Geekbench Comparison Redux
The other day I posted an article to highlight our new processor benchmark application Geekbench. It received a lot of attention, but there was some concern about the machines used in the testing (mainly, the Athlon 64 and Pentium 4c are kind of old). Despite the fact that the article was meant to be more about the testing than the results, some people refused to let it pass.
Luckily, a few people responded to my request for results from more machines, so here is a smaller article comparing just three machines: A PowerMac G5 Quad, an Athlon 64 X2 and a Pentium D.
Test Machines
- Power Mac G5 ‘Quad’ 2.5 GHz
- 1 MB L2 cache per core
- 1.25 GHz system bus
- 1.5 GB DDR2 533 SDRAM
- Mac OS X 10.4.4
- AMD Athlon 64 X2 3800+
- 512 KB L2 cache per core
- 1000 MHz system bus
- 2 GB DDR 400 RAM
- Windows XP SP2
- Intel Pentium D 3.2 GHz
- 1 MB L2 cache per core
- 800 MHz system bus
- 2 GB DDR2 533 RAM
- Windows XP SP2
The G5 is the same machine previously tested. My thanks to stanner and Jonathan for Athlon 64 X2 and Pentium D results respectively.
Also, just a few notes before I get started:
- Geekbench is new, and still a preview release
- Geekbench is not optimized for any CPU
Floating Point Performance
Geekbench measures floating point performance by computing a Mandelbrot set in one or four threads, using the square root function in one set of tests, and without using the square root function in another set of tests. All results are in megaflops.
On the single-threaded square root tests the Pentium D bests both by over 150 megaflops, which is not that surprising as the Pentium D has the highest clock speed. On the multi-threaded square root tests though the G5 takes a commanding 800 megaflop lead over the Pentium D, which is also not a surprise because the test runs four threads and the G5 has four cores. The Pentium D bested the Athlon by just under 600 megaflops on this test also.
On the single-threaded no square root test the machines all performed within 200 megaflops of one another, with the G5 on top. On the multi-threaded version of this test the G5 takes an even more staggering 2560 megaflop lead over the Athlon 64 X2, but again this is not surprising as the test is again four threads and the G5 has four cores. The Athlon bested the Pentium by about 400 megaflops on this test.
Integer Performance
Geekbench measures integer performance by performing Blowfish encryption on one or four threads, using a data set small enough for a processor cache in one set of tests, and a data set too large for a processor cache in another set of tests. All results are in megabytes per second.
On the two tests that run in the CPU cache, the G5 again takes commanding leads: over 160 MB/sec on the single threaded test, and 666.5 MB/sec on the multithreaded test. Again, there is a lack of shock on my part because the G5 has a total of 4 MB of L2 cache, and the multi-threaded test is again 4 threads. The Athlon 64 X2 and Pentium D scored fairly closely, the Pentium pulling ahead on both.
Moving to the Blowfish memory test the Pentium D scored highest on the single threaded test, and the G5 again took home top scores on the 4 thread test. The Athlon stayed fairly close to the Pentium on both these tests again, less than 26 MB/sec separate them on both.
More Integer Performance
For another measure of integer performance, Geekbench executes code for the MOS Technology 6502 CPU in a virtual CPU. This test is single threaded, and results are in megahertz.
The Pentium 4 was able to emulate the 6502 the fastest here, clocking in at about 372 MHz. The G5 was not too far behind at 331 and the Athlon brings up the rear at just over 300.
I wasn’t sure what to expect on this test, given that the older Pentium did so poorly in the previous batch. Still, this processor is more than a year newer, so I guess Intel had lots of time to address some of Pentium 4c’s drawbacks.
Memory Performance
Geekbench uses two methods to test memory performance. The first set of four all use functions from the standard library, first to fill a block of memory, then to sequentially access chunks of memory within a larger block, then randomly access chunks of memory from a larger block and finally to copy one block of memory to another. These tests are single threaded, and results are in gigabytes per second.
These tests are dominated by the G5; it scores at least one and a half times better than both the Athlon and the Pentium on all four tests1. I’m guessing this has more to do with the 4 cores than anything else, but the G5 does have faster RAM than the Athlon 64 at least.
The second set of memory tests uses processor operations, first copying data from one block to another, then three operations that manipulate data (scale, add and triad) before writing it back to memory. All of these tests are single threaded, and results are in gigabytes per second.
This series of tests resulted in a much closer race, the Pentium D taking top spot on the copy test, and the G5 taking top spot on the rest. The Athlon came in last on all four, but that’s not too earth shattering given that it’s using older DDR 400 RAM.
Conclusions
As with the last article there are some technical aspects of the CPUs in question that I am unaware of with regard to how the results panned out. That said, here (again) are what the results tell me.
The Pentium D and Athlon 64 are both good CPUs, and their performance was fairly close. The Pentium D has twice the L2 cache of the Athlon (1MB per core vs 512KB per core), uses faster RAM (DDR2 533 vs DDR 400), and has a higher clock speed (3.2GHz vs 2.0GHz), yet the Athlon was able to keep up to (and in a couple of cases outperform) the Pentium. This, I think, highlights the fact that there is more to performance than megahertz.
I hope this gives you a better idea how higher-end processors perform. I apologize for not including more machines, but I’m hoping to publish an even larger article in a couple of weeks. For those of you who are interested, the raw number from the testing are available for viewing.
Cry for Help
I wasn’t able to get any Athlon 64 FX scores; is there anyone out there willing to help out with testing? !
Brief Addendum
There are now two mailing lists for Geekbench. geekbench-announce, wherein we will announce things like new versions, and geekbench-discuss, in which we will discus new features, feature requests, methods, etc. You can subscribe to either (or both) list here
Notice
Since the writing of this article a new version of Geekbench has been released. Please see the Geekbench page for more info.
Or, if you prefer your browser to your email client there is now a forum too. Check it out.
1 Technically, on the copy test the G5 only did 1.48x better than the Pentium D, but I rounded it up. Sue me :p
I want a G5 Quad.
I’m glad I bought a Quad.
je ne voit pas l interet d un tel test.vous auriez dut plutot tester un powermac mono processeur sinon vous pouvez faire un comparatif entre un quad et un pII 266 mhz pour etre sur que le quad est plus performant
c koi ce test, le 3200+ en x2 n’existe meme pas, la gamme comense a 4200+!!! et pis le P4, dans le genre reference…
Since a single G5 is a dual core processor it should be only fair to compare it to a single X2 or D processor.
If you want to test a Quad G5, bring the Quad Xeon and the Quad Opteron (i think it’s the only quad proc combination from AMD)
Looks like Apple made a BIG MISTAKE switching to Intel!
Hey moronns (both french and english) he can only test what he has!
Why dont you donate a quad xeon or quad opteron (do they even exist?) instead of getting your M$ issued panties in a bunch.
Robert: The Pentium D is not the same as the processors inside the new iMac and MacBook Pro. Know your facts before opening your mouth.
I really can’t trust these benchmarks. Given the setup you’re going with, we may be seeing differences in operating systems over differences in Processors. And likewise as far as compilers go, depending on what you used (you neglect to even mention that). To make things fair all the way around you’d be better off runnning a multiplatform OS, something along the lines of Linux or NetBSD, as well as compiling your code with the same compiler on each system. Even then there might be differences in software which affect the benchmarks, but at least they’d be kept to a minimum.
Geekbench isn’t just measuring hardware performance, it’s also measuring software performance (including the compiler used to build Geekbench, and the operating system used to run Geekbench), and this is intentional. After all, compilers and operating systems have an impact on performance (especially when you consider day-to-day use).
So if that’s the case, why aren’t you listing which compilers were used, and testing more than just one per platform? There’s GCC, the Portland Group’s compiler, Visual Studio, and Intel’s C compiler, just to name a few.
Also, it’s rather suspect that the code to geekbench isn’t readily available. Like the previous poster just stated, depending on what you’re doing, you could be testing the C library more than anything.
Decker,
I wrote a brief post in our (really quiet) forums about the compilers we’re using for Geekbench. Hopefully it answers some of your questions. Essentially I don’t think it’s valuable to test with a number of esoteric compilers, considering most applications for a certain platform are compiled with the same compiler.
Geekbench reports what compiler it was built with, and we’ll be sure to include that information in subsequent articles. For this article and the last article, we used Visual C++ 2003 for Windows, and GCC 4.0 for Mac OS X.
Geekbench uses the default compilers on whatever system it is being run on. So on OS X it uses GCC and on Windows is uses Visual C++.
For those of you wondering, there is more information in the previous article and the Geekbench announcement, both linked in the first paragraph of this article.
Good to see someone trying to make a comparison on some VERY different processors. I’d love to see the figures for a dual G5 at the same clock speed – and also something from AMD that is not just their entry level dual-core processor. Even a 4400+ X2, but the 4800+ would be even better.
Well, these tests don’t surprise me much seeing how the G5 has 8MB of L2 cache total, and 2×1.25GHz bidirectional FSBs. Even (although it’s directly applicable now) now I’d be more interested in seeing comparisons between various G5 models, intel’s newer desktop processors, AMD’s highers, and the core duo in specific. Of course, he can only test what he can get, and I’m in no position to donate (wow am I poor), but it’d still be interesting to see. However, on the terms of testing, I think some more things could be tested. For example, the same gfx cards in the different processor configurations.
Well that is nice that you have a benchmark but to be honest you should be outsourcing your tests to a more capable firm such as anandtech, ocworkbench or overclockers.com/ocforums.com to give you more range in hardware testing. At this point your benchmarks are a little sketchy at best and do nothing more than provide you an outlet to say that your software compiles.
the result is absolutely biased. how on earth is any scientist benchmark/compare something without using a “control” environment? None of the hardware is the same!!! It’s like comparing an orange with an apple and saying the orange has more juice than te apple, even though it’s smaller… plain unconcludable.
sorry i didn’t make it more clearly.
has anyone tried to use the same CPU, same motherboard, same type and amount of RAM, same Operating System… just everything the same. The special thing is to swap the Display Card?
Try to use a AGP type ATI X700 and try a PCI type Intel i740 (yes, old enough) then use the same benchmark program to test the result. Yes, you’ll see differences in performance. Why? You’ve got a bottleneck somewhere and in this case happened to be the display controller. The overall system will wait until the display is finished then proceed with the next big instruction. Don’t believe it? Try it.
In this Geekbench Comparison, what if the system bus is set to 66MHz? (if one can do it) It slows down everything because that’s the limit of data transfer! No matter how fast the CPU is, how much RAM you got, what greatly optimised code you use, the system is basically slowed down by that bottleneck.
On the otherhand, eliminate all the bottleneck (that means the CPU is slowest of every other components), then may be you can say how the CPU performs.
Glad I bought an AMD X2, for games is simply the best see ya sucker
with your weird variable types 
Knew the G5 would destroy the Pentium & Athlon.
After all, why are Macs reknowned for their multimedia capabilities? Better processing.
The G5 beat the Athlon64 and Pentium in this test not because it’s better, but because it was 4 cores vs. 2 cores vs. 2 cores.
if g5′b beat x2’s then why are we not using them for games????
afterall most high end pc equipment is aimed at gamers who are allways hungry for more powerfull hardware.
just a thought
sat here smugly with my x2 4400 and my ati x1900xtx
Glad to see everyone’s still getting the point.
If anyone’s interested, I have access to a FX60 machine for a couple of hours on March 31 and I’ll be happy to benchmark it with Geekbench. Just tell me where to post the results. I’m not in it to rubbish Mac’s or PC’s, as it is not a straight forward comparison to say one is better than the other, there are so many variables, it is impossible to get around them all.
If you like Mac’s, that’s great. If you prefer a PC, well that’s great too.
Peace brothers.
Geekbench features results submission, so when it asks if you want to submit just say yes and we’ll get them.
My Intel Celeron 266Mhz 56k Cache
would destroy all ur CPUs on benchmarks, ok maybe not!
Doesnt the XBOX 360 Use a 3 Core G5 CPU????? Maybe it is
faster for games hence why Microsoft are using it. Sony didnt bother with AMD or INTEL, they just went for the CELL!
What planet are you on? Why choose the crappest X2 you can find? Unless it nicely skews the resutls – and it STILL don’t look too bad from where I am sitting. So, just out of interest if you used a 4600+ I wonder what these great scientific graphs would look like – COME ON BIASSED PEOPLE – be fair.
P.S> I still have a PII 450, if you’d like to throw that in the mix…and and my digital watch…that would be a good fair comparison too.
Well you guys are comparing a computer with really 4 cpu’s on 2 cores against pcs which have just two! Sure it should be faster:)
Not to mention the fact that you were using low end parts for the pcs which cost a fraction of the quad. Try comparing against the newer x2 5000 or the newer preslers.
Apple computers are a rip off compared to the value of my pc, you should compare products in the same budget category.
omg how exacly did apple make a BIG mistake by joinin intel?
intel beat amd in most of the test and i realy think the xeon would beat alot of things even though it is a server processor its one of the most powerful processors there is
ok to the guy named X2 rules..
the reason why you are not using G5’s for games is that apple do not sell the motherboard and processors seperately, you must buy a complete system. with the best system costing around £4000 or about 6000USD.
as for the X2 well it probly matches the P4 whereas the FX60 will beat the P4 EE but intel have the xeon, alot of people say that its no good for games cos its used for servers, they are wrong. xeons are great processors just like the G5 because wot do some servers do? process games for thousands of people accross the world and if a single dual core or dual single core processors can do that then they must be good.
if intel made a xeon for home use and home gaming then it would be the best processor out
Hehe so what a few benches lol, you people must have money to burn to use quads! My x2 4400 with an ati x1800xt will blow it away in any game and cost about half as much:))))))))