Page 13 of 13 FirstFirst ... 3111213
Results 121 to 126 of 126

Thread: memory question

  1. #121
    Join Date
    Mar 2009
    Beans
    1,711

    Re: memory question

    Oops.

    It seems that I posted without reading the extra page or two of posts that happened in the meantime.
    Last edited by 1clue; January 21st, 2011 at 12:58 AM.
    Help stamp out MBR partition tables. Use GPT instead!

  2. #122
    Join Date
    Jan 2011
    Location
    columbia,sc
    Beans
    163
    Distro
    Ubuntu 10.10 Maverick Meerkat

    Re: memory question

    so, will actually listen to any ideas that come about because I've learned my lesson.
    does anyone have anymore ideas to work on the speed here? After checking my logs several times, i have verified everything i can think of and found no further issues. so i think i'm ready to actually LISTEN this time.

    And btw, sorry for being a pain in the **** to begin with

  3. #123
    Join Date
    Mar 2009
    Beans
    1,711

    Re: memory question

    OK Gary,

    First thing to do is to start doing what you normally do, and keep tabs on what sort of performance you currently get. Save this one someplace, most people who try to optimize performance like to have a benchmark from day zero, but often they never think of it until months later.

    You'll want to keep tabs on this sort of stuff:
    1. Physical memory usage.
    2. Virtual memory usage
    3. Memory used as disk cache (probably irrelevant if you're using virtual memory at all)
    4. CPU usage on all cores.
    5. Disk usage on all disks, partition space.
    6. Disk bandwidth used on all disks individually
    7. Disk bandwidth used on the disk IO interface (sata card, whatever)
    8. Video card CPU load, etc.
    9. Network load.
    10. How choppy is the graphics rendering?
    11. How degraded is the sound?
    12. How long does the foreground task freeze while something goes on in the background?
    13. Are there deadlocks?
    14. I'm sure I missed some things.


    You'll also want to find out what sort of benchmarks pertain to your particular type of use, and perform them on your system whenever you change something. But keep in mind that while a general benchmark can help you see where you stand with others by that benchmark, the one that really counts is the task you run when you're using your system. 3x that game, for example.

    Some of this stuff you can get a GUI applet for, and have it always running. When you're gaming, have the applet visible so you can see how much CPU, memory etc is being used. If something pegs to 100% for any significant time, then that's the bottleneck. Look into improving that thing.

    Look for the tools that make sense to you for measuring your system performance. The best tool is not always the one that dumps the most irrelevant information on the screen, but the one that clearly and concisely shows you what you're interested in knowing in a way that you can tell what's going on at a glance. I 'grew up' on the command line, so things like top and htop and other command-line tools don't bother me at all. Even so, I still use things like Gnome panel for my normal system monitoring.

    Don't bother upgrading something unless YOU see it being a bottleneck on YOUR system, when you're doing something that you think is unacceptably slow. What I do is going to be drastically different than what you do. You could buy exactly my hardware, and chances are we won't have the same bottlenecks.

    If your system suddenly degrades, look in the logs. Now that you've cleaned them up you should be able to spot errors easier. Get familiar with the entire log, as bizarre as that seems it can teach you a lot about what your system is doing.

    You don't have to do all this at once. It's an ongoing process and a moving target. I'm still learning a lot, and I'm 45. Keep in mind that with anything, the amount of effort you put in to understanding and maintaining a process, the better the results you will get out of it.

    I would suggest that rather than having all your questions under this thread, separate them and make a relevant title for each topic. If necessary post links to other threads that might describe your situation accurately. Use the search feature to see if somebody else already asked that question, but don't interrupt their thread with your own questions unless you feel it's REALLY pertinent. Sometimes I begin my post with "I was reading <link to a thread> and am having a similar problem but didn't want to confuse that thread. Here's how it's different on my system...."

    Good luck and have fun.
    Help stamp out MBR partition tables. Use GPT instead!

  4. #124
    Join Date
    Jan 2011
    Location
    columbia,sc
    Beans
    163
    Distro
    Ubuntu 10.10 Maverick Meerkat

    Re: memory question

    as ive said many times, thank you much for fast and concise answers 1clue. its people like you that keep technology growing and moving without losing sight of where it started.



    and like i said, thank you very much. im going to continue looking into this. ive already been watching some of these the lst week or so to see what happens, so i must have learned something.

  5. #125
    Join Date
    Aug 2009
    Location
    Brisbane
    Beans
    3,779

    Re: memory question

    Damn it, I though I hit 'post'.....ohh well, I cant be bothered to write out in detail what I wrote last time, but here is the gist of it.

    Gnome system monitor is a bit of a resource pig. Its only really handy for the ability to see CPU use over time (and then you can correlate what caused that RAM spike, CPU use spike, etc).

    Onboard video can really screw up results. They use CPU power and main system memory bandwidth, so graphically 'heavy' tasks will create more load than a standalone video card (and because its using those system, anything else that uses them will be slowed down by the onboard video)

    Checking var/logs inst a bad idea. I normally only do it when I think there is a serious problem though.

    Dont go to crazy on any single benchmark. I've seen people 'tune' thier machines from a single benchmark, then find that all they've done is make that benchmark better, but hurt everything else.

    Easy hint- in your BIOS, disable all the junk you dont use or need (eg, firewire, serial ports, parallel ports, unused IDE ports, etc). Set your HDDs, dont leave them on 'atuo-detect' (detecting the HDDs does take time). You might only gain a tiny amount, but liek they say in drag racing "seconds are made up of lots of hundredths of seconds"

  6. #126
    Join Date
    Mar 2009
    Beans
    1,711

    Re: memory question

    Cascade,

    I agree with pretty much everything you said in that last post.

    The things you bring up are what you need to keep in mind when you're trying to make sense of the results, and figuring out what to do about it.

    At the stage Gary is at, he's just trying to find out what's slowing his system down. Not details exactly, but big bottlenecks. Not enough RAM, not enough CPU (for whatever reason) or whatever.

    I completely agree about not focusing on one benchmark, unless that benchmark is the system's performance when running your most demanding real-life scenario in which case it's perfectly fine to do that.

    IMO, if your system hits swap at all more than once a day you might consider getting more. I might be an extreme case, but if you know what's going on when you swap, it's like being in an airplane that's about to crash, and you stop to read a book before fastening your seat belt. You never hit swap when your system isn't working. When it's working, it means you're waiting for something to finish, and here you have to swap RAM over to disk?

    I actually go the opposite direction here. I use tmpfs anywhere I think it will do any good. I have 12G RAM, so when it uses the swap partition that really means it's using RAM as a disk drive. If I run out of RAM then it writes to the disk, but not until then. It's been several months since I actually swapped anything out, and that was when I bumped it up to 12.

    Another thing that gnome system monitor does is tell you how much of your RAM is used as a disk cache. In my case, usually more than what the system thinks of as active.

    I'm not sure if I would go nuts with PAE on a 32-bit system. That's a cluster ****. I don't know how well it handles that sort of thing, back when they went from 16-bit to 32-bit they did something similar and it was a noticeable performance hit when you had to page out. The more you paged out the worse it got.

    If you're using onboard video and you're running out of CPU, then consider getting a video card, even if you're just browsing the web. The performance improvements for even a cheap video card are stunning, and to get a decent one helps a lot even if you don't consider yourself to be video-bound. Try to gauge how much video performance you need and scale the card to that.
    Help stamp out MBR partition tables. Use GPT instead!

Page 13 of 13 FirstFirst ... 3111213

Tags for this Thread

Bookmarks

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •