Benchmark: Dual G5 vs Athlon 64 vs Itanium 2 vs Pentium 4 vs Power 4 vs Power 3

[GL]

The University of Hamburg has done some benchmarks of its atmospheric modeling software on several platforms. Information about the program is here..

Benchmark background information

The Results

 

[Monoman]

well, the A64 comes in 3rd soo... IT'S DA BESTEST!!!!! j/k


good read

 

[Eug]

The G5 2.0 rulez all, but it isn't a fair test since it's using Altivec.

 

[IvanAndreevich]

>>but it isn't a fair test since it's using Altivec
Uh.. nice logic there.

It isn't a fair test since it's using AMD64
It isn't a fair test since it's using SSE2/SSE3
It isn't a fair test because the mac wins?

I don't like macs and I don't like G5 as opposed to K8. However, this logic is baaad...

 

[addragyn]

^

"run what ya brung"

 

[mechBgon]

What kind of a test is that? They left out the VIA/Cyrixes!



(or wait... maybe the VIA/Cyrix's bar on the graph is just too short to see. n/m! )

 

[GL]

Originally posted by: Eug
The G5 2.0 rulez all, but it isn't a fair test since it's using Altivec.

The compiler he used for the x86 chips generates MMX/SSE/SSE2 code.

 

[addragyn]

"(or wait... maybe the VIA/Cyrix's bar on the graph is just too short to see. n/m! )"

That would make it faster than everything else.

 

[mechBgon]

Originally posted by: addragyn
"(or wait... maybe the VIA/Cyrix's bar on the graph is just too short to see. n/m! )"

That would make it faster than everything else.

Oops! So much for that

 

[Eug]

The G5 2.0 rulez all, but it isn't a fair test since it's using Altivec.

The compiler he used for the x86 chips generates MMX/SSE/SSE2 code.

Good point, but he's hand coding Altivec (since xlf doesn't auto-vectorize).

BTW, I know nothing about coding, but I'm just pointing a potential argument.

It isn't a fair test because the mac wins?

I don't think I'm anti-Mac, considering I'm typing on my PowerBook right now.

 

[Ardan]

The Athlon 64 did nice there....I like it . The G5 is a powerful CPU, and it is in a dual configuration there. Why would it be surprising that it wins? . I would say it is *gasp* fair enough, since you're NEVER going to get those CPUs using precisely the same things, might as well deal with it. When comparing them, i'd rather compare them like comparing cars. Sure, the parts inside two different cars may be from different makers (like Chevrolet Carburetors from Rochester, and Ford's typically from Holley) and sometimes may be quite different inside, but that doesn't mean the comparison is unfair. They could both be meant for families, and be sedans...it doesn't matter if the insides are precisely (or nearly) the same, what matters is the performance. Likewise with these, it doesn't matter...they make it quite fair, as people have pointed out. If the Dual G5 wins, well then hey Apple sure is making a great OS and IBM are producing nice chips (and AltiVec would sure seem to work well)....shouldn't be skeptical just because it is a Mac (regardless of whether we use Apple computers or not) and is on a different architecture. Thats my 2 cents, and i'm not subscribed to the thread because I came to add in to the discussion and nothing more. I hope I saved you some typing .

 

[IvanAndreevich]

We need some Opteron & Xeon. Especially in Quad flavours.

Eug
Alright! A bit off with my judgement there.

 

[Sunner]

Strange that the POWER4 doesn't do better since the PPC970 is derived from it.

# 1 IBM pSeries 690 computer with 32 CPUs and 256 GB RAM
# 1 IBM pSeries 690 computer with 32 CPUs and 128 GB RAM
# 10 IBM pSeries 690 computers with 32 CPUs and 64 GB RAM

I'd love to play around a bit with those

 

[n0cmonkey]

Macs and AMD64 at the top... Was there ever any doubt?

 

[txxxx]

Was the athlon64's code compiler with 64-bit switches active? If not, still impressive, although its not far off a P4 2.6ghz CPU.

 

[Eug]

Originally posted by: Sunner
Strange that the POWER4 doesn't do better since the PPC970 is derived from it.

The POWER4 has a much lower clockspeed, and I'm thinking the test doesn't really require the huge amounts of cache the POWER4 comes with. And I wonder how much of the speedup for the G5 is coming from Altivec.

 

[sharkeeper]

A G5 would make a nice XOR processing unit for an upcoming storage project.

Cheers!

 

[GL]

Judging from the comments, I think some of you are missing the other benchmark results there. The G5 only slaughters the competition in the first benchmark. The Athlon 64 and the G5 trade top spots in the remainder of the results. When you open the results page, there are 5 buttons "Lorentz", "Mdwarf", "Full NLTE", "NLTE Gauss" and "NLTE Voigt". Click on them to see the other results.

 

[JBT]

Originally posted by: GL
Judging from the comments, I think some of you are missing the other benchmark results there. The G5 only slaughters the competition in the first benchmark. The Athlon 64 and the G5 trade top spots in the remainder of the results. When you open the results page, there are 5 buttons "Lorentz", "Mdwarf", "Full NLTE", "NLTE Gauss" and "NLTE Voigt". Click on them to see the other results.

Yeah I was wondering the same thing all the other chips are much better in the other tests, either way where is the dual setups I know Operton's, Xeon's Itanium's and the Power chips can all have dual? rather than just sinlge vs setups...

 

[DAPUNISHER]

It's a shame they used a P4 2.6, a P4 3.4 would have made a nice showing me thinks.

 

[Eug]

Originally posted by: DAPUNISHER
It's a shame they used a P4 2.6, a P4 3.4 would have made a nice showing me thinks.

Yeah, but so would an Athlon64 at 2.4 GHz. BTW, where is that 2.5 GHz G5? It's long overdue to come out. (I'm betting by the end of the month, but as usual IBM and Apple are mum on the subject.)

 

[DAPUNISHER]

Which A64 did they use Eug? I don't keep up with Apple, but it seems to me@2.5ghz they will make a few waves no?

 

[HokieESM]

"Warning: Intel's C++ compiler cannot compile the QD library with anything above -O0 correctly, that puts all Intel based machines at a severe disadvantage compare to PPCs or Power systems. g++ does compile the QD code correctly, however it currently produces not quite so fast executables and it seems to be impossible to link g++ code with Intel ifort (version 8) code, that is a massive performance problem for all IA-32's, IA-64s and AMD64s. "

From the Benchmark background. I don't think this was exactly a fair test, once again. He also notes its a test of the compiler... and, to be blunt, IBM rules all in this area (xlf95 is amazing). The performance of the G5 is not surprising.... although, I think with a different piece of code (without having to do the link from fortran to c that the intel cannot handle, apparently), the results might be "closer". I'm shocked with the Itanium results.

Example: the code I'm writing for my dissertation runs 400% faster with ifc8 than with g77. Same computer, same operating system, just different compiler.

 

[Sunner]

Originally posted by: HokieESM
"Warning: Intel's C++ compiler cannot compile the QD library with anything above -O0 correctly, that puts all Intel based machines at a severe disadvantage compare to PPCs or Power systems. g++ does compile the QD code correctly, however it currently produces not quite so fast executables and it seems to be impossible to link g++ code with Intel ifort (version 8) code, that is a massive performance problem for all IA-32's, IA-64s and AMD64s. "

From the Benchmark background. I don't think this was exactly a fair test, once again. He also notes its a test of the compiler... and, to be blunt, IBM rules all in this area (xlf95 is amazing). The performance of the G5 is not surprising.... although, I think with a different piece of code (without having to do the link from fortran to c that the intel cannot handle, apparently), the results might be "closer". I'm shocked with the Itanium results.

Example: the code I'm writing for my dissertation runs 400% faster with ifc8 than with g77. Same computer, same operating system, just different compiler.

Is Intel's Fortran compiler better than ICC?
I've heard many people talking about the shortcomings of ICC in other areas than performance.

 

[Pariah]

Wow, a dual 2 GHz G5 beats up on a 2.6GHz P4 and 2 GHz A64 in number crunching, who'd have thunk it? A dual Opteron 248 or dual 3.2GHz Xeon would have put up a much better showing. Heck, a single FX-51 or P4EE 3.2GHz would have looked a lot better.