on SSD since April 2009. Lots to learn, and good to know about history of SSD, as some people in this thread are saying things that are no longer valid concern about SSD.

Tip: With vertex 2 OCZ screwed up and made them slower than originally, so have to watch out for that, and normally the larger the SSD drive the faster it is. Usually 120gb SSD is around fastest (sometimes bigger than that is faster, but not as many can afford it anyway).
If vertex3 120gb starts selling for $300, and if eventually they come out with 60gb version, as long as they can keep read speeds real high (500mb/s sequential) and also random read/write (I don't care about sequential write, as long as not slower than mid range HDD), and sell for $150, could time to buy into SSD.

The new Intel look good, especially if their quality control is good. I don't suggest older intel 40gb models as write speeds are only like 40mb/s sequential. And have slow random read/write.

Sammi

March 25th, 2011, 10:25 PM

andrewabc speaketh the truth.

Welly Wu

March 25th, 2011, 10:55 PM

I own an ASUS N61JV-X2 notebook PC. I upgraded it to include Crucial 8.00 GB of DDR3 PC-8500 SODIMM SDRAM and an Intel 2nd Generation 34nm 160.00 GB Solid State Drive. I installed Ubuntu GNU/Linux 10.10 Maverick Meerkat 64 bit on it.

I can cold boot in 5 seconds, reboot in 2 seconds, and shut down in 3 seconds. I recently created a 50.00 GB virtual disk in Sun/Oracle Virtualbox for Microsoft Windows 7 Professional 32 bit and it only took 4.00 minutes to complete the task.

I cannot afford to waste my time waiting for tasks to finish on my notebook PC. I am a graduate student at New Jersey Institute of Technology in Newark, New Jersey and I am studying for a Masters of Science in IT Administration and Security degree program. The courses are challenging and the work is intense and it is very fast paced.

I am also impatient so purchasing my Intel SSD was the right choice for me.

In short, my laptop is future proof for the next several years especially given the fact that I chose to switch to Ubuntu GNU/Linux which will prolong the usefulness of my computer and I am no longer locked into the planned obsolescence of the Microsoft Windows platform.

andrewabc

March 25th, 2011, 11:11 PM

andrewabc speaketh the truth.

Thanks.

I'd say if you got dual core 1.8ghz (what my old desktop is), or higher, an ok vid card (mine had GMA965 x3000) or higher, and find desktop/laptop a bit slow and don't mind spending $100 to speed it up (even for web browsing/app speed increase), go for an SSD. I've seen dual core laptop with intel GMA g965x3000 vid card, 1gb ram, but came with 80gb 2.5" hdd. Slow as hell. But only because of HDD. I've seen this on pretty much all cheap laptops. The HDD is slowest part of it by far. Kinda hard to open firefox/openoffice while doing something that involves writing to disk when top speed is <1mb/s random read/write. Firefox database is mostly random read/write. Same goes for most cookies/cache (<1mb files), hdd can't handle it.

If you already got desktop, then you just add SSD inside and keep HDD hooked up. SSD for OS/apps, and old HDD for data storage. I even got firefox downloads going onto HDD as there is no speed increase to read a pdf file or program installer on SSD vs HDD (ok there is, but how often you use program installer? just once until new version out, so don't even bother putting installer on SSD, just the installed program).

For laptop, crazy to buy HDD for it. Buy SSD instead. Only reason for HDD is if you absolutely need >100gb of stuff with you at all times and not willing to use external enclosure or internet storage.
Heck if your HDD on laptop works fine, and replace with SSD, you can still put HDD in external enclosure (they sell for $30), and use for storage, hook up via usb. Put external enclosure in laptop bag if needed (although if no desktop at home, maybe leave external enclosure at home and at end of day can transfer data back and forth between SSD/HDD). Say want to watch tv/movie while away for day or plane trip? Just put the 2gb worth of video on SSD for trip, and watch.

In 2009/2010 or so I remember people freaking out about nand write failure on SSD and how to 'prevent' it and they were putting firefox profiles and swap and other stuff on HDD, which completely negates the usefulness of SSD being fast.

HDDs will *definitely* break at some point too...

SSD are quieter, faster, cooler, consume less energy for the same amount of data written/read, and have no moving parts to break.

True: Quieter, faster, consume less energy, no moving parts.

False: Cooler. SSD can get hot. In one review they found SSD to have similar temps as hdd. Can't find specific review but I know for sure that was their conclusion.

Yes, but you KNOW that SDD's have a limited amount of writes to them, while HDD's do not. I dunno, I guess I just can't get over that.

But I guess if solid state drives really last 171 years, there's nothing to worry about. ;)
You should get over it. :)
I've had 60gb vertex since oct 2009 and I think around 50% health (not sure what exactly means, but I don't see it as problem.)
As long as you don't plan on non stop write 5gb daily on 120gb drive, then you shouldn't worry. Newer drives using newer/smaller nand have less writes allowed before failure, which does make some nervous, but for normal usage scenarios and you try to avoid constantly writing, shouldn't be a worry. Use HDD for large files/storage.

1. SSDs are fast - in general, but there are many exceptions.
Only latest medium to high-end SSDs with built in cache are faster than HDDs, especially in writing. Older and low-end SSD are slower or sometimes even deadly slow in comparison to new HDDs.

2. SSDs can last for 1.5 million of hours - FALSE: marketing or just lie (pick one)
Nand Flash data retention guaranteed by most manufacturers reaches up to 10 years for first written data -> every next write is shortening data retension, and first write errors can appear even after several thounsands of writes <to single flash page>
Thanks to speciall agorithms that are spreading writes to different flash pages (by dynamically remapping LBA) this weakness of Flash could be reduced to considerably low level. However, there is a catch: If the "disk" is nearly full, the spreading of writes becomes ineffective - and in some heavy write load conditions it's easy to wear several pages in several hours. The problem is, that SSD firmware cannot remap used pages, beacause it would need full implementation of file system core. Instead, semi-solution was invented: the TRIM command. This command is used to inform SSD firmware which pages can be remapped (because FS knows which pages (LBAs) are free).

3. SSD manufacturers are still experimenting with algorithms - there was many FW failures in not so far past - even OCZ and Intel have had problems.

SSDs are very good solution for netbooks/laptops (extending on-battery time) and for desktops (faster loading of games, etc). For Servers SSD technology is still unreliable, but there are experiments with using SSDs as cache for Databases.

1. True, at first SSD were crap. You had to watch out for what you bought. 1st gen Intel and Indilinx drives mostly solved this problem in 2009.

2. True. I would not suggest anyone go past 80% full capacity of SSD. I forget how much anandtech said, but around this amount.

3. True. There have been firmware problems. My vertex has probably had 10-15 firmwares released since early 2009 (I only upgraded the once about 6 months ago). Intel 2nd gen had firmware problem that delayed release by 1 month. OCZ still releasing firmware for Vertex2 sandforce drives (if you have these drives make sure to update, just released new one this week!).

As for servers and such, if your server requires high IOPS then SSD are best option.
Old article on why SSD better than HDD when it comes to IOPS
Solid State Disk will change the storage world… (http://virtualgeek.typepad.com/virtual_geek/2009/10/solid-state-disk-will-change-the-storage-world.html)
$ per IOPS SSD crush HDD.

I can buy a 7200RPM Sata II 2TB hard drive for under $100, or Maybe get a 60GB SSD for almost $100. Boot time isn't really that bad. If SSD was around 10 years ago it might have made a bigger dent and be more popular than it is now in its short life.

I am looking forward to Hybrid drives. 30-60GB for the OS, and maybe 1TB for data storage would be an amazing data drive in a small package.

On my old desktop:
SSD: 20 seconds to grub, 18 seconds to desktop = 38 seconds total
HDD: 65 seconds from power to desktop
Firefox no longer had hdd thrashing when typying into awesomebar. Awesomebar appeared instantly and searching was instant.

Hybrid drives currently not good, but in future maybe. Only useful application would be laptop with no access to any other storage device.

markp1989

March 25th, 2011, 11:22 PM

right now I have ssds in 2 of my desktops , one is about 2 years old, one of the original ocz vertex 32gb drives (paired with 2 1.5tb Sata drives) and the one in my desktop is a super talent 64gb SSD.

right now, my laptop has the hard drive that was installed in the factory, I had to replace the hard drive (free under warranty) as it has 1365 bad sectors on it! resulting in it being impossible to read/write the drive

Im seriously considering taking the 32gb SSD out of my htpc and putting it in the laptop for battery life and durability reasons, will probably do it when 11.04 is released (as i will be reinstalling the OS again anyway)

I brought mine at the time for noise and speed reasons, as it was for desktops I didnt need to worry about power usage/durability

andrewabc

March 25th, 2011, 11:39 PM

Im seriously considering taking the 32gb SSD out of my htpc and putting it in the laptop for battery life and durability reasons, will probably do it when 11.04 is released (as i will be reinstalling the OS again anyway)

I brought mine at the time for noise and speed reasons, as it was for desktops I didnt need to worry about power usage/durability

That is good idea. Sometime in May put vertex 30gb in laptop. I dunno if your desktop support SATAIII, but if it did put SATAIII SSD in it. Dunno speed of your supertalent (I'm gonna assume indilinx controller based upon google), then new SATAIII drive, even if limited to SATAII should run good. I know you won't go back to HDD only for desktop. Especially if the corei7 rig. Such a waste using hdd only on really fast desktops.

EDIT:
you have 32gb vertex in HTPC? what software you use? ubuntu+xbmc? Might still be good to have cheap/small SSD even for HTPC. Although small wait for os/apps to load isn't big deal if just watching video.

markp1989

March 25th, 2011, 11:49 PM

That is good idea. Sometime in May put vertex 30gb in laptop. I dunno if your desktop support SATAIII, but if it did put SATAIII SSD in it. Dunno speed of your supertalent (I'm gonna assume indilinx controller based upon google), then new SATAIII drive, even if limited to SATAII should run good. I know you won't go back to HDD only for desktop. Especially if the corei7 rig. Such a waste using hdd only on really fast desktops.

EDIT:
you have 32gb vertex in HTPC? what software you use? ubuntu+xbmc? Might still be good to have cheap/small SSD even for HTPC. Although small wait for os/apps to load isn't big deal if just watching video.

32gb in ssd was spare when I updated my desktop, uses ubuntu+xbmc, pc is also used as a fileserver/torrentslave (files stored on 2 1.5tb hard drives) put the ssd in there just to give it a purpose, this was before I brought the laptop.

Cant remember the model SSD i have in the desktop, but I have attached a screen shot of a benchmark.

edit: its this ssd FTM64GX25H http://www.newegg.com/Product/Product.aspx?Item=N82E16820609393&cm_re=FTM64GX25-_-20-609-393-_-Product

Cracklepop

March 26th, 2011, 01:28 AM

Bottleneck is the connection is my whole point. However, if spending an extra $1000 on a system that already costs $12,000 is going to provide some significant improvement then big companies won't bat an eye at the price. Or take it further, if you're talking about something really big then $1000 is a drop in the bucket. Which means that the numbers the starry-eyed SSD proponents are quoting can't be that much better, or there is some extenuating circumstance.
Maybe a large company would pay whatever it takes for a one-off purchase where only the best will do, but 99.9% of the time they weigh up the benefits against the cost, and the cost of SSDs is just too high (exactly why you yourself haven't bought SSDs).

Slightly less on topic: Oracle sells SSD enteprise solutions, so expect them to become commonplace in the near future.
Also I stumbled across this interesting little fact: Oracle advertise their HDDs for 1.5 million hours MTBF, and their SSDs at 2 million hours! (https://docs.google.com/viewer?a=v&q=cache:v4r77AnyuDQJ:www.oracle.com/us/043969.pdf+&hl=en&pid=bl&srcid=ADGEEShJjgz48RJA3NCgdK03snKWSG7_vVWc2AavD_kF BD4I61CfNmuUdHrdYV0rTbtbDgDIcKAzbAbHhXVIG4gXkgtJVv cwojbTKDMVRlKovkQb5ShiHKa46gTD404QIwCN35JTTpWn&sig=AHIEtbRPyfD5xK7qQDMcYTFioRLFDHiw7w) ;) (that's average time between failures before anyone asks)

If you're getting good drives then the cache built into the drive is large enough to absorb most of any speed increase, and once they have it cached then it's in RAM and there's no SSD that can come close to competing with that. So the speed increase is irrelevant because it will only ever get used once. There's some very convoluted logic there.
1. 'most of' would still not be as good.
2. SSDs have buffers too (please stop calling it a cache), so there is no advantage to the HDD
3. data from the SSD goes into RAM too, so there is never going to be a comparison between RAM and SSD.

In summary: unless you have a considerable number of very high speed disks then not only is SSD access time faster and latency lower, but users will get a noticeable, real-life, actual speed increase when going from HDD to SSD.

Less noise, true. Less power, true. Less heat, probably. Physical damage, that's true but IMO irrelevant because other things will break first. Less power 'true' but less heat only 'probably'!?
Think about that carefully for a minute...
SSDs are much, much cooler.

My own experience makes me strongly disagree about less susceptibility to physical damage being irrelevant, but obviously there is no data to support either argument. Perhaps other UF readers could tell us what has failed first in their laptops due to movement/violence: disk or screen?

Cracklepop

March 26th, 2011, 02:47 AM

True: Quieter, faster, consume less energy, no moving parts.

False: Cooler. SSD can get hot. In one review they found SSD to have similar temps as hdd. Can't find specific review but I know for sure that was their conclusion.

Perhaps a basic physics lesson is in order?
SSDs consume less total energy(say, at most one fifth the wattage), AND have higher energy efficiency (no moving parts).
They *cannot* emit as much waste energy (heat), it's physically impossible.

disabledaccount

March 26th, 2011, 09:18 AM

False: Cooler. SSD can get hot. In one review they found SSD to have similar temps as hdd.

Perhaps a basic physics lesson is in order?
SSDs consume less total energy(say, at most one fifth the wattage), AND have higher energy efficiency (no moving parts).
They *cannot* emit as much waste energy (heat), it's physically impossible.
Both of You are right. It's because there's big difference in thermal capacity. If you transfer 1W of heat to f.e. 10g of water it's temperature will raise equally to when transfering 10W of heat to 100g of water.

...Reviews... most unreliable and inaccurate source of information.

Cracklepop

March 26th, 2011, 09:51 AM

Both of You are right. It's because there's big difference in thermal capacity. If you transfer 1W of heat to f.e. 10g of water it's temperature will raise equally to when transfering 10W of heat to 100g of water.

...Reviews... most unreliable and inaccurate source of information.

SSDs are far, far cooler than HDDs. I'm guessing the heat capacities of both are fairly similar.

disabledaccount

March 26th, 2011, 10:20 AM

Cracklepop

March 26th, 2011, 10:48 AM

It depends on model an measuring techniqe. HDD is largely made of alluminium + some mechanical parts, SSD is made of PCB + thin case, mostly made of plastic -> there is really big difference in thermal capacity and thermal resinstance.

If you measure temp with pyrometer, then You can get chip temp inside SSD - it can easily exceed 40deg. C.

Define "far cooler".

There's a lot empty space in disks, and the aluminium platters are thin. The cases are the same. I'll stick my neck out and estimate the difference at definitely less than 20%, probably within 10%.

Anyway, the point is that SSDs are much cooler. Given the energy consumption and efficiency of SSDs the difference in heat capacity in favour of the HDD would have to be absolutely huge for us to see SSDs radiating anywhere near as much heat as HDDs, and:
- without knowing temperatures we can look at both disks and estimate the heat capacities are not that different
- we know from experience that SSDs are much cooler, so we know there is not enough extra heat capacity in HDDs to make up for their high consumption and poor efficiency.

This isn't particularly relevant or interesting. If my answer to andrewabc explaining *why* HDDs aren't as cool as SSDs turns out to be not accurate enough, I'll humbly withdraw it.
The 'why' is not the point.
The point is (again): SSDs are much cooler.

disabledaccount

March 26th, 2011, 11:19 AM

Cracklepop

March 26th, 2011, 11:32 AM

You are messing conceptions.
SSD power consumption is far lower than for HDD's - that is confirmed and measurable - but power consumption has nothig to do with temperature or to be precise - power consumption does not directly affects device temperature.

Peoples usually call something cooler, because it appears to be cooler when they touch it.
Wrong - this way of "measurement" is highly affected by thermal resistance. Peace of wood heated to 60deg. C will appear "warm", while peace of steel at the same temperature will cause thermal shock "wow, it's hot!" - because it has hundrets of times less thermal resistance, so it can transfer heat faster to human's hand.

So HDD can appear "hotter" when you touch it, because it has metal case -> SSDs appear "much cooler" because they have plastic case - thats all. And to avoid further misunderstanding -> it still depends what models are compared.

Quite right, I apologise for my vague usage of ''cooler''. The temperature of the computer around the disk is the only value of interest, thanks to heat transferred from the disk, and it's lower for SSDs.
The SSD in my laptop has a metal case btw, indistinguishable from an HDD.

As for power consumption, it most certainly does affect temperature. All other things being equal, a system with higher power consumption will be hotter than one drawing less power. HDDs consume more energy, and waste more in the form of heat (in both absolute and relative terms) than an SSD.

To repeat the point, a little more accurately: SSD's heat your computer less than HDDs.

andrewabc

March 26th, 2011, 12:36 PM

I found the review stating that SSD are not marketing 'cooler', just around the same temp as hdd.
Heat Output Results (http://benchmarkreviews.com/index.php?option=com_content&task=view&id=299&Itemid=60&limit=1&limitstart=11)

The message here is simple: Although the heat produced by SSD's under load is usually the same as what the Hard Disk Drive generates at idle, Solid State Drives still produce heat. Don't let marketing hype fool you into believing that Solid State Drives are cold-operating devices just because there are no moving parts. Cooler, yes. Cold, no.

Pretty much +-5 Celsius of HDD.

disabledaccount

March 26th, 2011, 01:31 PM

The SSD in my laptop has a metal case btw, indistinguishable from an HDD.
I suppose that Your "metal case" is only thin layer over plastic or just very thin metal case - it looks great, but that's all. That was the reason why I've mentioned about differences between particular models

Cracklepop

March 26th, 2011, 02:57 PM

Sean Moran

March 26th, 2011, 03:05 PM

If you format a flash drive partition, and encrypt it, you can keep all your 'sensitive' educational material in your pocket and not on your computer's hard disk. Keep your computer squeaky clean, and passwords and c/c digits and other educational material for your personal use can stay encrypted on the flash drive.

andrewabc

March 27th, 2011, 12:53 AM

Mm? Did you read what you quoted? SSDs *under*load* run at the same temp as an *idling* HDD.

And NOT '+-': only '-' ('cooler').
SSDs are much cooler than HDDs: cooler than an HDD that never does anything but idle.

Thanks for the correction.

Author also adds some more in a comment:

I've discontinued the measurement of heat output from SSDs. This is because each series of controller operates at a different temp, and the enclosure makes a difference as to how much heat is dissipated out through the walls. In general, SSDs emit a low level of heat, much less than any hard drive, but they do still produce heat.

Guess I misread that article back when it was released.

Cracklepop

March 27th, 2011, 01:31 AM

Guess I misread that article back when it was released.

You're not the first to make the same mistake about that very same article, I remember seeing this same conversation (different people) a year or two ago in this same forum ;)

So, any advances from anyone on: faster, cooler, quieter, more efficient, more rugged?
The benefits of going from moving parts, to none.

All that's left is price - still very high - and longevity, and it seems like although on paper an SSD should have a shorter life than an HDD, we can expect very long lived SSDs, far beyond the average time people keep a computer, and maybe longer than the average HDD life, maybe longer than you or me...

1clue

March 27th, 2011, 06:45 AM

Maybe a large company would pay whatever it takes for a one-off purchase where only the best will do, but 99.9% of the time they weigh up the benefits against the cost, and the cost of SSDs is just too high (exactly why you yourself haven't bought SSDs).

Dude. I'm saying that if the actual performance benefit were as high as you are saying it is, then nearly EVERY enterprise machine would have an SSD -- for a boot drive if nothing else.

Slightly less on topic: Oracle sells SSD enteprise solutions, so expect them to become commonplace in the near future.
Also I stumbled across this interesting little fact: Oracle advertise their HDDs for 1.5 million hours MTBF, and their SSDs at 2 million hours! (https://docs.google.com/viewer?a=v&q=cache:v4r77AnyuDQJ:www.oracle.com/us/043969.pdf+&hl=en&pid=bl&srcid=ADGEEShJjgz48RJA3NCgdK03snKWSG7_vVWc2AavD_kF BD4I61CfNmuUdHrdYV0rTbtbDgDIcKAzbAbHhXVIG4gXkgtJVv cwojbTKDMVRlKovkQb5ShiHKa46gTD404QIwCN35JTTpWn&sig=AHIEtbRPyfD5xK7qQDMcYTFioRLFDHiw7w) ;) (that's average time between failures before anyone asks)

Good for you. You're showing me a single instance of what, if your performance figures were sound, should be overwhelmingly commonplace.

There's some very convoluted logic there.
1. 'most of' would still not be as good.
2. SSDs have buffers too (please stop calling it a cache), so there is no advantage to the HDD
3. data from the SSD goes into RAM too, so there is never going to be a comparison between RAM and SSD.

In summary: unless you have a considerable number of very high speed disks then not only is SSD access time faster and latency lower, but users will get a noticeable, real-life, actual speed increase when going from HDD to SSD.

1. Pay that much for a few seconds of time savings a month? Less than a minute of speed advantage? Why?

2. Disk buffers are not pure FIFOs. If there is a piece of data in the buffer and the computer wants to read it a second time, the buffered data will be sent again without reading from the platter. That makes it a cache regardless of what it's called.

You're evading my point entirely. My point is that first-time reads, you get the speed advantage you're describing, but the SECOND read you get a RAM response with either SSD or HDD due to Linux's caching of anything coming through the disk interface. And if you have sufficient RAM, then any piece of data on the disk will only be read once, unless it's data you're actually operating on.

As well, if you use tmpfs aggressively for your temporary files (which you want to do anyway with SSDs especially) then very little data is actually written back to the disk for most users.

Finally, if you're using a high performance filesystem, writes won't block either, and in general very little of a typical user's work experience involves heavy writes for long enough to make a noticeable lag.

Less power 'true' but less heat only 'probably'!?
Think about that carefully for a minute...
SSDs are much, much cooler.

Less power does not automatically mean less heat. For an extreme example, consider a hot tub in the winter being heated to 103F, compared to a soldering iron designed for modern cell phone circuits. The hot tub heater has thousands of times more power, but never exceeds the boiling point for water. The soldering iron on the other hand can melt tin and lead.

I'm saying "probably" cooler because I don't have specific data. Nor will I any time soon, unless I buy another laptop soon. My desktop machine will get an SSD after they become commonplace on server hardware. Until then I see no significant advantage in an SSD in desktop or server hardware.

My own experience makes me strongly disagree about less susceptibility to physical damage being irrelevant, but obviously there is no data to support either argument. Perhaps other UF readers could tell us what has failed first in their laptops due to movement/violence: disk or screen?

If you get disk damage more than screen, then it doesn't matter at all what my experience is. Same thing goes the other way.

If you really want to believe that your SSD is a thousand times faster than an HDD, then you can believe it if you want. All I'm saying is that I don't see a market response which supports your claims about SSD performance.

The only market that seems to support obvious superiority of an SSD is in battery operated devices.

Copper Bezel

March 27th, 2011, 07:00 AM

Mobile devices, more generally. Again, they're also far more shock resistant because there's no needle to strike the disk on an impact. And, again, the lack of enterprise adoption is a result of the clear disadvantages in storage space and cost, so it doesn't reflect directly on performance (because performance is just one consideration among others.) The performance really could be, as you say, 1000 times faster and not overcome other difficulties.

Cracklepop

March 27th, 2011, 07:25 AM

Dude. I'm saying that if the actual performance benefit were as high as you are saying it is, then nearly EVERY enterprise machine would have an SSD -- for a boot drive if nothing else.
Er, how high did I say it was?
Regardless, the point is (as I've already said) that decisions are made by comparing perceived benefit to perceived cost. The absolute figures are irrelevant: if cost is higher, people won't buy.

Good for you. You're showing me a single instance of what, if your performance figures were sound, should be overwhelmingly commonplace.
What figures?
Give it a few years. It's an incoming technology, but it'll be commonplace soon enough.

1. Pay that much for a few seconds of time savings a month? Less than a minute of speed advantage? Why?

2. Disk buffers are not pure FIFOs. If there is a piece of data in the buffer and the computer wants to read it a second time, the buffered data will be sent again without reading from the platter. That makes it a cache regardless of what it's called.
1. You're being completely unrealistic. 2. The main use of disk buffers is matching the interface speed to the disk speed, not caching.

You're evading my point entirely. My point is that first-time reads, you get the speed advantage you're describing, but the SECOND read you get a RAM response with either SSD or HDD due to Linux's caching of anything coming through the disk interface. And if you have sufficient RAM, then any piece of data on the disk will only be read once, unless it's data you're actually operating on.
RAM is measured in GB, disk buffers only in MB. The chance of a cache miss in RAM *and* having the data waiting in the disk buffer is absolutely minuscule. Lower level caches need to be larger than higher levels to make any real difference. Disk buffers have almost zero effect on system performance due to caching, so please stop going on about it.
And stop ignoring the fact that faster first-time reads is *faster* overall.

I'd prefer hundreds of GB in RAM to a faster disk too, but meanwhile SSDs are a significant speed improvement for a system.

Until then I see no significant advantage in an SSD in desktop or server hardware. That's unfortunate for you. The rest of us will enjoy our higher performance, lower temperatures, less noise, lower consumption, higher efficiency, and higher durability.

If you really want to believe that your SSD is a thousand times faster than an HDD, then you can believe it if you want. All I'm saying is that I don't see a market response which supports your claims about SSD performance. You're putting words into my mouth again...
Where is the 'thousand times faster', and which claims are incorrect exactly?

The only market that seems to support obvious superiority of an SSD is in battery operated devices.You mean the market more than any other where perceived benefit outweighs perceived cost. As prices come down you'll see them everywhere.

1clue

March 27th, 2011, 10:02 PM

Er, how high did I say it was?
Regardless, the point is (as I've already said) that decisions are made by comparing perceived benefit to perceived cost. The absolute figures are irrelevant: if cost is higher, people won't buy.

What figures?
Give it a few years. It's an incoming technology, but it'll be commonplace soon enough.

Looking back, it was others posting performance figures.

1. You're being completely unrealistic. 2. The main use of disk buffers is matching the interface speed to the disk speed, not caching.

1. No, I'm telling you what I get with my home system.

2. That does not mean they don't cache.

RAM is measured in GB, disk buffers only in MB. The chance of a cache miss in RAM *and* having the data waiting in the disk buffer is absolutely minuscule. Lower level caches need to be larger than higher levels to make any real difference. Disk buffers have almost zero effect on system performance due to caching, so please stop going on about it.
And stop ignoring the fact that faster first-time reads is *faster* overall.

And cache hit rates are essentially the same as for CPU caches. And yet the people who design CPUs continue to put caches on them, in fact several layers of caches.

I'd prefer hundreds of GB in RAM to a faster disk too, but meanwhile SSDs are a significant speed improvement for a system.
That's unfortunate for you. The rest of us will enjoy our higher performance, lower temperatures, less noise, lower consumption, higher efficiency, and higher durability.

You don't need hundreds of GB of RAM. I've been using 12G on my system and been up for months, and still had free space in RAM which was not being used as cache. Disk management in Linux is very good, and the longer your uptime the less benefit you get from a first-time read.

You don't have to read the entire filesystem into RAM in order to cache everything you actually use on a regular basis, and if you operate on the same space over and over it doesn't overflow the cache.

All I'm saying is that if you have a sufficient amount of RAM which is extremely reachable on modern systems, and as near as I can tell is cheaper than one of your drives, then the only benefit you'll get in terms of performance is the first-time read.

You're putting words into my mouth again...
Where is the 'thousand times faster', and which claims are incorrect exactly?

You mean the market more than any other where perceived benefit outweighs perceived cost. As prices come down you'll see them everywhere.

The thousand times faster comment was sarcasm, but it obviously didn't come through that way.

Eventually I'll have one too, but not for awhile yet.

debiansu

March 27th, 2011, 10:06 PM

Cracklepop

March 27th, 2011, 10:45 PM

And cache hit rates are essentially the same as for CPU caches. And yet the people who design CPUs continue to put caches on them, in fact several layers of caches.
So what? Notice how the higher level CPU caches are smaller than the ones below them?
You're talking about a rate, not absolute numbers.
The *caching* (as opposed to 'buffering') ability of disk buffers has very nearly zero impact on system performance: when a higher level cache is potentially over 100 times larger than the lower level cache, the lower level cache is going to do squat.

All I'm saying is that if you have a sufficient amount of RAM which is extremely reachable on modern systems, and as near as I can tell is cheaper than one of your drives, then the only benefit you'll get in terms of performance is the first-time read.

I agree, but a benefit is a benefit, and performance is a genuine advantage (just one of a list) that SSDs have over HDDs.
System-wise, for most of us with most of our systems swapping the HDD for an SSD gives a significant boost.

Not 'my' drives. Don't know where you do your shopping, but current prices are about $2/GB for SSDs, and $10/GB for RAM, and even though there's overlap in their function one is not a full replacement for the other.

SSDs are faster than HDDs, and will improve system performance.

disabledaccount

March 28th, 2011, 12:27 AM

The *caching* ability of disk buffers has very nearly zero impact on system performance.You are simply wrong. HDD caches have *huge* impact on performance, especially for writing. It also saves a lot of time under mixed I/O load providing re-read and queuing functionality. Most of I/O transfers have scattered LBAs with relatively small portions of data - cache saves seconds of seeking and makes today's drives fast.
Besides: write cache makes SSD fast too - without it SSDs are getting sloooooow due to write amplification.

You can try disabling HDD caches with hdparm.

Cracklepop

March 28th, 2011, 12:36 AM

You are simply wrong. HDD caches have *huge* impact on performance, especially for writing. You aren't paying attention. Disk buffer's ''buffering'' makes a huge difference to performance, but ''caching'' very little difference - it's very rare for something to be cached in the disk buffer but not in memory.
You'd also notice that we're talking about reading, not writing.

It also saves a lot of time under mixed I/O load providing re-read... So, so rare. I have potentially 4GB (or more) of memory for caching above the disk buffer. The chances of the disk buffer being helpful here are tiny.

...and queuing functionality. Most of I/O transfers have scattered LBAs with relatively small portions of data - cache saves seconds of seeking and makes today's drives fast. Yes: ''buffering''.

Besides: write cache makes SSD fast too - without it SSDs are getting sloooooow due to write amplification. Despite the fact SSDs have write amplification they're still faster than HDDs.
SSDs are not 'getting slow'...they're getting faster.
Also, OCZ claim they have achieved write amplification of 0.5: http://www.anandtech.com/show/2899/3

I'm aware that I'm blurring terms (''buffering'', ''caching''), but go back and read the conversation between myself and 1clue.

Time for a thread-specific sig, I think, because the point keeps getting lost in the fog:

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They are also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they are relatively very expensive.

1clue

March 28th, 2011, 03:49 AM

SSDs fall into the same category as Blu Ray to me; immature and unproven in the long term. I have no desire to save a few extra hundred milliseconds, at the expense of paying for a ridiculously inflated new technology that will likely fair a lot quicker than tech that has been around since the iron age. Blu Ray? have you seen the price of those discs? SSDs - ditto.

No thanks, I'm great with my "slow" sata "winchester" drives :lol:

Blu Ray is, above and beyond anything else, a 50G backup medium. If you want a big backup to a transportable/transferrable medium, Blu-Ray works wonderfully. The drive also reads and writes absolutely any CDROM/DVD format I've ever heard of.

So what? Notice how the higher level CPU caches are smaller than the ones below them?
You're talking about a rate, not absolute numbers.
The *caching* ability of disk buffers has very nearly zero impact on system performance: when a higher level cache is potentially over 100 times larger than the lower level cache, the lower level cache is going to do squat.

I'm not going to quote everything you say, it's both confusing and hugely redundant.

We're talking about several different things here, and we keep going back and forth and crossing topics which makes this discussion nearly pointless.

Performance of the storage device itself.
Performance of the interface (sata/sas/etc)
Performance of the system as a whole.

You mostly talk about the performance of the device--which, I don't argue, is faster on a per-device scale. And you also seem to be talking about mobile devices, which in my first post on this thread I acknowledged is well worth it.

I mostly talk about the performance of the system as a whole, on non-mobile systems. And by my argument, I claim that the device speed difference between SSD and HDD is almost irrelevant for a carefully configured system which rarely shuts down.

I agree, but a benefit is a benefit, and performance is a genuine advantage (just one of a list) that SSDs have over HDDs.
System-wise, for most of us with most of our systems swapping the HDD for an SSD gives a significant boost.

I don't disagree with that, I only say that if you were a bit more careful you could dramatically improve your HDD system in a way that minimizes the difference between SSD and HDD. The way I have configured mine, the SSD really can't save any significant time at all.

Not 'my' drives. Don't know where you do your shopping, but current prices are about $2/GB for SSDs, and $10/GB for RAM, and even though there's overlap in their function one is not a full replacement for the other.

An SSD needs to contain the entire filesystem, including software that is installed but rarely or never used, documents that you stored once but never open, and a whole lot of documentation that most Ubuntu users don't even know is there. It needs to be larger, large enough not only for what you use, but what you don't use and what you MIGHT use. Because of that your price per gigabyte doesn't really apply. How expensive is your SSD compared to 12G RAM?

My /usr partition contains 2.8G. My root partition is less than 1G. All that fits extremely easily in 12G, even running a virtual machine or a game.

I ran Gentoo on this system for quite awhile. Gentoo compiles all installed software right on the system. That includes downloading source files, compiling everything, and then installing it. I ran for an awful long time with that, and still had RAM left which had not been used as disk cache after several months of uptime. Moreover, there's not really much reason for the temporary files used for compiling to ever go over the disk interface. I put all that on tmpfs, and I could watch the disk light stay dark during the entire compilation process and only light up when it installs the app.

FWIW, Linux won't empty out disk cache or tmpfs unless it needs to put something else there. If you have pristine RAM left after a few months of running, your system has never used it for anything at all, and Linux automatically uses all available space for disk cache.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They are also very long lived, although it's too early to claim they live longer than HDDs.
Unfortunately, they are relatively very expensive.

And, once the information comes off it and is stored into RAM, absolutely none of that speed makes any difference whatsoever. On a system sitting on or under a desktop for its whole life, toughness doesn't really matter either. If you allow them to spin down, the power savings is marginal, and disk buffering and high performance filesystems can negate all write lag from spinning the disks back up.

If you're running something like an i7 with a couple 24" monitors on it, spinning disks every now and then is going to be a marginal power drain.

We can talk all day about what COULD happen. Theoretically, my system could suffer huge delays for lack of an SSD. In practice, I need to wait a few seconds when the system boots once a month or more, and never really have to wait for anything at all afterward.

I can't say it any more clearly than that. Technically, that first-time read is always going to make the system faster with an SSD, but if you use your system like I do, the advantage you're bragging about is like a sailboat going around the world and saving about 2 minutes of the trip because of a technical advantage.

I've pretty much chewed all the flavor out of this topic, so if you still need to argue about it I'll let you have your last word and get on with life. Have fun.

Cracklepop

March 28th, 2011, 04:08 AM

.

Performance of the storage device itself.
Performance of the interface (sata/sas/etc)
Performance of the system as a whole.

1. better
2. same
3. better

You mostly talk about the performance of the device--which, I don't argue, is faster on a per-device scale. A good start.
And you also seem to be talking about mobile devices, which in my first post on this thread I acknowledged is well worth it. I haven't brought them up. I'm talking about all devices.

I mostly talk about the performance of the system as a whole, on non-mobile systems. And by my argument, I claim that the device speed difference between SSD and HDD is almost irrelevant for a carefully configured system which rarely shuts down. You may be able to tweak your system to minimize the difference but there is still a difference, and for most of us it's quite a significant one.

Because of that your price per gigabyte doesn't really apply. How expensive is your SSD compared to 12G RAM? I'll follow this (rather pointless - this SSD vs HDD, not RAM) line of thought as soon as you give us the official ''SSD to RAM conversion rate'' ;)

And, once the information comes off it and is stored into RAM, absolutely none of that speed makes any difference whatsoever. On a system sitting on or under a desktop for its whole life, toughness doesn't really matter either. If you allow them to spin down, the power savings is marginal, and disk buffering and high performance filesystems can negate all write lag from spinning the disks back up. But still a speed advantage. No-one is arguing RAM is faster/slower than RAM (?).
Still a durability advantage that matters to some. We aren't limiting this discussion to desktops.
Still an efficiency advantage. Under load we're talking an order of magnitude.

We can talk all day about what COULD happen. Let's stick to facts instead. SSDs are faster, cooler, etc, etc...

... but if you use your system like I do, the advantage you're bragging about is like a sailboat going around the world and saving about 2 minutes of the trip because of a technical advantage. Most of us don't, and no it isn't.

I've pretty much chewed all the flavor out of this topic, so if you still need to argue about it I'll let you have your last word and get on with life. Have fun. Ah, the good old ''I'll let you have your last word'' attempt at avoiding a reply...;)
I c what u did thar...

Sig:
SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They are also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they are relatively very expensive.

Exodist

March 28th, 2011, 04:19 AM

Interesting....I ran across a page of failure percentage statistics the other day for solid state drives and regular hard drives....the solid state drives had a slightly higher failure rate,,even though they had fewer moving parts....odd....
This is due the amount of times the RAM chips on the SSDs can be re-wrote to. Also the tech is a mix of old and new. They will get better over the next 5 years or so, but unless I am running a enterprise level server as RAID I am sticking to regular drives for desktop usage.

I have 4, 320GB WD Blue drives setup as RAID0 and get 380 MB/s Reads and 370 MB/s Writes. At 45USD a drive at the time and over 1.2TB of space. This was the best performance/price ratio IMHO.

@OP,
IMHO SSDs are for enterprise level computing in RAID configurations or for those enthusiast that have the money to waste and think they need them.
You will also see them becoming popular in laptops, this is due to lack of raid options and they do have great vibration resistance. Which makes them a great choose for laptops.

1clue

March 28th, 2011, 04:49 AM

1. better
2. same
3. better

A good start. I haven't brought them up. I'm talking about all devices.

I'm only talking about non-mobile devices, because I've already said there is a worthwhile advantage for mobile devices.

You may be able to tweak your system to minimize the difference but there is still a difference, and for most of us it's quite a significant one.

I'll follow this (rather pointless - this SSD vs HDD, not RAM) line of thought as soon as you give us the official ''SSD to RAM conversion rate'' ;)

There is no SSD to RAM conversion rate and there does not need to be one. The only amount of "spare" RAM you need is the amount of data that is loaded during the time your system is up. Once you've loaded everything you're going to use, it's in RAM and your SSD is not even touched unless you ran out of RAM. If you read a sector, then rewrite it, then read it and write it a thousand times, that sector is only ever read from the platter once, and the buffer will absorb any write time in any practical application.

The necessary amount of RAM could be under 2G if you have a simple system. Mine comes up and runs for a week using less than 33% most times, which means a full installation of Ubuntu with dual monitors, a development environment, a database and a really big flight simulator, along with a couple system updates, and I'm using 4G total RAM right now.

It is absolutely positively NOT pointless, unless you're selling SSDs.

But still a speed advantage. No-one is arguing RAM is faster/slower than RAM (?).
Still a durability advantage that matters to some. We aren't limiting this discussion to desktops.
Still an efficiency advantage. Under load we're talking an order of magnitude.
Let's stick to facts instead. SSDs are faster, cooler, etc, etc...
Most of us don't, and no it isn't.

There are two situations where your strategy can be worse than mine for strict performance:

If your disk cache flushes due to lack of RAM, then is read back in, you have two disk events for the same data, and it's happening during real work, which is a delay WHEN NEEDED.
If you have inadequate RAM and keep hitting your SSD for virtual memory, then your SSD is slower than having sufficient RAM, not only for the fact of a SWAP event but because a page gets written through the disk interface and another one gets read out, which again is a disk delay WHEN NEEDED. Then, when the first page gets swapped back in, that's another pair of disk events WHEN NEEDED.

So go take a look. If your system EVER shows that virtual memory was used, OR that your entire RAM has been used for disk cache, then you are unnecessarily using the disk interface. Which means that you're hitting the disk interface when you are under a load, which is worse than when you are booting.

All my disk activity happens when I could care less. When I reboot, I usually get up to get a snack or go to the bathroom, because it's a reboot and it takes time no matter what your system is.

Ah, the good old ''I'll let you have your last word'' attempt at avoiding a reply...;)
I c what u did thar...

Ah, the old 'you're using a ploy to avoid a reply' ploy...

Sig:
SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They are also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they are relatively very expensive.

Sig:
HDD + RAM is every bit as fast as an SSD under normal operation after the first load, and they are not very expensive at all.

Cracklepop

March 28th, 2011, 07:39 AM

I'm only talking about non-mobile devices, because I've already said there is a worthwhile advantage for mobile devices.
Ok, if you like. But all advantages bar durability still apply for other devices.

There is no SSD to RAM conversion rate and there does not need to be one. I was being facetious, of course there is no such thing. My point was that if you want to compare total price you need to know the quantities, which are not defined.

It is absolutely positively NOT pointless, unless you're selling SSDs.It's price of SSDs vs price of RAM, which is of zero interest to me.

There are two situations where your strategy can be worse than mine for strict performance: You are joking, yes? No-one is going to compare any disk, SSD or HDD, to RAM - certainly not me.

All my disk activity happens when I could care less. When I reboot, I usually get up to get a snack or go to the bathroom, because it's a reboot and it takes time no matter what your system is.It's nice that you can time your disk accesses so well, but the very simple fact remains that swapping the HDD for an SSD in most people's systems will give them a significant performance increase.

Ah, the old 'you're using a ploy to avoid a reply' ploy...:rolleyes:

Sig:
HDD + RAM is every bit as fast as an SSD under normal operation after the first load, and they are not very expensive at all.I see, so, every bit as fast except when it's slower. Got it. ;)

Or, to put it another way: HDD+RAM is as fast as SSD+RAM if you ignore the HDD and SSD and only look at the RAM...

You can keep going as long as you like, but you know and I know that SSDs are faster, and even if you go out of your way to try and minimize the difference, you can't eliminate it.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They are also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they're relatively very expensive.

billdotson

March 28th, 2011, 08:23 AM

SSDs are good for OS installs or programs that are disk intensive. However, they are very expensive. They also use flash memory instead of platters like a normal hard drive. I seriously doubt the flash memory will wear out before you get a new computer or a new SSD (out of boredom of having the same one for forever :))

If the speed of an SSD is worth the cost of one then go for it. Personally, I can't justify the cost for such little storage space.
The use of SSD in mobile devices is definitely useful. You can't have a big hard drive platter rattling around in your smartphone!

However, if you store any information on your computer that is sensitive, such as personal finances (or if you live in the United States and you foolishly write down your social security number, address, bank account number in files, hahaha) be careful. SSDs are, for all intents and purposes, impervious to normal hard drive methods to erase data. Writing zeroes to the SSD? Will not make [most] previously deleted data unfindable.
The only way to ensure that nobody gets previously deleted data off an SSD drive is to encrypt the entire drive (with a good encryption algorithm) and then format it.

Like every other technology (almost anyway) prices will continue to go down. For now, if you just want some really fast storage, you can buy some extra RAM for your machine and create a RAMdisk. I made a 2GB RAMdisk last year and set up a script to copy OpenOffice and other binary files to the RAMdisk and create links to the install folder. OpenOffice opened in less than a second.

disabledaccount

March 28th, 2011, 08:30 AM

So, so rare. I have potentially 4GB (or more) of memory for caching above the disk buffer. The chances of the disk buffer being helpful here are tiny. Have You heard of read prefetch? - it seems You haven't. Re-read takes place every time - due to read-modify cycles caused by misalignment or when application is writing small portions of data (smaller than physical block size, wich is usually 4KB for newest HDDs).

wizard10000

March 28th, 2011, 09:45 AM

However, if you store any information on your computer that is sensitive, such as personal finances (or if you live in the United States and you foolishly write down your social security number, address, bank account number in files, hahaha) be careful. SSDs are, for all intents and purposes, impervious to normal hard drive methods to erase data. Writing zeroes to the SSD? Will not make [most] previously deleted data unfindable.

Actually it will. Since SSDs don't have platters there's nothing to hold a residual magnetic field and almost all disk forensics tools are useless against SSDs. You might find this interesting -

http://news.techworld.com/security/3263093/ssd-fimware-destroys-digital-evidence-researchers-find/

The PDF referenced in TFA is a little too big to attach but explains how standard SSD garbage collection routines made deleted data unrecoverable even with the data connector to the SSD blocked. Check it out -

4. EXPERIMENTAL INVESTIGATION OF SSD & HDD BEHAVIOUR

4.1 Experiment 1: Do HDDs and SSDs preserve deleted data equally?

Purpose of experiment: This experiment examined whether SSDs continue to store data after deletion,
(providing it is not overwritten) in the same manner as traditional magnetic hard drives.

Method of experiment: We filled each drive with files containing the word 'EVIDENCE', quick-formatted the
drive (which resets the filesystem but does not delete the file contents), and then measured the extent to which
the drive continued to contain data from the 'traditionally recoverable' data files.

Detail of experiment: The experiment was carried out as follows. The standard experimental machine and
configuration was used (see Appendix I). The 64GB P64 Corsair SSD and the 80GB Hitachi HDD were tested.
During each test, the drive was connected to the secondary SATA channel on the motherboard and the machine
powered on. After login the drive was 'quick formatted' with the command diskmgmt.msc, producing a new
single primary NTFS partition covering the entire drive. A 196KB template file was selected containing the
word 'EVIDENCE' repeated 25000 times. The partition was filled with copies of the template file using the Fill
program in MSDOS. The copied files were numbered 1.txt, 2.txt, 3.txt... ending when the filesystem would not
allow any further files to be written (all logical blocks were full). Upon the drive being essentially filled with
data, the fill program was stopped. The drive was then immediately quick formatted as before using
diskmgmt.msc, and then immediately shutdown to a power-off state. From the time the drive was filled with
data to the time of complete shutdown was approximately 1 minute, including the time to carry out the quick
format. The machine was then left turned off for approximately 10 seconds.

The machine was then started up, and was logged into as soon as the login screen became available. The
program Go was then started in the Cygwin environment. The Go program repeatedly ran a second program,
Comparer which sampled data for approximately 3 seconds, with 5 second gaps between each set of samples.
The overall statistics for the sample from comparer were recorded into a file automatically by the Go program.
The Go program was started as soon as possible after login (approximately 1 minute from the time of power
on) and ended 15 minutes later. Comparer sampled 10KB of data from 60 positions in the drive separated
regularly by 1000MB (and 80 positions, in the case of the 80GB drive)2. This data was tested by Comparer to
see what portion of the sampled bytes were 'zero bytes'. Experiment 1 was repeated three times in total to the
SSD and HDD to determine whether the results occurred consistently and reproducibly.

Expected result: We expected to see both the SSD and HDD start with a very high fraction of non-zero data
representing the 'pre-formatting' files. We expected the HDD to continue to hold this data without modification,
as there is no process erasing or overwriting the old data. We originally had expected to see the SSD 'forget'
data over 30-60 minutes as the Garbage Collection inside the SSD explored the disk and erased the unused
parts of the disks by itself.

Actual result: The results found are shown in Figure 2. We were astonished to see that contrary to the informal
forum comments we had noted, suggesting that the drive should be left for one hour (or in some cases,
“overnight”) to carry out a partial garbage collection process, Garbage Collection consistently began around
three minutes after power-on, and consistently wiped almost all the drive (according to the samples taken) in
around three minutes. The traditional magnetic HDD results stood out in stark contrast, with the disk data remaining unchanged, and thus available for forensic recovery.

Cracklepop

March 28th, 2011, 10:52 AM

Have You heard of read prefetch? - it seems You haven't. Re-read takes place every time - due to read-modify cycles caused by misalignment or when application is writing small portions of data (smaller than physical block size, wich is usually 4KB for newest HDDs).

You're really scraping the bottom of the barrel...
Sequential reads aren't where you'll really get a feel for the speed of your disk - random reads are.
If you don't align your partitions I have no sympathy for you.
Don't talk to me about writes - this little offshoot topic stems from 1clue and his ideas about reads and disk buffers.
What's important to remember here is that:
...you know, I know, and 1clue knows: SSDs are faster.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They're also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they're (relatively) very expensive.

disabledaccount

March 28th, 2011, 11:56 AM

You're really scraping the bottom of the barrel...
Sequential reads aren't where you'll really get a feel for the speed of your disk - random reads are.
If you don't align your partitions I have no sympathy for you.
Don't talk to me about writes - this little offshoot topic stems from 1clue and his ideas about reads and disk buffers. Who's talking about sequential read? It's not about partition alignment - it's about data alignment - applications are reading/writing data at random offsets - and this is where cache works. Modern HDD's Cache can compensate seek delays in many workload scenarios - You dont have to like HDD's, but You cant beat tha facts. (You can of course say that manufacturers are stupid and are spending money for useless technologies)

wizard10000: That's very important aspect - data recovery from damaged SSD is practically impossible, while with HDDs it's reletively easy (years of experience and great set of tools).

Cracklepop

March 28th, 2011, 12:01 PM

Who's talking about sequential read? It's not about partition alignment - it's about data alignment - applications are reading/writing data at random offsets - and this is where cache works. Modern HDD's Cache can compensate seek delays in many workload scenarios - You dont have to like HDD's, but You cant beat tha facts. (You can of course say that manufacturers are stupid and are spending money for useless technologies)
Perhaps you can apply your own advice to yourself: you don't have to like SSDs, but you can't ignore the facts: they're faster than HDDs.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They're also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they're (relatively) very expensive.

Copper Bezel

March 28th, 2011, 12:02 PM

wizard10000: That's very important aspect - data recovery from damaged SSD is practically impossible, while with HDDs it's reletively easy (years of experience and great set of tools).

I like that you can even use the same facts against each other.

As one of the committed SSD people, I'll go ahead and point out that a more durable drive is less likely to be damaged by accident but just as easy to "shred" with a $7 hammer, even if such a thing were necessary, which it's apparently not.

Grenage

March 28th, 2011, 12:08 PM

This really is a 'horses for courses' argument. Both have their strong and weak points, albeit SSD's only major downside is the cost.

disabledaccount

March 28th, 2011, 12:08 PM

Perhaps you can apply your own advice to yourself: you don't have to like SSDs, but you can't ignore the facts: they're faster than HDDs.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
But what you're talking about? First, You are messing conceptions of cache, Second: nobody here is saying that SSDs are slower - your behaviour looks childish.

Cracklepop

March 28th, 2011, 12:12 PM

But what you're talking about? First, You are messing conceptions of cache, Second: nobody here is saying that SSDs are slower - your behaviour looks childish.

I'm not going to get into any name-calling with you, and I'd rather you didn't do it either.

I'm glad you have no problem saying that SSDs are faster, perhaps you can convince 1clue of their performance benefits.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They're also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they're (relatively) very expensive.

Copper Bezel

March 28th, 2011, 01:14 PM

The broken record bit is getting a little old, and I'm one of the people who agree with you.

I honestly think that magnetic drives will eventually cease to be relevant. They haven't done that yet, and cost and density are the big reasons why. You want to try those weight and heat figures if you're calculating them by the gig?

wizard10000

March 28th, 2011, 01:44 PM

I like that you can even use the same facts against each other.

As one of the committed SSD people, I'll go ahead and point out that a more durable drive is less likely to be damaged by accident but just as easy to "shred" with a $7 hammer, even if such a thing were necessary, which it's apparently not.

Given the proper equipment it's not real difficult to recover a significant amount of data from a mechanical drive that's been physically damaged. The professional companies have recovered data from drives that have been hit with hammers, shot with handguns, been through fires, floods and so on. If the law enforcement or intelligence agency wants your data badly enough (and if there's no or weak encryption) they can get it off the drive.

mamamia88

March 28th, 2011, 01:55 PM

yeah i had an ssd for awhile before i sold it because i couldn't get it to work in the machine i wanted. the machine i did put it in was a toshiba laptop with an i3 and 4gb ram with intel graphics. booted a fresh install of ubuntu in under 10 seconds. i have to say that is really fast. on the note about durability there is a video called ssd awesomeness or something to the effect where there is a guy jumping up and down on a trampoline with about 24 ssds dangling by their cords with absolutely no interruption to the os

sdowney717

March 28th, 2011, 02:04 PM

Perhaps you can apply your own advice to yourself: you don't have to like SSDs, but you can't ignore the facts: they're faster than HDDs.

SSDs are faster, quieter, cooler, tougher, and more efficient than HDDs.
They're also very long lived, although it may be too early to claim they live longer than HDDs.
Unfortunately, they're (relatively) very expensive.

in the future perhaps less than 50 years, people will use ssd all the time and spinning platters will be like yesterdays record players, cd's and dvd drives.
People will also likely wonder about reciprocating piston internal combustion motors and gasoline diesel fuels.

MasterNetra

March 28th, 2011, 04:29 PM

Woah! What in the world?! I'm thinking something must be off here, but you never know. It's a relatively new technology.

Stop right there! This can't possibly be correct. I'd say most HDD's definitely surpass three years of usage.

One another note... I've heard that it's better to leave your computer in standby mode so you hard disks can remain spinning rather than starting and stopping (when turning off and shutting down the computer). I have no idea if this is true or not.

I never said the flash drive was a hard drive. Its a USB that I use frequently. And I was replying to the first post. At anyrate I do have a laptop, however it doesn't get moved much, and it uses a regular hard drive.

1clue

March 28th, 2011, 05:13 PM

I never said they were slower in any normal use case, I never said they would never replace HDDs.

What I said was, you can make your HDD system fast enough to essentially eliminate any benefit of speed that the SSD gives you, for at most the price of a bit of RAM but possibly with the RAM you already have.

The most consistent comment about SSDs on this forum is, "Boy it sure boots fast." The second most consistent comment is, "Boy it sure is expensive."

Nobody's really got any speed comparison on actual real-world tasks. Not benchmarks, but actually running something from an SSD and an HDD that they actually use every day. Try running OpenOffice, then close it, then run it again. Compare the two times on your SSD and see how much time it saves you compared to the total time it takes to run it. The difference between the first time and the second time is the SSD access time. After that, you have the time it takes for the app to initialize itself in memory, which happens no matter what kind of drive you have.

If you hit swap at all, or if your disk cache uses up all available space, then that comparison goes out the window and you're bound by RAM.

Back in the 80's when I was in college, the professors ranted about virtual machines. They said that no matter what you did, a virtual machine could not be as fast as the hardware you ran it on. Technically they were correct and that is still true. However, now we know that virtual machines can dramatically improve the effective speed of your server farm and at the same time reduce the total cost of it. Technically "bare metal" is faster in any benchmark, but in real practice it makes zero difference and costs more.

The same exact thing is true for servers or desktops and SSDs right now.

You're like the guy selling a new bomber to the Pentagon, you have exactly one advantage and you're trying to gain all possible advantage from it. You have a very slight speed advantage that only happens when it doesn't matter, and triple the cost.

You're hung up on one thing, and you think it's the only thing that matters. I'm not arguing your technical details at all. I'm arguing your assertion that an SSD WILL improve the performance of your system in some significant way.

SSDs WILL replace HDDs the same way HDDs replaced tape and tape replaced punch cards. It's how technology works. But right now, on a desktop or server, it's not enough of an advantage to even justify the changes you'd have to put in to make it reliable for long service, no matter how much money you have.

At some point, maybe in the next few years, SSDs will probably be cheaper than HDDs. They'll work the kinks out, improve the number of writes you can make, and nobody will use HDDs anymore. I can't wait. But right now, if somebody leaves their system on all the time, they won't get any noticeable improvement in speed over an HDD for anything they do. You can rant about "SSDs are faster" all you want, but in reality you haven't improved anything significant.

andrewabc

March 28th, 2011, 06:25 PM

The Intel SSD 320 Review: 25nm G3 is Finally Here (http://www.anandtech.com/show/4244/intel-ssd-320-review/1)

They are competing with year old sandforce drives (vertex2 etc).

Nothing wrong with this. More competition better. And still lots of SATAII computers that don't need SATAIII SSD.

Although if they had released this back in November 2009, Intel would have sold a lot more, since it uses the same controller (they just eliminated the throttle).

gnomeuser

March 28th, 2011, 08:01 PM

1clue

March 28th, 2011, 11:16 PM

The complete silence, having just bought one of those 199$ ARM netbooks what gets me about it mostly is how quiet it is.

I would LOVE to put SSDs in all my machines, and move the big storage to a central place (kinda like a personal in house cloud or what Project Cãua aims to do).

For that, your wireless router (if you're doing that, and if it's recent) probably has a USB port on it, you can stick a USB drive on that and configure it as a file server by going to your router configuration.

On a home wireless network that makes a huge amount of sense. You only get half the traffic you would if you were sharing a drive from one wireless node to another through your router.

Cracklepop

March 29th, 2011, 07:37 AM

What I said was, you can make your HDD system fast enough to essentially eliminate any benefit of speed that the SSD gives you, for at most the price of a bit of RAM but possibly with the RAM you already have. You can never eliminate it. Only minimize it. As soon as you replace the slowest part of your system with something faster, the system has become faster, end of story.

Try running OpenOffice, then close it, then run it again. Compare the two times on your SSD and see how much time it saves you compared to the total time it takes to run it. The difference between the first time and the second time is the SSD access time. After that, you have the time it takes for the app to initialize itself in memory, which happens no matter what kind of drive you have. The time to look at is your 'SSD access time', and then compare it to an HDD.

If you hit swap at all, or if your disk cache uses up all available space, then that comparison goes out the window and you're bound by RAM. How is this relevant? It applies to both HDDs and SSDs....except that when you do use swap, SSDs do it faster.

You're like the guy selling a new bomber to the Pentagon, you have exactly one advantage and you're trying to gain all possible advantage from it. You have a very slight speed advantage that only happens when it doesn't matter, and triple the cost A speed advantage that happens when it *does* matter, for everyone besides you personally, and heat, noise, efficiency, etc advantages.
Absolutely right about price: I've said it over and over. If the prices were the same I'd say HDDs would become very obscure very quickly - who would choose the inferior product?.
Currently they simply don't offer the value for money that HDDs do (although this is an opinion that people richer than myself may disagree with), which is why I own just one SSD, but many HDDs.

You're hung up on one thing, and you think it's the only thing that matters. I'm not arguing your technical details at all. I'm arguing your assertion that an SSD WILL improve the performance of your system in some significant way. For the vast majority of systems an SSD absolutely positively will give a significant performance increase.

I can't wait. Why not? You keep saying it won't make any noticeable difference...

But right now, if somebody leaves their system on all the time, they won't get any noticeable improvement in speed over an HDD for anything they do. You can rant about "SSDs are faster" all you want, but in reality you haven't improved anything significant. This is just false. The system will be faster. There isn't any room for denial, it's just a simple fact. No, not everything you do will be affected, and you can take steps to minimise the difference, but no matter what you do there will be times when it makes a difference, and for most systems the change from HDD to SDD is a significant improvement.

Copper Bezel

March 29th, 2011, 09:14 AM

SSDs WILL replace HDDs the same way HDDs replaced tape and tape replaced punch cards. It's how technology works. But right now, on a desktop or server, it's not enough of an advantage to even justify the changes you'd have to put in to make it reliable for long service, no matter how much money you have.

This is true.

Nobody's really got any speed comparison on actual real-world tasks. Not benchmarks, but actually running something from an SSD and an HDD that they actually use every day. Try running OpenOffice, then close it, then run it again. Compare the two times on your SSD and see how much time it saves you compared to the total time it takes to run it. The difference between the first time and the second time is the SSD access time. After that, you have the time it takes for the app to initialize itself in memory, which happens no matter what kind of drive you have.
I'm not arguing your technical details at all. I'm arguing your assertion that an SSD WILL improve the performance of your system in some significant way.
But right now, if somebody leaves their system on all the time, they won't get any noticeable improvement in speed over an HDD for anything they do. You can rant about "SSDs are faster" all you want, but in reality you haven't improved anything significant.

This is not. Speed, not durability or temperature or weight or boot time, was the primary reason I initially switched. As I've said earlier in this thread, the difference is immediate and makes for a more responsive desktop. OpenOffice is a great example - the time between double-clicking a file and typing was cut in half when I switched, and whether or not that's a real, practical benefit per se, it was worth the ~$100 USD for the drive.

That has nothing to do with a server situation, and it's a benefit exclusively relevant to the responsiveness of a graphical desktop environment, so don't misread me as getting back into that server nonsense, where calculations per $ per foot are the only consideration.