Here are some of the most powerful PCs vs the G5. The G5 does very well on cpu crunching, but not on graphics tests.
Apple PowerMac G5 vs. PC
About The Author
Eugenia Loli
Ex-programmer, ex-editor in chief at OSNews.com, now a visual artist/filmmaker.
Follow me on Twitter @EugeniaLoli
link works for me
also…
“even 1GB would have produced the same results as 4GB. But for the record, this was the configuration:”
wtf? once the windows based systems use up their 1gb and 2gb of RAM they use virtual Memory on the HD so hows that a fair test? Rendering on my system usually takes up my 1Gb of RAM plus 512mb of Virtual memory and VirtMem is sh*t loads slower than RAM so this test isnt fair
let us wait, until they put scores of dual 2.6Ghz opterons there…
Now here’s a subject that unfortunately always brings out the worst of the x86 and PPC camps. I wonder what the subject of the first argument will be about this time…
Consider this a pre-emptive “Can’t we all just get along?” Play nice..
😉
Now here’s a subject that unfortunately always brings out the worst of the x86 and PPC camps. I wonder what the subject of the first argument will be about this time…
Consider this a pre-emptive “Can’t we all just get along?” Play nice..
😉
a dual G5 @ 2.5ghz is a bit more expensive then a AMD64 3500+ system…
this benchmarks make no sense… as usual =:P
First off, the AMD 64 box had 333Mhz ram versus the 400Mhz ram of the other boxes.
Secondly, how can you compare single versus dual CPU boxes? That’s not a fair comparison. It’s like saying our horses are so much faster than your horses, but you can only race one of your horses against two of ours.
They never explain HOW they ran their tests. Where are the sources to the benchmarks? People should be able to idependently verify the results at home.
If this was a high school science project, these guys would flunk.
If you want the fastest OS X box you buy a Power Mac, if you want to run another OS on another platform feel free… I’ve never understood this Apple versus x86 thing… oh well can’t say I care one way or the other.
/me goes back to playing World of Warcraft on a Power Mac Dual G5
to me this test is cool it just shows that pc as some advantage that mac dont and thats it so if your happy with the performance of your pc or mac than stick to if not well then you may want to conssider the other side if the fence and i agree with risc i never understood whats the point of arguing for it there both capable type of machine with each there own set of flaw
I think the main point here would be the “Bare Fact” lies.
They used a 3500+ Amd 64,why not the 3900 or the FX.The amd and even worse the P4 had less memory,3gig for the amd and 2 gig for the P4,this doesn`t effect rendering preformance?
Was the dual g5 running on the panther 64bit optomised OS X?
And if it was ,surely the tests for the 64bit amd was using a 64bit optimised Windows?
This well thought out test is probably as good as pitting a MacLaren F1 against a Fiat 126
The last paragraph of the article says :
“Xi Computer let us test the Dual Opteron 2GHz a year ago. We pulled those results from this page at their request because, in their words, “that was a bad motherboard.” They’ve invited us to come test a newer Dual Opteron 2.6GHz with PCI-Express bus. We’ll post those results in a few days.”
So, be patient and stop thrashing Rob Morgan.
I am all for comparing single cpu versus dual cpus. The best benchmark is the one where you see power based on the dollar spent. That a dual 2.5Ghz G5 wth 4gb of ram beats an Athlon 64 3500, or a machine that probably cost 1/3 the price, doesn’t surprise me much. If you want to see what single cpu PCs can do to dual cpu Macs, look here: http://pcversusmac.com/benchmarks.htm If you got the newest top end end Athlons, Opteron’s, Xeon’s, and Pentium 4’s and coupled them with the newest motherboards with ddr2 memory, PCI Express, faster busses, top end video cards, etc., I’d be willing to bet that we would see them thrash the dual 2.5GHz just as the above beat on the dual 2.0GHz last year.
Would be interesting to see a more comprehensive suite of benchmarks.
And it would give us all a bit more trust if we got the exact specs for all the machines.
Finally, if you are benchmarking the two most powerful and very expensive Powermacs made, please pit them against some PCs that are a bit higher up the food chain. I know, I know the Opteron’s are coming supposedly, but Bare Feats is notorious for saying they will post something and then it doesn’t come or comes much later. They said this test would be done back in September and here it is February and it’s finally going online.
This well thought out test is probably as good as pitting a MacLaren F1 against a Fiat 126
Whell, why not. The Fiat would probably win on usability.
…because no one can agree on the terms of the testing. You readers will never be happy unless it is your particular PC or Mac that you have at home that wins the freakin’ tests.
Cripes.
…because no one can agree on the terms of the testing. You readers will never be happy unless it is your particular PC or Mac that you have at home that wins the freakin’ tests.
Cripes.
_____________________________________________________________
Sorry Tuishimi,
To an extent I have to agree but also, I have to disagree. For the hardware config’s,
1) Same memory should be used, PC3200
2) Same drive types S-ATA vs P-ATA (same speed and potentially same drive).
3) Were both OS’ 64 bit?
I have no issue w/PPC winning these test. However, I believe that the testing of the PC hardware platform varied too much.
With less memory like the P4 or slower memory with the AMD will cause some potential issues. The disks could have started swapping. All, I would like to see is better hardware line up on the Intel/AMD side of the house.
My interest sits with in the performance differences between Intel not AMD not between PPC vs x86. So, I basically could care less if PPC’s win the race. There is nothing wrong with Apple (hardware or OS).
But your right… No one will be happy with any benchmark. Because people cannot decide.
One more benchmark about g5 vs pc:
http://www.pcworld.com/news/article/0,aid,112749,pg,8,00.asp
The graphics issue is very clear, the Windows drivers are far more optimized than the OSX ones and so are the games. If you want to play games get a console.
The most interesting fact about the CPU performance is that you get the same horsepower using 1/10th of electricity. These Athlons and Xeon CPUs are a total joke, using as much as 75W per cpu. AMD64 is a nice arch, but it’s too little too late, PowerPC got it right, 10 years ago. It’s time to let the IA-32 crap finally die. When you have the source for everything available all you need is a recompile.
I couldn’t disagree more, and I say that for two reasons:
– No matter how similar the configurations are, someone will always claim that you are comparing apples to oranges (eg. bad choice of applications, poor compiler, etc.).
– People tend to be more interested in performance per dollar for real-world applications (ie. they know that benchmarks are easy to fudge).
The best way to go about this, IMHO, is to choose machines with similar prices from large vendors. Do not customise those machines in anway, either once you receive the machine or when you order it. Test it with a bunch of popular applications. List the results. List the bottlenecks. Publish the results in an appropriate venue (ie. don’t pretend that results from graphics applications will be suitable for scientific computing needs).
Oh, and try to find someone who doesn’t have a bias one way or the other when designing the tests.
These comparisons on these different hardware platforms are rather pointless, let’s wait a decennia until both platforms run a mature OS.
Comparing is good, but if you have to slant the tests, be honest.
Really… comparing a dual processor unit to a single processor??? Never should have been done. It is like a truck pulling contest, with the same RPM’s and gear ratios, a 4 cylinder cannot compete with an 8 cylinder. In the same way, a dual processor unit just has more power (it is like doubling your resources).
And worse yet… 4 GB of 400 to 3 Gb of 333 ??? ??? ???
OK, I know you did not pay for the units, and ran a test to the best of what you had, but present it and take in account the differences. Make sure the readers know about the disadvantages, don’t hide them in the results like you did.
PS are you a Democrat, you sure tell the story like one.
🙂
“Apple PowerMac G5 vs. PC” again, don`t you tire that…
( We want a G5 vs Opteron, and the fully fatcs or functionallity on Linux + AMD64 vs G5 + OSX )
And not one single scientific application.
Besides http://smc.vnet.net/timings50.html which has Mathematica benchmarks and shows a Dell Precision 650, 4X3.06GHz Xeon, 512KB L2, 4GB, being killed by an AMD xp-2700, 2.17 GHz, 333 FSB, 1 GB, win-xp-pro, does anybody know other other sites that compares different hardware platforms using number crunching scientific software rather than artistic software as benchmarks?
If you read the entire article, you’ll find the reason the Mac does so badly is that the programmers are lazy when they do their ports. As Quake Arena shows, a Mac optimized version will smoke the PC version.
Quake also makes use of a 2nd processor. I wish more games would do so – again it all comes down to programmer laziness. The rush to meet deadlines wins out over quality code every time.
But it works as an end product comparison.
No Opterons in the mix ?
Guess they didn’t want to make the Intel’s and G5s look bad across the board. *shrugs*
I did find it interesting to see that the dual 2.5 Ghz G5 outperformed the dual 3.4 Ghz Xeon. The Xeon CPU’s are supposed to be Intel’s best chips. I’m not sure about the price tag they have, but i’d guess they aren’t much cheaper than a dual G5. And with that in mind, it’s nice to see the G5 outperform the Xeon on all but one test (not including the 3D tests).
It would have been nice to see a G5 like the one in the iMac (as in ONE cpu) being compared to ONE Pentium4 and ONE AMD64. I wonder if a single G5 cpu would also outperform a single Pentium4 or a single AMD64 cpu. Right now i would assume that the G5 would win because the dual G5 beats the dual Xeon, but i’d love to see some benchmarks that would prove me either right or wrong.
What methodology was used to produce these results? Where are the details, especially when the comment was made about video cards not making any difference? Where is the data to support that conclusion.
If you are going to go through the trouble of benchmarking systems, then you should have a sound methodology (on paper) and full deatils as to how the tests were conducted (OS and application installation, twaeks, etc.). And the information should be available for independent review by anyone. If not in the article, a separate document should be made available with the testing information.
the use of a 3d test that uses code not optimized for the G5.
the use of hardware that is less powerful than the G5… but then again, did this guy ask for a companies top end machine or something similar?? if so, it is less a test of the hardware architecture and more a test of a companies product against another companies product.
Waht about the Opterons?
What you want to with single vs dual comparation?
What abouts the 64 bits OS instead old and sucked Windows?
Sorry, but this kind of comparation show up everday.
I just will agree when somebody with the real know how do it. (Like THG).
Sorry…
Secondly, how can you compare single versus dual CPU boxes? That’s not a fair comparison. It’s like saying our horses are so much faster than your horses, but you can only race one of your horses against two of ours.
ROFL!! That’s the most pointless and dumb analogy I have ever read!
a good term of comparation would be ubuntu
it runs on powerpc and AMD64 and provides a lot of productivity aplications
that would be a good way to compare hardware platforms
i would like to see a 2500$ opetron workstation vs. a 2500$ power pc workstation. now THAT would be realy interesting !
does anyone seen something like that before ? i’m realy curious
64 bit OS? why OS X is not 64 bit. why should windows be? besides that the OS really does not need to be 64 bit… it will just slow it down.
Why are people so upset over this? I find it just to be interesting reading…sheesh.
The G5’s do so well in performance because they have dual 1 Ghz + frontside buses, and dual FPU’s. This allows for the processors to always be fed data.
The G5’s do terrible in the graphics tests because OpenGL is optimized for X86, and not PPC. Also a lot of games for Mac are converted from DirectX to OpenGL, which also brings slowdowns.
Unless the OpenGL orginazation can release a specific backwards compatable PPC version, graphics performance will always suck on PPC.
No Athlon 64 or Opteron, not to mention a single G5 for refrence.
I think it is impressive that one man, without help from the computer manufacturers has been able to do this. Still without a lab with the full spectrum of hardware, and full control of that hardware, turns this into a rough estimate at best.
Many of you stated that it’s an endless argument because all we want is to say “our boxes” are faster and that’s it. I can’t say it’s wrong but at least give ms something reasonable.
In here, I use the word reasonable because it’s really funny to compare with different hardware and say “o, the dual CPU one is faster, so go for that”. Why don’t you put 127 linked 486 with optimized drivers and software comparing with a dual G5, if that 127 linked 486 wins, are you going to tell us to buy that 127 linked 486 as ONE computer?
This review is just not professional to me. Like Pcworld or PcMag, they tend to like Intel products even thousands of benchmarks show that AMD64 is faster than P4, I still don’t know where they can find software showing that P4 is “outperforming” AMD64. Anyway, my point is that even those magazines tend to show their good relationships with Intel, they still did hardware tests with same configurations including software.
In this case, while you can’t run Windows on Mac or OSX on x86, you can run PPC linux on Mac and Gentoo on x86, at least then, you can see both of them have some suitable software to do the test.
This review is not pointless, but it is done in incorrect way or the author or company has good relationship with Apple but not “orange”.
😛
line: “at least give us something”
Sorry
By Uno Engborg
“This well thought out test is probably as good as pitting a MacLaren F1 against a Fiat 126
Whell, why not. The Fiat would probably win on usability.”
Because the test was about CPU benchmarking NOT usablility.
Forget about the exact numbers or saying that this is comparing apple to oranges. That isn’t true. A scientific benchmark should try to compare same things, but on the real world, you just compare, daily tasks, things you do, in computers you can buy and you might buy. If you want to test the real hardware optimizacion, then an OS independant test would have to be performed on each bus and hardware component.
Perhaps the G5 is moving data faster along its Front Side bus, and therefore, Quake III works faster, you never know. Who cares?
And… add as many Operton CPU’s as you want. You can’t have OS X.
“Perhaps the G5 is moving data faster along its Front Side bus, and therefore, Quake III works faster, you never know. Who cares?
And… add as many Operton CPU’s as you want. You can’t have OS X.
”
Sorry i thought that the fsb for the dual g5 was 1000mhz,the same as Amd64`s oh and wait same as some of the P4`s.
And as for the OS X,seems to work fine for me on pearpc,i even selotaped the two extra buttons over to have that genuine mac feel.
The funniest thing about the past postings as per usual is that almost all of them are answered on the benchmark site.
1st, he can only test systems he has access too, it’s not a big mac conspiracy, he is waiting to do the Opteron tests ASAP.
2nd, he explains why the disks performance (ATA vs SATA)* & the graphics card performance (AGP8X vs PCIxpress)* are not an issue on the CPU tests because the CPU has been isolated, in the testing metodology.
*the performance difference between AGP8X vs PCI express and ATA133 vs SATA; has been proven repeatedly recently to be a non issues. ATA and SATA hd’s right now are the same drives with different interfaces neither of which are utilized at even 50%;
3rd, He didn’t have access to an AthlonFX, Opteron, Magic carpet, etc, with DDR400 or DDRII memory so he couldn’t test any of these. (ps AMD and Apple do not support DDRII.)
He is just one bloke on a mission from god to test machines around him and post the results on HIS PERSONAL site.
You want him to test your dream config, well then send him any hardware/money you want and he will do it.
4th, He explains his testing methodology throughout the site, please read before bitching.
5th Please stop making teenage boy bs fantasy statements about the price of a Mac vs a DYO PC; THEY ARE NOT THE SAME.
IF you are going to speak of the price of a Mac vs a PC it has to be between a G5 and equivalent workstation from a top tier vendor, ( HP, DELL, IBM, Sun) or at least a reputable vendor (Alienware) with similar finish as APPLE. And this workstation has to have all of the same options. (Firewire800/400 and gig ethernet, and 802.11g and bluetooth, etc)
(Apple is not in the hobby market, they dropped that in the early 80s with the Apple III.)
The Mac will never compare to a BYO machine, which inherently will have a better initial price/performance point.
The BYO is designed with this # in mind.
BUT…
It will fail in the price per effort front –researching parts and building the awesome machine– and will also fail in support costs –when something in the system goes wrong, you will be the sole person responsible for this machine, no warranty (apart for individual parts) no support apart from your time, and since you built it anything you did wrong will be your fault (like forgetting to plug in the fan therefore frying the amazingly hot cpu).
Again in long term price comparison, let say over a 2 year period, PCs generally cost more, and BYO cost even more.
6th
At no point does he say the MAC is inherently better.
So get over your conspiracy theories.
He is not out to get you.
He is not painting the MAC as the better machine.
7th PPC cpu is better designed, more efficient, and more powerful then both X86-64 cpus.
In the scientific community where # crunching counts and price/performance is measured in terms of electricity and heat, a 2GHZ PPC 970fx, and a 2GHZ Opteron are not in the same league AT ALL.
The following are Clockspeed + heat/power consumption + theorical GFLOPS
PPC970fx 2Ghz –> 24.5w, for 10GFLOPS
Opteron 246 2Ghz –> 70W, for 19GFLOPS
Xeon (prescott) 3.4Ghz –> 132W, for 24GLOPS
If you read the brief that Virginia tech wrote about it’s choice of Apple G5 based systems, it explains that the equation between the # of processors vs the power consumption vs processing performance, of the PPC970 is in a completely different league then a comparable Opteron or Xeon machine.
It might take twice as many PP970 cpus to equate the theorical GFLOPS of an AMD64 chips but this is not the whole story.
For each Opteron you can have almost 3 PPC970s, and for each Xeon you get 5 PPC970.
Because of the 970s faster cache/frontside bus (always half of cpu bus speed, as opposed to 800Mghz and now 1Ghz Hypertransport.)
The difference between theorical and real GFLOPS on the PPC970 is reduced significantly, where as an Opterons real GFLOPS is about 9.5 (vs 19 theorical)
The caches bus speed is very important in a distributed application.
So technically the ppc design is better suited for future domination where more and more computing is being distributed in more power conscious situations.
8th.
Neither the present osX nor WinXP are 64 bit oses.
9th a real comparison would be between a Xeon/Opteron/G5 all runing the latest Suse/Red Hat linux. Both in 32bit mode and 64 bit mode as this is possible.
10th I’m out.
“Besides http://smc.vnet.net/timings50.html which has Mathematica benchmarks and shows a Dell Precision 650, 4X3.06GHz Xeon, 512KB L2, 4GB, being killed by an AMD xp-2700, 2.17 GHz, 333 FSB, 1 GB, win-xp-pro, does anybody know other other sites that compares different hardware platforms using number crunching scientific software rather than artistic software as benchmarks?”
Mathematica is benchmarked here:
http://pcnmac.com/benchmarks.htm
Mathematica 5.0
AMD Athlon FX-51 2.2 GHz 572 sec
Intel P4EE 3.2 GHz 639 sec
Apple Dual 2.0 GHz G5 997 sec
The Mac is soundly trounced by single cpu PCs with slower hard drives. Almost all other specs are the same (memory, video card, etc)
These kinds of tests (G5 vs fastes PCs) came around from the moment G5 was released. First they were compared to the fastest CPUs there existed. Then to the newer ones. Then the newer ones. Then… and it just goes on, here’s another one. So, I’d say in 5 years these G5s will lag quite behind compared to the will-be new CPUs.
In other news, this disk we’re living on is actually round and it goes around the Sun.
do terrible in the graphics tests because OpenGL is optimized for X86, and not PPC
frag that
I just laughed my a$$ off for the day 
ignorant lameness rules, my friends 
As a programmer for about 10 years now, I can understand alot of what is being said here. I use x86 Linux personaly, OSX on a G4 at work (for video editing) and Win32 everywhere else. What I’m waiting for is a full blown “geek” benchmark. Somthing allong the lines of top-of-the-line Dual G5 running Gentoo with full optimization run against a Dual Xeon 3.xGhz running Gentoo against a Dual Opteron 3.xGhz Gentoo (against Irix on a dual Tezro? but I degress). I want to see someone sit down and optimize the life out of each of these machines and then look at the benchmarks. Hard Drives are not the same speed? Netboot, Load the entire system into RAM Disk and do your benchmarks that way. Not the same amount of ram? Cut the ram out of some of the systems and make them all equel. In the end you will be left with these factors: FSB, CPU power (Speed and otherwise) and RAM speed.
The question is, when I go to my boss for an upgrade and he asks me how I can get the most bang-for-the-buck. I want to be able to say unbiasedly that yes the system I quote to him will be the best his money can buy
I don’t care much about benchmarks, especially for photoshop comparisons. But I must say that after looking into it, a dualie Opteron is more expensive than a Xserve. Both with comparable specs. I was quite surprised to see this. Plus, I really like what the OS X Server software has to offer, although i’ve never used it. So maybe a bit OT, but thought I’d mention that.
The methodology might be there (if you can call it that), but specifics on how the various tests were conducted is not there. Let’s say I wanted to repeat this test, just exactly where do I start once I have collected the equipment and software? How do I set up the machines for the tests, is there anything enabled or disabled.
I like comments like this: “The boot drives varied in speed and interface from system to system, but the tests we ran and the way we ran them did not involve any disk activity.” And this is based on what? Just exactly how were the programs and data loaded if they were not read from the various systems’ hard disks? And where is the data proving this, sorry I just don’t take anyone’s word for it.
And I didn’t even know there was a second page until I found the link to it (site design leaves something to be desired). I also find it interesting how this guy links to AnandTech, look at that site, in particular this page:
http://anandtech.com/video/showdoc.aspx?i=2335&p=3
I would actually use AnandTech’s results because they list most (if not all) the technical information that makes this kind of testing valid. I am not just concerned with the “results”, I am as equally concerend with how they got the results. So it is NOT bitching, it is a valid point.
Just as biased as the benchmark on Apple’s website.Even if you could spend money to buy the fastest and the most allround system that exists the mac wouldn’t win.Where are real 3D benchmarks?No games tested?The AMD64 3500+ is a fine one,but why not test a dual 3500+ against the G5? or even a dual FX54.
Macs are for people who have to much money and are in a hurry,other than that a AMD64 could win in any other case.I would like to see some gaming benchmarks.Don’t get me wrong,the G5 is a very fine piece of equipment,for the lazy,money plenty in a hurry amongst us.If you would be able to build a PC out of the best individual parts money can buy the mac doesn’t stand a chance.For the > 2800 euro i could assemble a faster better PC.For those not able to,or simply don’t have the time to build one the G5 is a very good choice.But to the point,the test results are to say it mildly,flattered and not complete.For those who like to game a lot realy don’t need a G5 perse, a single AMD FX54 CPU machine would do just fine whith about everything you trow at it and much better when gaming.And you could save a lot of money.
After Effects rendering is very fast on the G5 and that’s what big movie houses look at. As far as the two high end machines tested, both around the same power, but the G5 does blow it away on most of the test from what I see. IBM has made a great processor and the others are following their lead.
Apple, Adobe, Newtek, alias, discreet takes advantage of Tiger’s Core functions(like core-audio, video) we will see even more improvements. Like real time results on every piece of software.
I don’t think that’s to far off. Apple is leading the way in those cocoa-apps that utilize the technology and customers will want Adobe and others to do the same once they have used an Apple made app that uses the Core funtions.
Somthing allong the lines of top-of-the-line Dual G5 running Gentoo with full optimization run against a Dual Xeon 3.xGhz running Gentoo against a Dual Opteron 3.xGhz Gentoo (against Irix on a dual Tezro? but I degress
that’s assuming it was a decent port of linux to PPC.
Not the same amount of ram? Cut the ram out of some of the systems and make them all equel. In the end you will be left with these factors: FSB, CPU power (Speed and otherwise) and RAM speed.
the systems should also be approximately the same cost. if the Mac comes with monitor, RAID, whatever, then set up something equivalent on the PC. if there is a several hundred dollar disparity between the systems, the results are meaningless to a purchaser.
Why can’t they choose decent processors for the benchmark? Honestly, the P4 and Athlon they chose are good enough in their own way, but are obviously going to bring up the rear compared to a top-of-the-line dual G5. The Xeon was the only one ever in the running in the CPU tests.
Where was the Athlon 64? Where was the dual Opteron?
They might have given the G5 a spot of competition – instead they’ve put in a couple of tired single processors. The P4 3GHz is a couple of years old now, and it and the Athlon can be bought for 2/3 of bugger all – they’re simply not in the same league as the others.
As far as I can see they’re only there to make the Mac look better because it’s got more to beat. To be fair, it does get the Xeon pretty easily, but Opterons piss all over them too – why can’t we see a battle between them?
photoshop is made for windows and optimised for windows then ported to mac os x
that’s assuming it was a decent port of linux to PPC.
I think we can assume that Gentoo on PPC is about as good as it’s going to get – it’s one of the few ways to compare without the OS getting in the way (although arguably that could be part of the benchmark anyway – depends exactly what you want to test).
Agreed about cost – that was exactly my point with the P4 and Athlon in that test. They’re practically budget CPU’s, whereas a dual 2.5 G5 is anything but.
For those not aware Alias’s Maya is ported to Linux, Windows and OSX where as Apple’s Shake is ported to Linux and OSX. In either case the benchmarks posted in the links below prove the G5 is not all that Apple hypes it to be. It’s also reason why consumers serious about animation, compositing, etc for visual effects are complaining about not having DCC cards for Apple systems. Unlike what some would lead others to believe the GPU isn’t the only factor when using such software.
Maya Render Benchmark:
http://www.zoorender.com/
Shake Render Benchmark:
http://homepage.mac.com/breadboi/shake/
How about lets see some Gentoo Linux v. Darwin Unix (OSX).
As many people have remarked, if we want a real comparison, rather than a fluff piece for cosy little trade fair faux presentations and to drive clicks to web sites through controversy, we need to focus on some proper benchmarks.
Benchmarks like LinPack and Sciencemark are essential tools that we should be utilising. These are pure benchmarks that should be understood as artificial examiners of architectures and valued as such.
For application benchmarking, we need Maya, 3DS Max, Lightwave, Cinema 3D and all the major renderers to be tested. We need to test a minimum of three audio and video encoders across all platforms. We need to test HTTP server performance and database performance across these platforms in structured ways, as well (see Anandtech’s server benchmarking procedure).
If we have a test suite of 30+ benchmarks, common across platforms, I believe that the outcomes will be evident to all and satisfactory.
I fail to understand how duplicating the benchmarks that *Apple uses for promotional purposes* is going to generate an impartial and valuable comparison, nor when using half a dozen or so benchmarks only. All x86 benchmarking sites would be embarrassed to display such a limited range of tests for differing platforms, even when examining only a single component of a system; so we need more here. We’ve been shown no methodology, no detailed breakdown of what tasks were performed by the aforementioned applications, with what settings, and no rationale for why the chosen benchmarks *ought to be insightful* and what they are designed to be testing. Why choose a certain combination of Photoshop filters as a fair benchmark of all computing systems?
Finally, I’d love to know how Photoshop is a good benchmarking setup in the first place, given that we have no code, nor understanding of how the filters used are actually constructed. For all we know, there could be a section of code:
if(architecture == “x86”)
{
wait(300);
}
I don’t believe that there is, but an authoritive study must be able to validate the integrity of such benchmarks.
oh, and P.S:
“After Effects rendering is very fast on the G5 and that’s what big movie houses look at. As far as the two high end machines tested, both around the same power, but the G5 does blow it away on most of the test from what I see. IBM has made a great processor and the others are following their lead.”
Nah… Big movie houses use clusters and render farms, with custom rendering code, rather than After-Effects. Your little local TV station and quite a few advertising design companies probably use After Effects, though. After-Effects really isn’t a good benchmark to predict high-end video processing, though.
“Secondly, how can you compare single versus dual CPU boxes? That’s not a fair comparison. It’s like saying our horses are so much faster than your horses, but you can only race one of your horses against two of ours. “
Can I just point out that this analogy doesn’t make any sense..
Two fast horses racing aren’t going to be any faster than one horse racing.. there’s just more horses, they will still go the same speed (all things being equal- of course).
It’s interesting how this character never tested a DUAL Athlon system. And what about the Athlon 64? They have certainly beaten the G5 in MANY, MANY tests.
Maybe because he didn’t have access to them? Who really cares anyway? You want OS X you buy Apple, you want another OS buy something else, none of this my CPU is faster than your CPU crud means anything in the real world.
doods
Benchmark all of your dream machines yourselves wtf.
HE IS NOT anand tech, and if you just read the second page he is not praising the mac in anyway.
what is wrong with you sensitive bastards anyway.
They man is not a full fledged lab..
he just shows you what he has found out.
damn.
READ before BITCHING at him.
quoting the site:
iXcomputer let us test the Dual Opteron 2GHz a year ago. We pulled those results from this page at their request because, in their words, “that was a bad motherboard.” They’ve invited us to come test a newer Dual Opteron 2.6GHz with PCI-Express bus. We’ll post those results in a few days.
But once again who cares OS X doesn’t run on Opteron – so NEXT… this really is just more pointless OSNews.com flamebait.
Apple has always lagged behind the PC in the graphics arena despite media claims to the contrary.
I love how their benchmarks list only the Xeon in the dual CPU config, comparing the P4 3ghz and A64 3500+ in the single cpu and not even bothering to list a single CPU G5 for comparison… OF COURSE a dual 2.5ghz G5 buries the 2.2ghz A64.
But it ISN’T an Athlon 64. It’s just a regular Athlon XP. If you’re going to compare a dualed processor to other processors, those other processors should be dual as well. Besides, all of the benchmarks I’ve seen, other than this one, compare the processors fairly. Oh yeah, what’s so great about OS X? I use it all the time and I don’t see anything special about it. Linux seems faster on my outdated machine than OS X does on a machine that came out the same year. Linux isn’t worse or just simpler.
Just because Apple tells you that their hardware is higher quality doesn’t mean that it is. Apple is known for their visual gimmicks. Just because aqua is pretty doesn’t mean the OS is better.
–the benchmarkers are supposed to pull one of the of the processors out of the G5? Give me a break. The fastest single processor Mac is only 1.8 GHz. You can’t get anything faster from Apple without it being a DUAL PROCESSOR unit. I’ve never met a PC user that used a dual processor rig, have you?
The only multi-processor AMD, INTEL rigs I’ve been in contact with are clusters and they weren’t playing Half-Life2 on those.
Oh I didn’t see. He does say that it is 64 bit, which makes no difference unless the OS is a 64 bit os.
I believe thi link is really interesting.
The tester is using what most users would be accesible to.
There is no need to try to force results one way or another… If we have more information about that, or about the other. Some people say the method is incorrect, etc.
However, this seems not to be a science degree in benchmarking.
There are PPC fans, Athlon fans and Intel fans, and at some point all have many reasons to claim one processor is faster than the other.
Could an Intel processor be faster than AMD processors? Yes. Why? Because the compiler that was use to build the application. This does not mean the AMD is a bad design. We just do not have information about compilers used to build the apps. A company can make a build that is faster for Xeon, because it is optimized for Intel, and it is not good on AMD. That’s very common on games because most of the market is on cheap Intel boxes and not on AMD64.
Today’s technology on compilers is complex even on x86. Using only the new AMD64 revised x86codec is better for AMD but it is not optimal for Intel chips.
The same happens on PowerPC chips. After the G5 introduction, IBM demonstrated that using its special G5 compiler you could get up 40% increase on speed, but most of the time, G4 processors would suffer using G5 optimizations… Why? Different processors. Having many different king of processor on the market explains many trade-off of applications development.
One thing is true. None of the programs tested were G5 optimized (Because of the date of creation), most of them not even heavily G4 optimized. Most of the programs tested were Intel optimized (Intel sells more).
Apple is using GCC compilers even to code the operating systems…
I think AMD and PPC are doing really respectable benchmarks considering most of those programs are not optimized for their architectures.
ONE thing about the people who claim about building your own computer is cheaper. Most of them, NEVER consider the price of the software included with Mac computers, which is the most expensive thing Mac users pay…iLive, the OS, iDVD, iMovie… They are high quality programs, and they are expensive to develop. You may have the FASTEST computer on Earth, but without the software it is not worth a penny. On Linux you can find great free programs… But they are not as polished on user interfaces and design, as those found on Mac, or even in Windows.
ONE thing about the people who claim about building your own computer is cheaper. Most of them, NEVER consider the price of the software included with Mac computers, which is the most expensive thing Mac users pay…iLive, the OS, iDVD, iMovie… They are high quality programs, and they are expensive to develop. You may have the FASTEST computer on Earth, but without the software it is not worth a penny. On Linux you can find great free programs… But they are not as polished on user interfaces and design, as those found on Mac, or even in Windows.
Agree,the PC or whatever you name it is only a dumb tool unless you feed it with sophisticated software and start using it.When you reach a certain degree of proficiancy the OS itself doesn’t matter anymore.You can make allmost anything secure,fast and stylish.Only thing that matters is the supported hardware you might need or desire to use and the applicatications avaible for your platform.The amount of money you would save by assembling an dual (socket939) AMD64 PC is significant.You know what what hardware is going in to the box,with x86-(64) you are not bound to any hardware vendor.Pratically everything is supported.The afore mentioned money saved could be spend on software *you* choose like the hardware *you* choosed for the mentioned assembly PC.
Unfortunately a lot of clients don’t have the time and desire to get to much involved with hardware and or the actual OS.A G5 or better yet any mac is worth considering in this referential fit.Everything works right out of the box and has a great deal of comfort.I wouldn’t say the OS is better but they have left little to think about.It’s like driving and tuning a car or being driven by the car.
My concern are those one-sided tests.Even if someone would only use the mentioned appications,the test still remains somewhat limited.For example a single AMD 3500+ isn’t a match for a dual 2.5GHZ G5.What about installing some Linux distro on both a G5 and a dual AMD64 3500+ ,install Doom3 on it and see what the performance specs are.
Someone whose focus isn’t playing games wouldn’t mind missing some points on that one.On the other hand someone who does gets the relevant info.
The site (the link ) is a self fullfilling economic prophecy.Heavily sponsored by Apple and the local PC shop.How could you get objective results this way?
“Can I just point out that this analogy doesn’t make any sense..
Two fast horses racing aren’t going to be any faster than one horse racing.. there’s just more horses, they will still go the same speed (all things being equal- of course).”
Sure, it was worded badly and not thought out well. The analogy breaks if all the horses start running at once, but if the side with the two horses run in a relay fashion then the side with one horse won’t be able to run as far and will lose.
But if you could get a monkey, a babboon, and a turkey…just kidding. You know what I meant, let it go.
Two fast horses racing aren’t going to be any faster than one horse racing.. there’s just more horses, they will still go the same speed (all things being equal- of course).”
Two horses are pfysically bound to some maximum speed just like a single horse is.However with two horses you don’t have to worry about aerodynamics and weight to much ,more torque.:-)
>The only multi-processor AMD, INTEL rigs I’ve been in >contact with are clusters and they weren’t playing Half->Life2 on those.
Refer to the target market for NVIDIA’s nForce Pro SLI.
As all PC benchmarks that claims that PC are faster than MAC without giving anything on how they do the benchmark…
I am a PC user who will not have the money for a dual G5 anytime soon. However, I notice that this benchmark, if nothing else, dissipates the notion that x86 PCs are much faster than macs.
Seems to me that apple+IBM is starting to put a bit of fire under the x86 world’s ass. Maybe even an increase of marketshare is possible, for apple.
This a no value since it’s missing Dual Opteron and Athlon FX-55. Simple as that.
These performance tests would always be biased . Why not just compare Intels to AMD’s and that too is not fair. ANyways I personally prefer Opterons to Intels anyday. I own a PowerMac G5 and an Opteron machine. I don’t compare them both are good machines for respective purposes. Computers are tools not a religion. Just my 2 cents.
Good work seabasstin.
It’s nice to see some maturity on the site once in a while.
Actually, you can only get it on the Mac.
Also, It’s NOT an encredible waste of time to become and Expert User!
if ( architecture != “x86”)
{
wait(300);
}
The benchmarks are not current. The machines except the G5s are old. The processors in the PC world are a lot faster these days.
Benchmarks are funny becuase depending on what you show you can show one processor platform to be a lot faster than another.
After reading all of the “I can build it cheaper and it’s faster” and “they did it all wrong” comments I have come to one conclstion: If my favorite setup comes out on the top it was a great benchmark and if it didn’t the test was flawed.
Whooooopeeeee!
Let us now all pray to our god/idol
“What about installing some Linux distro on both a G5 and a dual AMD64 3500+ ,install Doom3 on it and see what the performance specs are.”
It would be the same…. Doom3 is not even available on the Mac yet. The compiler technology still would show a slow Mac.
Doom3 is heavily optimized for x86 architecture, DirectX and a lot of things. Not to mention the graphic card issue which shows the more interest there. Even ATI drivers and NVidia drivers can make a huge difference.
Even Scientific code, written by experts, can show compiler advantages one platform over the other. Again I have to say, processor speeds are not that relevant, if it were, no one would buy Sun Sparcstations.
So again we are getting into something that is pointless. In my experience, people who buy Macs most of the time do not EVEN consider games. They are more interesting in other things, usually some kind of artistic work. Well, I have to say also, that most of the PC users only care about Word, PowerPoint and Excell. Gamers are even a minority in the PC universe.
I have to say also, most of the custom apps today on the PC world are built using Visual Basic. Do not get me wrong, it is a good tool to deliver fast solutions… But the performance hit is a enormous. I wouldn’t even dare to say Java speed apps.
Now, what I really would like to see is a comparison between PCs and game consoles…. I mean, trying to justify a 3 thousand dollars PC against an PlayStation 2, XBox, or whatever. That would be interesting in a gaming topic.
I repeat again, trying to find out what is faster is a terrible waste of time, and you are buying computer vendors adds. Apple says it is the fastest, Dells says is not, AMD is better than Intel, Intel is making a Pentium 15…. blah blah blah.
I would prefer to pour the money on software development than on Computer marketting. I cannot believe we have faster computers (Macs and PCs) and we use basically the same applications over the last 15 years.
On the PC market there is more variaty, but there is also more redudancy. Sometimes I am speechles when a company like Apple with 3 percent of the market share (maybe more or less, I don’t know) can make such a great apps, and software. The 90 % PC world (or more) should be ashamed about how the resources are wasted.
I still cannot believe Linus Torvalds and a bunch of part-time workers and computer geeks can deliver more promise than Microsoft.
Maybe if Microsoft don’t spend so much in advertising…. Maybe….
“Can I just point out that this analogy doesn’t make any sense..
Two fast horses racing aren’t going to be any faster than one horse racing.. there’s just more horses, they will still go the same speed (all things being equal- of course).”
What if the horses were pulling chariots of equal weight and construction? The chariot with two horses would win.
I think the analogy needs tweaking but AS IS doesn’t work really.
When is someone going to put Linux on all the systems involved and get a real measurement of how the HARDWARE compares. These screwed up PC/MS vs. OSX/Mac tests have too many variables to give an accurate picture of how the machines truly work. To be fair, as much of the same hardware should be used, video cards, hard drives, and memory. This test is so inaccurate on so many levels that is it useless for any comparison.
When is someone going to put Linux on all the systems involved and get a real measurement of how the HARDWARE compares. These screwed up PC/MS vs. OSX/Mac tests have too many variables to give an accurate picture of how the machines truly work. To be fair, as much of the same hardware should be used, video cards, hard drives, and memory. This test is so inaccurate on so many levels that is it useless for any comparison.
You are assuming that even linux is an equal playing field on all these hardware platforms?
Linux on x86 is far more mature and optimized than on hardware like the G5. You are also asuming that the optimized graphics drivers from Nvidia and ATI are available for the PPC and perform equally to thier x86 versions.
I also find it very strange that no matter how well a bencmark is performed the party whose favorite platform loses, almost always finds fault with the methodology used.
Suffice it to say that benchmarks will never be accepted AS IS, escpecially by the losing party.
See Topic – and move on.
IBM just need to turn up the cooker and get more powerful 970 chips to Apple. I wish it was that easy, but 2.5 aren’t the fastest anymore, but it would suit me fine!
Im sure we will see new iPods before any new Powermacs or New Power books with either a G5 or a MP G4 under the hood.
Doom3 is heavily optimized for x86 architecture, DirectX and a lot of things
Do you have any idea what you’re talking about? Doom3 uses OpenGL, not DirectX. Hence it’s obviously not going to be optimised for DirectX now is it?
And how do you know it’s optimised for x86 if it’s not even out for PPC yet? It’s possible (unlikely, but possible) that the PPC version will outperform the x86 one.
Benchmarking Doom3 isn’t actually such an awful idea – it shows one facet of performance in a reasonably cross-platform manner. It’s at least no worse than benchmarking Photoshop.
Again I have to say, processor speeds are not that relevant, if it were, no one would buy Sun Sparcstations.
But hardly anybody does buy them… 😛
I agree they’re not really important any more; the fastest CPUs make little difference unless you’re playing games, doing a lot of compiling or something similar.
“In the latest Pentium 4s, cache is king.
Intel introduced a line of Pentium 4 desktop chips Sunday that contain 2MB of secondary cache, twice as much as current Pentium 4s, as well as technology from its notebook line that’s designed to cut power consumption. Larger caches, a pool of memory located on the processor, generally improve performance.
The four chips included in the new 600 series of Pentium 4s range in speeds from 3GHz to 3.6GHz, and all of them feature an 800MHz bus. The top-of-the-line 660 Pentium 4 sells for $605 in quantities of 1,000, while the 630 sells for $224.
The Santa Clara, Calif.-based chipmaker also unveiled another member of the Pentium 4 Extreme Edition processor family, which runs at 3.73GHz and contains a 1.066GHz bus. These chips are used inside computers for gamers and serious hobbyists.
The new chips in some ways can be seen as some of the last of their kind. Starting in the second quarter, Intel will begin to sell dual-core processors. These chips won’t run as fast as single-core chips (or at least won’t continue to maintain the pace of the current speed increases) but they will contain two separate “brains,” thereby improving overall performance.
Rival Advanced Micro Devices has said it will come out with dual-core chips in the summer and deliver a new family of energy-efficient notebook chips called Turion.”
“Do you have any idea what you’re talking about? Doom3 uses OpenGL, not DirectX. Hence it’s obviously not going to be optimised for DirectX now is it?
And how do you know it’s optimised for x86 if it’s not even out for PPC yet? It’s possible (unlikely, but possible) that the PPC version will outperform the x86 one.”
Sorry, I did not know Doom 3 uses OpenGL… That’s very good in my opinion. I just assumed DirectX was used, because most of today’s game use item.
I really doubt the PPC version will outperform, but who knows… Maybe if they build it using the IBM G5 special compiler? Maybe, but still we got the videocard drivers issue… Which I doubt it to change any time soon.
Nvidia drivers are extremely optimized for games in PCs. On Mac, ATI drivers are usually better… I have seen the opposite on PCs.
“It’s at least no worse than benchmarking Photoshop.”
Until Photoshop gets no use of hardware accelaration libraries, such as those provided by “Core Image” in Tiger, Photoshop uses only processor power for filters and renderings… Doom3 is a mixed performance between processors and graphics cards.
“hardly anybody buys them”
Sun machines are still sold, maybe not as much as Sun wish, but there are many people who need them. What I tried to say is there are some computers that are not for mainstream business, but are very useful. Think about it, POWER5, Itanium and SPARCS still have customers. They are not best sellers, but they are useful. If you see them, all of those processors are slow on megahertz, but they are very powerful chips, especially POWER5 and Itanium architectures.
Note, AMD is currently introducing stepping E4 (initially with Opterons).
>Opteron 246 2Ghz –> 70W, for 19GFLOPS
Refer to Opteron HE @2Ghz 55Watt class chips.
For 2Ghz PowerPC 970FX’s max power @2Ghz refer to
http://www-03.ibm.com/technology/power/newsletter/august2004/articl…
>*the performance difference between AGP8X vs PCI express(SNIP)
It would be a non-issue with today’s GPUs and applications, but effective use of Clearspeed’s 96 processor array would require such an I/O setup. Full-duplex bus would be important in the future GpGPU applications.
you are comparing apples with oranges…err well…apples with pc´s in that case
>For each Opteron you can have almost 3 PPC970s, (SNIP)
With Opetron HE 246, you get almost one Opteron HE*(130nm) for one PowerPC 970FX (90nm).
*Currently being replaced with E stepping.
Refer to Eugenia Loli-Queru’s postings i.e.
“Here are some of the most powerful PCs vs the G5. The G5 does very well on cpu crunching, but not on graphics tests.”
2.13GHz Dothan: 10 cycles => 4.69ns
3.6Ghz Prescott 2M: 27 cycles => 7.50ns
I just bought a Motorola 68060 off of eBay for nostalgia’s sake…..at the time it was the world’s best, but Apple killed it as it smoked the forthcoming PowerPC chips.
Wish someone would benchmark the ‘classic’ chips……
Remember the days when you didn’t need a 10 pound metal heatsink or a liquid freon cooling solution?
And as someone pointed out, we’ve all been running basically the same software for the past 15 years anyway.
Bring back the 68060!!!
To be fair, you try getting hold of top spec machines for every platform to perform a benchmark
get a high end sparc,
Opteron,
Xeon,
P4,
AMD64 (in all its forms,
G5
then make an os that is equally optimised for each one…
the benchmarks were for the systems they had available… you want other systems tested GO BUY THEM!
//I wish more games would do so – again it all comes down to programmer laziness. The rush to meet deadlines wins out over quality code every time.//
This *hardly* comes down to programmer “laziness.” The programmers _rarely_ have the time they need to implement the best code. It’s just the way many software development houses work.
//besides that the OS really does not need to be 64 bit… it will just slow it down.//
That PROVES you know less than zero about operating systems.
Thanks for coming. Bye now.
dear Tetsuo The body Hammer:
Pentium 4 @ 2.8 Ghz: 68.4 Watts
Pentium 4 @ 3.2 Ghz with HT: 103 Watts
PPC 970 @ 1.8 Ghz: 42-55 Watts
PPC 970FX @ 1.4 Ghz: 12.3 Watts
PPC 970FX @ 2 Ghz: 24.5 Watts <– This is the processor we are looking at
PPC 970FX @ 2.5 Ghz: ~ 40 watts?– This is not clear
G4 (MPC7447A) @ 1.42 Ghz: 21 Watts
G4 (MPC7447A) @ 1.167 Ghz: 16 Watts
G4 @ 500 Mhz: 5 Watts
links for more referrences:
http://www-3.ibm.com/chips/techlib/techlib.nsf/techdocs/7874C7DA860…
http://www.theregister.co.uk/2004/01/22/ibm_claims_massive_power_cu…
http://www.aceshardware.com/read_news.jsp?id=80000467
http://www.tomshardware.com/hardnews/20050117_225242.html
http://www.aceshardware.com/forums/read_post.jsp?id=115063089&forum…
http://www6.tomshardware.com/cpu/20041114/index.html
Why am I calling you Tetsuo?
Because of one of the craziest Japanese cyberpunk movies ever, “Tetsuo II: Body Hammer”
check it: http://project.cyberpunk.ru/idb/tetsuo2.html
Crazy stop motion animations on live footage, really insane body transformations and gore. really great film.
Smoking a lot today?
What the hell are you talking about?
it’s not “it would be a non issue”, it is a non issue.
GPU’s today do not even flood the AGP8x BUS.
PCI express’s value is in its being ready for the next generations of GPU’s and for graphic cards with Multiple GPU’s.