PDA

View Full Version : 'four times faster'???



fuscia
January 19th, 2006, 06:06 AM
four times faster than what? eventually? how dumb does apple think their customers are? i'd buy 'one fourth as slow' maybe. i know some of you are enthusiastic about macs, but would you ever claim speed to be a mac quality?

stuporglue
January 19th, 2006, 06:44 AM
I'll agree with you on clock speeds, Macs aren't (or haven't been) very fast.

I've never had to sit and wait for it to do things though, unless I was preparing to burn a DVD. I think Apple has made good use of their resources. Or maybe we're just to the point where processors are fast enough for our tasks, and any more speed isn't usually needed.

Derek Djons
January 19th, 2006, 09:03 AM
I'll agree with you on clock speeds, Macs aren't (or haven't been) very fast.

I've never had to sit and wait for it to do things though, unless I was preparing to burn a DVD. I think Apple has made good use of their resources. Or maybe we're just to the point where processors are fast enough for our tasks, and any more speed isn't usually needed.


The results are based of course on nice 'softly modified & edited' standard Benchmark results. These indications are useless when comparing your real-life applications.

But that doesn't shut it all down. I've been selling and using countless Centrino notebooks. The CPU can't b compared with an Intel Pentium IV, but it can't be compared to a Celeron. Intel Centrino's are capable of doing a lot heavy duty and still let the user benefit his quiet and thin machine.

These days, and that counts for AMD as Intel the MHz or GHz doens't means anything these days. It's all about the technology in and around the CPU which let the user benefit from all key features, speed, silence, portability.

Once more, I also find the 1 to 3 times faster a fairytale, but still compared to the PowerPC this CPU and related hardware will kick ass!

nocturn
January 19th, 2006, 09:32 AM
You cannot compare CPU speeds in Mhz, the old Macs used RISC based processors which are completely different from the current CISC based ones.

RISC can do a lot of things faster, but has a smaller instruction set, so some things are in fact slower.

In general, a RISC processor at the same speed is much faster then a CISC one though.

The problem is that those chips haven't evolved much over the last years.

mcduck
January 19th, 2006, 10:16 AM
PowerPC processors aren't too fast in general things like web surfing. But when doing things that need heavy floating point calculations, like image and sound related tasks, they are way ahead of x86.

And, as nocturn mentioned, you can't directly compare speeds of different CPU architectures. Even comparing speeds of different x86 CPU's doen't work..

fuscia
January 19th, 2006, 11:07 AM
PowerPC processors aren't too fast in general things like web surfing. But when doing things that need heavy floating point calculations, like image and sound related tasks, they are way ahead of x86.

so, i guess the intel macs will be four times slower for those tasks.

prizrak
January 19th, 2006, 01:49 PM
You cannot compare CPU speeds in Mhz, the old Macs used RISC based processors which are completely different from the current CISC based ones.

RISC can do a lot of things faster, but has a smaller instruction set, so some things are in fact slower.

In general, a RISC processor at the same speed is much faster then a CISC one though.

The problem is that those chips haven't evolved much over the last years.
The current x86 line is not CISC and hasn't been in a while. They are a combination of both, PPC CPU's are not faster in ANYTHING than the x86 line and haven't been for years. One of the reasons Apple has switched is PPC not being competitive anymore, there wasn't much of a reason to make it better unlike the x86 line which had two companies fighting for the market forcing them to create better and faster CPU's.
If you want to compare x86 to the IBM CPU's compare the new Cell instead of the PPC that is where the real power and innovation is at.

Viro
January 19th, 2006, 02:20 PM
The current x86 line is not CISC and hasn't been in a while. They are a combination of both, PPC CPU's are not faster in ANYTHING than the x86 line and haven't been for years.

Wanna bet ;)? There is nothing similar to Altivec on the x86. SSE3 is close, but still no where near as well implemented as Altivec. Just read up the Apple Developer docs to see how many Altivec operations do not have SSE3 counterparts.

Now, you wanna see Altivec shine? I've posted this before, but I'll post it again. Look at the FFTW benchmarks of a 2GHz G5 (http://www.fftw.org/speed/g5-2GHz/), 2GHz Opteron(in 32bit (http://www.fftw.org/speed/opteron-2GHz-32bit/) and 64bit (http://www.fftw.org/speed/opteron-2GHz/) mode). At the start, it's mainly double precision benchmarks and the performance is not too shabby, with the G5 handily beating the Opteron. Scroll down to the middle of the page and look at the single precision benchmarks where Altivec is used. See that the G5 performs at the same clock speed performs 2x (!!)better than the Opteron. Basically, in anything that can be optimized for Altivec (media encoding, encryption[1 (http://n0cgi.distributed.net/speed/query.php?cputype=112&arch=all&contest=all&multi=0), 2 (http://n0cgi.distributed.net/speed/query.php?cputype=115&arch=all&contest=all&multi=0)], filters and DSP tasks, lots of scientific simulations, etc) practically nothing can touch the G5.

There seems to be a lot of hostility towards Apple and Macs from a select few on this board, as seen in anothe thread where the starting poster claimed that "Apple hardware is useless". My only advice is this. Take a deep breath, and breath out. These are only processors. Unlike what you might think, all processors have strengths and weaknesses and the G5 is a very amazing chip.

Nevertheless, raw performance isn't why Apple has switched to Intel. Rather, it's the almost cliched phrase "performance-per-watt". The G5's perform great. But they generate too much heat to be put into notebooks. Sure, you could get them into 'desktop replacement' notebooks like some Pentium 4 notebooks, but Apple wants thin and light notebooks, not huge bricks. As such, they needed a replacement and that's where Yonah fits in. In terms of "performance-per-watt", there is no processor around that can top it.

nocturn
January 19th, 2006, 02:33 PM
The current x86 line is not CISC and hasn't been in a while. They are a combination of both, PPC CPU's are not faster in ANYTHING than the x86 line and haven't been for years. One of the reasons Apple has switched is PPC not being competitive anymore, there wasn't much of a reason to make it better unlike the x86 line which had two companies fighting for the market forcing them to create better and faster CPU's.
If you want to compare x86 to the IBM CPU's compare the new Cell instead of the PPC that is where the real power and innovation is at.

Yes, the x86 is a hybrid, but the CISC architecture is still dominant.

PPC are not the only RISC based architectures arround. Sun also produces them and even at their lower clock speed, they beat x86 hands down.

I haven't seen much benchmarks comparing PPC and x86 at the same clockspeeds, that would be a good indication about performance.

frodon
January 19th, 2006, 03:44 PM
You cannot compare CPU speeds in Mhz, the old Macs used RISC based processors which are completely different from the current CISC based ones.

RISC can do a lot of things faster, but has a smaller instruction set, so some things are in fact slower.

In general, a RISC processor at the same speed is much faster then a CISC one though.

The problem is that those chips haven't evolved much over the last years.You're right, too many people think that they can compare 2 processors with their clock speed but you really cannot.
CISC an RISC architecture have both advantages and disadvantages. RISC is designed to perform small operations faster than CISC architecture but RISC architecture is slower when executing a high level instruction conmpared to CISC architecture which will use less instruction to perform the action.
I would prefer to compare 2 processors with the number of instructions run in one clock cycle, it's a more reliable information than the speed. This parameter depend generally on the level of pipelining.

prizrak
January 19th, 2006, 04:16 PM
Sun also produces them and even at their lower clock speed, they beat x86 hands down.
Was it Windows vs Solaris that was used for testing? I want to see both machines either running Linux or Solaris to be able to have a more fair match between the two.

Wanna bet ? There is nothing similar to Altivec on the x86. SSE3 is close, but still no where near as well implemented as Altivec. Just read up the Apple Developer docs to see how many Altivec operations do not have SSE3 counterparts.

Now, you wanna see Altivec shine? I've posted this before, but I'll post it again. Look at the FFTW benchmarks of a 2GHz G5, 2GHz Opteron(in 32bit and 64bit mode). At the start, it's mainly double precision benchmarks and the performance is not too shabby, with the G5 handily beating the Opteron. Scroll down to the middle of the page and look at the single precision benchmarks where Altivec is used. See that the G5 performs at the same clock speed performs 2x (!!)better than the Opteron. Basically, in anything that can be optimized for Altivec (media encoding, encryption[1, 2], filters and DSP tasks, lots of scientific simulations, etc) practically nothing can touch the G5.

There seems to be a lot of hostility towards Apple and Macs from a select few on this board, as seen in anothe thread where the starting poster claimed that "Apple hardware is useless". My only advice is this. Take a deep breath, and breath out. These are only processors. Unlike what you might think, all processors have strengths and weaknesses and the G5 is a very amazing chip.

Nevertheless, raw performance isn't why Apple has switched to Intel. Rather, it's the almost cliched phrase "performance-per-watt". The G5's perform great. But they generate too much heat to be put into notebooks. Sure, you could get them into 'desktop replacement' notebooks like some Pentium 4 notebooks, but Apple wants thin and light notebooks, not huge bricks. As such, they needed a replacement and that's where Yonah fits in. In terms of "performance-per-watt", there is no processor around that can top it.
I never believe Apple vs x86 benchmarking.
Reasons:
1) Different OS's
2) Highly optimized hardware
I'm not an Apple hater but in order to have a fair benchmark you need comparable environments. If you run PPC Linux vs x86 Linux with carefully selected hardware on the x86 and then run the same exact benchmark then I will believe the benchmarking. Remember MacOS was designed and optimized for very specific hardware architecture, Windows has to run on million different combinations. Hell even Linux got i686 and K7 kernels optimized for the specific CPU family. Also Apple is switching their desktops, where heat makes little difference to Intel so your thin laptop argument holds no water. The main reason Apple is not using PPC anymore is because IBM doesn't feel like innovating enough for Apple to stay competitive, they are concentrating on the console market where innovation cycle is about 5 years.

nocturn
January 19th, 2006, 04:27 PM
Was it Windows vs Solaris that was used for testing? I want to see both machines either running Linux or Solaris to be able to have a more fair match between the two.

Both Linux x86 vs. Solaris/Sparc as Linux x86 against Linux/Sparc.



I never believe Apple vs x86 benchmarking.
Reasons:
1) Different OS's
2) Highly optimized hardware


The problem of benchmarking two different architecture is that you can never level the playingfield completely. Even the choice of operations to benchmark will make a difference.

Viro
January 19th, 2006, 05:29 PM
Was it Windows vs Solaris that was used for testing? I want to see both machines either running Linux or Solaris to be able to have a more fair match between the two.


On processor intensive tasks that do not stress the entire system, the OS should have a negligeble impact on performance, unless one OS is horrendously broken. If you're just crunching data, why should the OS matter? That's the beauty of such microbenchmarks. They are useless in comparing operating systems, but they are very useful in understanding the CPU.



I never believe Apple vs x86 benchmarking.
Reasons:
1) Different OS's
2) Highly optimized hardware


If you believe the stuff that is published as OS X vs Linux benchmarks, you will agree that if anything, using OS X severely handicaps the Mac platform! Here's one example of such a benchmark (http://www.anandtech.com/mac/showdoc.aspx?i=2436). If you believe the flames about microkernels vs monolithic kernels, you will also agree that Linux has the edge. If you factor in the fact that GCC on PowerPC is not as well developed as GCC on x86, you will also agree that the benchmark would be in favor of the x86 machine. Given that Linux on x86 is far more widely used and is much more developed than Linux on the PowerPC, how can you say that running Linux on both architectures constitutes 'fair'?



Hell even Linux got i686 and K7 kernels optimized for the specific CPU family.

The question of course is whether these different kernels provide any noticeable difference :). I've switched from the default i386 kernel to a i686 kernel to a Pentium 4 kernel on my work machine, and aside from uname -a telling me there is a difference, I wouldn't have known which kernel I was using.

I'm sorry, but your excuses do not hold much water. You've gone from saying with absolute confidence that nothing on PowerPC is faster than on x86 and has been for years, and when evidence is shown that proves you wrong, you now claim that you don't 'believe' such benchmarks as it isn't fair.



Also Apple is switching their desktops, where heat makes little difference to Intel so your thin laptop argument holds no water. The main reason Apple is not using PPC anymore is because IBM doesn't feel like innovating enough for Apple to stay competitive, they are concentrating on the console market where innovation cycle is about 5 years.

Take a look at the iMac and tell me temperature doesn't matter. To be honest, nobody knows for sure why Apple moved from IBM and we can speculate all we want. The "performance-per-watt" excuse is what Steve feeds us, so we might as well take it at face value.

prizrak
January 19th, 2006, 06:53 PM
On processor intensive tasks that do not stress the entire system, the OS should have a negligeble impact on performance, unless one OS is horrendously broken. If you're just crunching data, why should the OS matter? That's the beauty of such microbenchmarks. They are useless in comparing operating systems, but they are very useful in understanding the CPU.

I dunno enough about OS design to say that you are wrong so lets just say you are right :) However OS makes a huge difference in the real world where there is no such thing as "raw" number crunching.

If you believe the stuff that is published as OS X vs Linux benchmarks, you will agree that if anything, using OS X severely handicaps the Mac platform! Here's one example of such a benchmark. If you believe the flames about microkernels vs monolithic kernels, you will also agree that Linux has the edge. If you factor in the fact that GCC on PowerPC is not as well developed as GCC on x86, you will also agree that the benchmark would be in favor of the x86 machine. Given that Linux on x86 is far more widely used and is much more developed than Linux on the PowerPC, how can you say that running Linux on both architectures constitutes 'fair'?

OS X has the Aqua interface that would slow it down as compared to your average Linux desktop. While Linux for x86 is alot more developed at the very least it can give you a bit more claim on fairness as opposed to a system specifically tailored for the hardware. As said by nocturn above there is no real way to compare the two.

The question of course is whether these different kernels provide any noticeable difference . I've switched from the default i386 kernel to a i686 kernel to a Pentium 4 kernel on my work machine, and aside from uname -a telling me there is a difference, I wouldn't have known which kernel I was using.
Yes you did it on a desktop, I saw no improvement between the 386 and K7 kernels on my desktop either since the video card is decent enough to take the bulk of the rendering load off of the CPU. On my P4M laptop tho the difference is noticeable not huge but noticeable.

I'm sorry, but your excuses do not hold much water. You've gone from saying with absolute confidence that nothing on PowerPC is faster than on x86 and has been for years, and when evidence is shown that proves you wrong, you now claim that you don't 'believe' such benchmarks as it isn't fair.
How about we halt this debate until we have figured from OS X on Intel vs OS X on the G5? You will be hard pressed to prove to me that OS X isn't developed enough for Intel.

Take a look at the iMac and tell me temperature doesn't matter. To be honest, nobody knows for sure why Apple moved from IBM and we can speculate all we want. The "performance-per-watt" excuse is what Steve feeds us, so we might as well take it at face value.
If Apple can't design a desktop to decipate heat properly that's THEIR problem. By and large heat means very little on a desktop, I have 12 fans in mine it's not gonna be overheating anytime soon, if yours does shove some fans in it or get a better case.

Viro
January 20th, 2006, 09:31 AM
However OS makes a huge difference in the real world where there is no such thing as "raw" number crunching.


Here we go again about the mythical 'real world'. My work requires me to do lots of time series analysis (i.e. number crunching). There are lots of other people working in similar fields, where the ability of a processor to number crunch is important, hence the G5 is a fantastic chip. This is what I have been responding to ever since you made the confident assertion that PowerPCs do not have anything that come close to x86 processors. I have provided benchmarks that show otherwise.




OS X has the Aqua interface that would slow it down as compared to your average Linux desktop. While Linux for x86 is alot more developed at the very least it can give you a bit more claim on fairness as opposed to a system specifically tailored for the hardware. As said by nocturn above there is no real way to compare the two.


What does Aqua have to do with anything? Besides, Aqua isn't slow. Don't just read everything you see on the Internet. There are real ways of comparing systems and that has to do with execution time. If computer A takes 10 minutes to complete a task, while computer B takes 5 minutes, regardless of the underlying hardware/OS used you can confidently state that computer B is faster than computer A. Now, if you ensure that computer A and computer B are comparable in specification, you've got a pretty good comparison there, aye?

Look with me at the results found on here (http://www.anandtech.com/mac/showdoc.aspx?i=2436&p=8). According to those results, OS X performs horribly compared to Linux, sometimes 2 - 5x worse. So as I have said previously, surely you must agree that running OS X on the G5 must severely hamper it? And yet, it still manages to perform certain tasks that blow pretty much everything out of the water.



Yes you did it on a desktop, I saw no improvement between the 386 and K7 kernels on my desktop either since the video card is decent enough to take the bulk of the rendering load off of the CPU. On my P4M laptop tho the difference is noticeable not huge but noticeable.


Here you make assumptions about the quality of my video cards :). I don't do 3D much, so I have a very poor video card. Besides, the things that I do don't really involve graphics apart from displaying some graphs.

Still, you have proved my point. At best, you see a small difference. How can such a small difference account for the great performance delta of the Athlon 64 vs G5 in tasks where Altivec can be used?



If Apple can't design a desktop to decipate heat properly that's THEIR problem. By and large heat means very little on a desktop, I have 12 fans in mine it's not gonna be overheating anytime soon, if yours does shove some fans in it or get a better case.

You are obviously not familiar with Apple's design goals or its target market. Suffice to say, Dell and gang can get away with stuffing as many fans as they want into a computer case, but Apple can't. If their machines had 12 fans (!!) they will lose nearly all their customers. The iMac cannot afford to have a very hot CPU given its form factor. The Powermac cannot afford to have a hot CPU for the more fans you have, the louder your machine is going to get, and if you consider that many audio engineers use Powermacs, Apple will lose business! Heat may mean little to most desktops, but there are some areas where heat matters.

Bandit
January 20th, 2006, 09:44 AM
I dont know about Intel, but ALL AMD CPUs are technicaly RISC based<--key word based!! But they have evolved so much since the K6 it is getting fuzzy to really tell what they are now.

Getting back to the orginal topic, one of my friends dad just purchased one of the new Intel based Mac's. His discription of it was "Its F'ing Fast!!". He said it was so much faster then the PPC Macs that you would have to see for your self to understand. Which is understandable being that the PPC architecture is getting pretty old by todays standards.

Cheers,
Bandit

fuscia
January 20th, 2006, 09:50 AM
He said it was so much faster then the PPC Macs that you would have to see for your self to understand. Which is understandable being that the PPC architecture is getting pretty old by todays standards.


my mother's faster than ppc macs.

the only reasonable comparison, for benchmark testing, would be how fast is their $1000 piece of crap compared to these other guys' $1000 piece of crap.

Bandit
January 20th, 2006, 10:00 AM
my mother's faster than ppc macs.

the only reasonable comparison, for benchmark testing, would be how fast is their $1000 piece of crap compared to these other guys' $1000 piece of crap.
ROTF.. Your pretty much right....

Viro
January 20th, 2006, 11:50 AM
Which is understandable being that the PPC architecture is getting pretty old by todays standards.

:rolleyes:
Check out PowerPC (http://en.wikipedia.org/wiki/PowerPC) and x86 (http://en.wikipedia.org/wiki/X86) histories. You are the first person I've met who says that PowerPC is aged when compared with x86. ;)

You are correct when you say that there is now no real distinction between CISC and RISC CPUs, as most x86 processors are RISC-like internally.

prizrak
January 20th, 2006, 02:05 PM
Viro,
Like I said above we will halt this debate until a meaningful comparison between OS X on Intel and OS X on x86 is done. If you we can go but Bandit's story it would seem that I'm the one who is correct.

You are obviously not familiar with Apple's design goals or its target market. Suffice to say, Dell and gang can get away with stuffing as many fans as they want into a computer case, but Apple can't. If their machines had 12 fans (!!) they will lose nearly all their customers. The iMac cannot afford to have a very hot CPU given its form factor. The Powermac cannot afford to have a hot CPU for the more fans you have, the louder your machine is going to get, and if you consider that many audio engineers use Powermacs, Apple will lose business! Heat may mean little to most desktops, but there are some areas where heat matters.
Dell didn't stuff 12 fans in my case, I got my own case that I stuffed the fans into :) There are many ways to dissipate heat and they don't have to be loud, water cooling comes to mind. I stand by my statement if ANY desktop overheats because of a hot CPU it has been designed by a retarded monkey since it's way too easy to dissipate heat in that kind of space. The only exception I could see would be the super small factor machines such as the mini or the shuttle but those aren't meant to be too high performance in the first place.
I would also say that a G5 cannot be compared against an x86 since it's a fully 64bit CPU it should be compared against an Itanium.

nocturn
January 20th, 2006, 02:23 PM
I stand by my statement if ANY desktop overheats because of a hot CPU it has been designed by a retarded monkey since it's way too easy to dissipate heat in that kind of space.

I did think that our Sun Ultra 60 at my previous job was pretty nice. Outperforming a desktop Athlon while the CPU had just a passive cooler.

AmboyGuy
January 20th, 2006, 03:30 PM
Here's a MacWorld test. http://www.macworld.com/2006/01/features/imaclabtest1/index.php

IntelMac 20% faster than G5 Macs

prizrak
January 20th, 2006, 03:38 PM
I did think that our Sun Ultra 60 at my previous job was pretty nice. Outperforming a desktop Athlon while the CPU had just a passive cooler.
Nothing can touch Sun I LOVER their systems too bad their a bit too expensive.

prizrak
January 20th, 2006, 04:10 PM
Here you make assumptions about the quality of my video cards . I don't do 3D much, so I have a very poor video card.
Of course I could be wrong but since it's a desktop chances are it is at least an nVidia/ATI based card with its own RAM. As opposed to a laptop, which has a built in Intel card that shares RAM with the system :)
My desktop Video isn't all that powerful either it's a GeF4MX440 but it still get renderaccell to make it faster ;)

fuscia
January 20th, 2006, 04:22 PM
Here's a MacWorld test. http://www.macworld.com/2006/01/features/imaclabtest1/index.php

IntelMac 20% faster than G5 Macs


Instead, our tests found the new 2.0GHz Core Duo iMac takes rougly 10 to 25 percent less time than the G5 iMac to perform the same native application tasks, albeit with some notable exceptions. (If you'd prefer, that makes the Core Duo iMac 1.1 to 1.3 times as fast.) And we also found that applications that aren’t yet Intel-native—which must run using Apple’s Rosetta code-translation technology—tend to run half as fast as the same applications running natively on the iMac G5.

maybe they should go for 'two times as elegant' instead.

poofyhairguy
January 21st, 2006, 11:07 AM
There seems to be a lot of hostility towards Apple and Macs from a select few on this board, as seen in anothe thread where the starting poster claimed that "Apple hardware is useless". My only advice is this. Take a deep breath, and breath out. These are only processors. Unlike what you might think, all processors have strengths and weaknesses and the G5 is a very amazing chip.


PPC chips are fun to play with. I want a g3 iBook or a first gen. Macmini to run Linux on....the CPUs are really fun.

ssam
January 21st, 2006, 12:00 PM
Here's a MacWorld test. http://www.macworld.com/2006/01/features/imaclabtest1/index.php

IntelMac 20% faster than G5 Macs

doesn't it depend a lot on the task?

the core duos should be a lot faster at integer tasks, but about the same at floating point (fp is a powerpc strong point).

however most task are limited by the hard disk seak time and data rate. apparently steve jobs said in the keynote that the disk were not twice as fast. http://apple.slashdot.org/comments.pl?sid=174537&cid=14521385 . If you swapped a 2ghz pentium 4 with a 4ghz pentium 4 and kept the other components in your computer the same then you would not see a 2x speed increase everywhere.

another post http://apple.slashdot.org/comments.pl?sid=174537&cid=14520791 talks about gcc. he says that the 2ghz core duo beats a 1ghz G4 at compiling by 6-7x. so for some tasks there is a lot of gain.

another point about the only 10% gain on desktop apps is, are those apps multithreaded? if not then they will only be using half the cpu.