Sidebar Sidebar
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Pentium 4 vs Athlon 64 vs G5 Benchmarked

Status
Not open for further replies.
Who says XP is bad for multiprocessing?

I did a test to push my 3.0c @ 3.5ghz and this was the setup:

Perfect disk doing its defragmentation
45 Instances of IE
Nero burning a cd w/buffer underrun protection disabled
UT running in the back bot match 16 bots, 1024x768
WMP9 and WMP6.4 both running xvid videos
Winamp playing an mp3
Bit torrent downloading
Hyper Cam capping a 1024x644 of real time video of the screen
Prime95 doing a torture test
1 instance of IRC

ALL RUNNING at ONCE for 1 hour without any errors from Prime95 and no buffer underrunsl Cpu temps were at 51c MAX using stock Intel Heatsink.

The results are from my personal configuration and custom installation of XP. Doing this used up all my cpu usage and alot of my physical and virtual memory. This was the max number of processes I could run before it wouldnt open any more applications. Even at this limit I was still able to go into UT and play in the Bot match that I had running. I had this running for more then 30 minutes. I had to put in more discs for nero since my burner is 16x about 5.30 minutes to burn a cd.

The ONLY time I ever see a Blue Screen is when I oc the pc at a 5:4 divider or 3:2. When im at 1:1 its as equaly as stable as stock speed.

Where is the instability in xp from multitasking? Give me a simular scenario for OSX. If anyone is interested I can post a link to the video or simply take a screen shot of the screen.
 
Originally posted by pgwalsh
Exactly.. If all we cared about was speed, we'd all be weenies. Just like dem utter folks on dah dark side. Dem boys like speed and blue screens. y'nah... Just can't figure dem out.
Click to expand...
I hope you were being sarcastic with that comment.

Seriously, no computer is perfect (yes, a Mac is a computer, not a god). The old stereotype that a PC is a beige box running Windows 9x/ME has to die. Windows XP is just as stable as Mac OS X. If Windows XP or Mac OS X seem to be crashing for no damn reason, then there's either a problem with the hardware or the user.
 
Originally posted by Independence
I hope you were being sarcastic with that comment.

Seriously, no computer is perfect (yes, a Mac is a computer, not a god). The old stereotype that a PC is a beige box running Windows 9x/ME has to die. Windows XP is just as stable as Mac OS X. If Windows XP or Mac OS X seem to be crashing for no damn reason, then there's either a problem with the hardware or the user.
Click to expand...
If you can't tell that I was being sarcastic then you need help..
 
Originally posted by Independence
I can never tell if a human is being sarcastic due to the many masks they hide behind.
Click to expand...
haha.. well anyway... :rolleyes: You should be able to tell just from the way I wrote it. But since we're all hiding behind masks, then maybe not.. :rolleyes:
 
Originally posted by Independence
You've got a real attitude problem.
Click to expand...
Look.. I apologize I was kidding around... I really find if funny.. But you don't and I don't want to hurt your feelings.
 
Irrelevant benchmarks again !!!!

These Photoshop, rendering and 3D tests only go to prove the lack of expertise in optimising apps for the G5.

If you look at Apples performance figures you will get a more accurate overview of the potential of the G5 and not a test of how well software is written. I understand that we can only use what developers create and if Adobe creates a faster version of PS on the PC then it's obvious that the PC benchmarks are going to be faster.

If that's the case then these threads should be titled:

Adobe's PC dev team VS Adobe's Mac dev team and not P4 V Athlon vs G5

For a more accurate measure of actual CPU performance you need to look at highly optimised code libraries available on All platforms. You need to find code that each platform vendor has had, the time, money and expertise to make their respective CPU architecture looks as impressive as possible.

A good set of code libraries do exist in the form of FFT or fast fourier transforms. These highly optimised code libraries and algorithms are at the heart of every decent app used for video, audio and imaging. Each vendor has now had the time to fully optimise their own libraries to compile and execute as fast as humanly possible. Each vendor is allowed and has used every trick up their sleeve to make their CPU work as fast as possible.

And you will agree that the benchmarks are very surprising indeed. Here is an overview of the results that are linked to below (from slowest - fastest):

These are peak results for double precision 1D transforms measured in mflops

Single 400Mhz G3____________________415
Quad 500Mhz PIII____________________500
Single 500Mhz Ultrasparc IIe____________810
Single 733Mhz G4____________________985
Dual 833Mhz Alpha Eu6_______________1600
Dual AMD Athalon XP1700_____________1650
Dual 1.4Ghz AMD Opteron 240 - 32-bit____1730
Dual 1.4Ghz AMD Opteron 240 - 64-bit____2075
Dual 2.0Ghz AMD Opteron 246 - 32-bit____2400
Dual 2.0Ghz AMD Opteron 246 - 64-bit____2900
Dual 2.2Ghz Xeon____________________2900
Dual 900Mhz Itanium II________________3025
Dual 2.8Ghz Xeon____________________3900
Dual 2.8Ghz Xeon New Code libraries______3900
Dual 2.0Ghz G5______________________4000

These are peak results for single precision 1D transforms

Dual 2.0Ghz AMD Opteron 246 - 64-bit___4300
Dual 2.8Ghz Xeon___________________6900
Dual 2.0Ghz G5____________________10000

These result make it perfectly obvious that the G5 under highly optimised conditions is by far the superior CPU architecture. The full benchmarks including all source code used is freely available from the following site FFTW . You will also see that this site has NO affiliation or preference for any particular platform. Their goal is to simply create the fastest possible FFT libraries for as many platforms as possible.

From the above figures the Dual G5 is:

144% faster then the Dual Xeon and
232% faster then the Opteron .

These results also have to do with the fact that Apple has produced the fastest FFT code library currently available anywhere, fullstop. These Apple code libraries are also the reason why FCP, DVD Studio Pro and Apple's other Pro apps are the fastest in their respective fields. If Apple were to release a competitor to PS I would stake my life on the fact that it would trounce any crappy Adobe version, even running on the fastest PC available.

So please stop pretending that these other published benchmarks are anything more than a test of the software and how well it is written.
 
zoetropeuk:
Please do not say that using the Apple site itself is nonBiased when it comes to accurate tests. I already posted a link proving otherwise. BTW where are you getting your numbers for all the processors? Barefeats is also a bit biased after more extensive research leanning only towards Mac. I wouldn't link sites to any Microsoft area if they ran a benchmark cause it would also be as lame. I can't think of one program that will run equal on both platforms.

Its said the G5 is getting its next boost in speed in the summer so if we say June then that would mean if anyone buys a G5 now then they will have the same tech for the same price as the time when it was introduced in August whereas the competition introduces faster speeds for their processors every 3 months. This really doesnt matter that much but, I am one that isnt interested in buying tech thats almost a year old if I am to pay top dollor for performance. I cant tell the speed diffrence when comparing an amd 3200xp from a p4 3.0c but, I can if we were using a p4 3.0b which was introduced 7 months prior to the 3.0c.

virividox:
What can't you do on a Win Pc that you can on OS X?
 
Haxial is not an un-biased site either. Benchmarks, like statistics, can be made to show whatever you want.

Apple is marketing their machines and is using data that shows what they want. People should be checking with Veritest, not railing at Apple.
 
Originally posted by Oblivion
Veritest is also a fraud.

Reread the link I first posted when i entered this forum. Ill repost it just so you dont have to search.

http://spl.haxial.net/apple-powermac-G5/
Click to expand...

They were proven by most of the Slashdot crowd to be very wrong. Simply read the test results posted by veritest and you can see that most of their main points were incorrect.

Now as for this pissing contest. It's probably a good idea to end it on this point:

An IBM PC running x86 architecture is in the end going to be faster than a Mac simply because of the choice of chips, motherboards, ram, and other I/O. This of course does not mention the wads of money behind the Wintel monopoly.

I use a Mac for two reasons:
- A PowerMac G5 is a compiling beast
- A PowerMac G5 runs Mac OS X

My personal prefernence. I don't run it because its the world's fastest personal computer. I have a PC for games but I get my real work done on my PowerMac.

At the end of the day I will use what gets the job done. I switched from Windows because Mac OS X got my job done faster. I will switch back to PC if it does my job better, but right now Mac OS X is assisting me more than Windows does.

When Longhorn finally comes out, I will sit down and use it and compare it to Mac OS 10.7 and make another decision.
 
The link is in my post...

Originally posted by Oblivion
BTW where are you getting your numbers for all the processors?
Click to expand...

but if it's not obvious enough then here it is again:
http://www.fftw.org/speed/

Have fun discrediting these results :D

Don't get me wrong, I'm not trying to say that the app's benchmarked aren't faster on a PC. That much is quite clear unfortunately. But it simply doesn't prove that the underlying architecture is faster. It simply proves that the apps ARE faster.

Now the results that I've posted do prove without a doubt that the dual 2Ghz G5 is the fastest machine currently available.

Like I said in my above post, these optimised libraries form the core processing ability behind most major video, 3D and image processing apps. If Adobe were to use Apples vImage and vDsp libraries in their apps then the Photoshop benchmarks would be very different indeed.
 
Those tests you wre comparing are interesting on a number of points. As already stated... macworld/pcworld did not use uniform tests. This is probably the only place where you will see results of the G5 have its ass kicked.
When you look at the tests its obvious why
Adobe Premiere was killed years ago on mac. Last i heard it was a'Classic ' application ... non SMP aware (so no benefit of dual cpu's). So they compared an unoptimized old version of premiere on mac... made the mac run it in essentially emulation mode (classic is emulation mode ..is it not?) and the app is still non SMP in theversion used on mac. They then compared the performance against the latest OPTIMIZED version of premiere on the PC. And from this they deduced that the G5 was twice as slow as the competition!!! Thats some journalism. For a start they didn't tell readers that it was running in Classic mode (emulation)... they didn't tell readers that whenever you run an app in emulation mode on mac, pc or any other platform there is a big performance penalty. They didn't tell readers that they were comparing version say 5 on mac with latest version say10 on pc.
Want to dig deeper...... Look at the MP3 encoding tests. Spot the flaw there yet? Well here's the problem... they didn't use the same mp3 encoding app on each platform. Hell the different apps use different mp3 encoders. Music match on pc beat out the competition. Im not surprised. its mp3 codec is tuned for speed not quality. Ask the guys over at hydrogen audio. They should have taken a cross platform mp3 encoder and compared results with say LAME. Just to reiterate what im ranting about... you can encode mp3 nearly 3 times faster on my Opterons with musicmatch than you can with EAC and Lame.
Looking at the other tests they used.... Games... well the mac is never gonna compete properly with games... regardless of a 10GHz G5 i would imagine PC's will have faster games... purely from the porting issue involved with games to the mac. Most games ported to the mac are sloppy ports and completely unoptimized. Regardless mac's are not bought for games. However at the very least when benchmarking games they should have used the same video card in all machines. They didn't. I would imagine that performance in the benchmarks run would be very different now if they were to rebench since apple has released new vid drivers in X.3.2 which saw huge FPS increases across the board in games.
One could then point out recent bench's over at barefeats which saw the G5 kicking booty. These are probably more credible results because at least they chose modern apps that are cross platform and compared. The G5 won all tests bar the one!
If we wanted to make things even more interesting we could point out that already the G5 is running in a 'worst case scenario' .... radically new architecture, with little compiler support except from XLC and XLF. It has very few apps with any kind of G5 awareness. Meanwhile AMD's Opteron automatically leverages existing Athlon optimizations since its essentially and athlon xp with SSE2, x86-64 and memory controller. Intel has made such a big push of SSE2 in pc apps that both P4 and Opteron automatically take advantage of this bonus. In terms of comiler support there is lots optimizations in GCC presently for Intel and AMD. Much more so that there is PPC support. Still in most tests the G5 keeps up or beats both.
Now take a look at recent results of apps that have merely been recompiled with XLC and XLF (IBM's G5 compiler)... performance in a lot of apps went up 2 to three times the performance of the GCC 3.3 'G5 optimized' binaries. Great example was the Jet3d program, and the Extreme Fluid Dynamics. With a recompile with XLC, the resulting binary ran 3 times faster than with GCC3.3's most settings.
Point being ... G5 is presently completely untapped and only now is see compiler support that is good enough. I reckon hence forth we will be seeing much bigger gaps in performance from recompiled apps on G5 compared with x86 equivalents.
If like me you use Open Source software and are willing to recompile apps whether for x86, or PPC or other platform, then without any doubt the G5 IS the fastest workstation or desktop you can buy sub 4000 or 5000 pounds. And probably the chip/platform with the most performance to be unlocked over the coming months.
 
Oblivion,

If you are cutting DV only yes you can find PC solutions that cost less than Mac solutions. But if you need to work with uncompressed SD or HD video the PC options become fewer and more expensive. When working w/SD or HD media a Mac + FCP is actually on the lower end of the price scale.

All NLE's, at one time or another, drop frames for a variety of reasons (most of the time resulting from user error). I've never heard of FCP having any problem of dropping frames more than other NLE's. On any given day I bet I could go to a support forum for any NLE and find posts about dropped frames. And of those I find I bet the vast majority of them are related to incorrect settings and/or storage/HDD issues.

Why did you bring up Flame and Inferno when others were talking about FCP? Flame and Inferno are very hi-end, very expensive FX/Compositors. FCP is an NLE w/above average compositing ability. Apples and organes man. I should hope Flame and Inferno are better compositors than FCP is.

And since you brought up FCP in the professional world. Starting a couple of years ago I noticed post houses replacing their Avid Xpress' and/or Media Composers they used for off-lining projects w/FCP and just keeping a Symphony or Smoke/Flame set-up to finish them on. Off the top of my head the TV show "Scrubs", the movies "Rules of Attraction", "Full Frontal", "Cold Mountain", and "Intolerable Cruelty" all were cut/are being cut on FCP. The Coen Brothers ("Intolerable Cruelty") have already said they plan to edit their next feature on FCP as well. And FCP has become the Indie/documentary filmmaker's weapon of choice.

If you think a faster machine inherently makes a better NLE you are mistaken. Speed may be the Queen but software and stability are King when it comes to NLE's. You will still find a surprising number of 5 or 6 year old PowerMac 9600 based Avids still cranking out your favorite commercials, movies, and TV shows. Needless to say those systems are still around because they are fast.

/rant


Lethal
 
Yea hmm Im going to have to dissagree greatly on this site.

http://www.fftw.org/speed/

First.... These are purly synthetic benchmarks which prove nothing.

Second.... If the Dual Xeon 2.8 is doing well against the G5 then why wasnt the dual 3.0 used? I mean are we just going to exclude the 3.0 all together even if the G5 still wins in this bogus test? After all we are taking a test of "current" processors aren't we?

This looks like another of the many biased sites used to be in favor of Mac.

As for Barefeats... Ill save my reasearch and just quote a little somthing from a member from this site.

cr2sh
"I thought we decided to ignore everything that barefeats has to say? They are not a reputable source at all, their tests are flawed and they have little metadata at all.... why even bother?"

PCWorld Benchmark.. well thats also another bogus test just like the rest of the bunch.
 
Originally posted by Oblivion
First.... These are purely synthetic benchmarks which prove nothing.
Click to expand...
The biggest problem is your lack of intelligence and complete misunderstanding with regards to the benchmarks I linked to.

The FFT benchmarks test the absolute processing performance of the CPU that they are tested on. They take away the variables that are introduced when comparing apps developed by different people to run on different platforms. In simple terms these routines level the playing field. So how are they synthetic ? These core routines are used in some guise in every video, imaging or audio app on any platform you wish to compare. So these results are extremely relevant to the true potential of each platform.

If you spend the time to analyse the results you will see that each chip manufacturer has supplied their own highly optimised FFT routines to be benched against.
If Intel and AMD can't fully optimise the code to fully exploit their own CPUs then who can?
Originally posted by Oblivion
This looks like another of the many biased sites used to be in favour of Mac.
Click to expand...

How is this site biased in favour of Apple. Anybody can download the code and run the benchmarks for themselves. The developers don't even use Macs. They were supplied a Dual G5 and simply ran the benchmarks to compare their own code.

They are not out to prove that one platform is better then the other, the simple goal is to compare their own FFT routines to the other freely available and vendor supplied code. They even freely admit that Apple's own FFT routines are the best in the industry. This to me shows a completely unbiased attitude towards Apple and their own results.
Originally posted by Oblivion
Second.... If the Dual Xeon 2.8 is doing well against the G5 then why wasn't the dual 3.0 used? I mean are we just going to exclude the 3.0 all together even if the G5 still wins in this bogus test? After all we are taking a test of "current" processors aren't we?
Click to expand...

Well until somebody supplies them with a Dual 3Ghz Xeon then there is not much we can do. They haven't deliberately excluded it from the test, they just haven't had the opportunity to run the benchmarks yet. The dual 3Ghz Xeon is only going to be 10-15% faster anyway so overall the dual G5 still wins hands down.

You just have to be man enough to face the facts.
 
"The FFT benchmarks test the absolute processing performance of the CPU that they are tested on. They take away the variables that are introduced when comparing apps developed by different people to run on different platforms. In simple terms these routines level the playing field. So how are they synthetic ? These core routines are used in some guise in every video, imaging or audio app on any platform you wish to compare. So these results are extremely relevant to the true potential of each platform."

That's true but, it ignores memory access times and the fact that operating systems do things differently. If your benchmark is aimed towards finding how fast a system does office tasks, the best benchmark is how fast office works. FFT benchmarks are a ****ty way of determining which system runs video games the best. That's why there are so many benchmarking utils aimed at certain things, such as 3dmark.

Cross-Platform benchmarking with an fft is insane. Any realistic software program is so far from the hardware that it won't mean anything. Not to mention compilers

It accurately measures how fast a given system can do an fft, i'll give it that but, real world performance, blabla
 
Originally posted by Oblivion
Veritest is also a fraud.

Reread the link I first posted when i entered this forum. Ill repost it just so you dont have to search.

http://spl.haxial.net/apple-powermac-G5/
Click to expand...


Read the post above you...

Haxial is not an un-biased site either. Benchmarks, like statistics, can be made to show whatever you want.
Click to expand...

Apple can maniupulate... Haxial can do the same.


But the fact is, an independent firm has NO reason to lie.

At all.


End of story.

GOOD DAY SIR.
 
I agree that some sites are trying to manipulate the results to suite their needs for the benefit of what they like over the competition. I've yet to find an honest Apple vs Pc site that doesnt lean in 1 direction everytime when clearly there are claims elsewhere to prove otherwise.

As for Hex...
Lets exclude what he claims to be faster then what along with the results he provides in his comparison and only include the many facts that he provides with evidence that Mac manipulated the test setup. That in itself is enough to convince me that Apple and Veritest are both frauds for displaying inaccurate results.
 
LeathelWolf:
"But if you need to work with uncompressed SD or HD video the PC options become fewer and more expensive."

How do you figure that? I not only have the same options as you in HD's but, the ability to add more then 2 hds making my total uncompressed video storage more widespread. Like having 2 250gb hd's is not enough as it is already but, I do have that extra option unless theres somthing I don't know about.

"When working w/SD or HD media a Mac + FCP is actually on the lower end of the price scale."

Now we are going for best bang for the buck. Ok I built my pc around the 1k range which is as follows..

P4 3.0c = $270
1gb Cosair 3700ddr ram = $280
Asus P4p800 Mobo = $130
Antec Tower w/PSu extra 120mm fan = $80
2x250gb hd WD 7200rpm 8mb cache = $225x2= $450
Sony Dvd burner = $150
ATI 9800 AIW = $339

This is a total of $1,699 w/o Montior all www.Newegg.com prices. Now we add Avid Xpress Pro which is $1,695 and we get a grand total of $3,394w/o monitor.

The entry level G5 1.6ghz stock is $1,799. Obviously we will have to add ram since more ram is better for video editing and give it 2x250gb HDs. Now the price is $2,824 w/o monitor. Lets now add FCP $999 and we get $3,823

Im obviously getting more bang for the buck in the same Pc category. I even have a better quality built machine. I can do anything you do as far as Video editing plus the extra ability to do 3d animation with my tweaked Radeon. You will have to add $350 to the finnal price to utilize the same video card as me. Where is the analog svideo input on any of the radeon cards offered for any G5? I also get that extra bonus.

If you are to use the Mac for the comfort of OS prefrence then thats understadable but, just know how much of a price diffrence you will be paying. I obviously showed above how XP is Stable while doing multiprocessing under the heaviest cpu stress so please dont attempt to belittle XP in terms of stability. Im still more then happy to send anyone a real time video if they arent convinced.

"Why did you bring up Flame and Inferno when others were talking about FCP?"

Well I just thought I would add an added feature the x86 has over the Mac in terms of professional work.

I will also add the Ease of use with FCP with its GUI just like OSX. That doesnt make it any better then what I can do.


Rower_CPU:

This is Tony on day 1 going against the claims....

http://www.theregister.co.uk/content/39/31405.html

And this is Tony the next day with a sudden change of thought.....

http://www.theregister.co.uk/content/39/31416.html

Makes you wonder who this guy is really working for.

You may want to look at this site which also claims otherwise.

http://www.extremetech.com/article2/0,3973,1136018,00.asp
 
Oblivion-

Tony Smith is reporting stories. He's not changing his opinion. How about you deal with Joswiak's statements instead of attacking the person reporting them?

That other site is a PC-centric site and does not deal with Joswiak's statements either.

For someone crying over an Apple-bias in the G5 benchmarks, you're doing a really bad job finding objective sources that go against them.
 
My Mistake on Tony.

"That other site is a PC-centric site and does not deal with Joswiak's statements either."

Are you going to deny the Spec numbers provided on this site which go against the numbers that were provided in favor of Apple against PC? If so then this is turning into a debate about who believes what.
 
Status
Not open for further replies.
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.
Top Bottom
Back