SPEC on G5

[sturm375]

Let me start by saying that I am very impressed with the new G5 PowerMacs. I do want to question the SPEC numbers posted here

Now compare to the SPEC numbers posted here

Lets not concern ouselves with the AMD v. G5 just yet. Look at the Xeon numbers on the two pages. Apple's posts a: 11.1 for the Xeon 3.06 Ghz, while AMD's posts a: 19.6 for a 2.8 Ghz. Maybe I am a noob, and not reading the fine print, but I think something fishy is going on here. I have not seen AMD "fudge" numbers on benchmarks, but I have seen Apple do some pretty strange things with benchmarks.

Again, this is not to say that the G5 is a door stop, in fact I'd say it's a pretty kick-@$$ machine. I just hope somebody can enlighten me on why the difference here.

Thanks

 

[daveg5]

Re: SPEC on G5

Originally posted by sturm375
Let me start by saying that I am very impressed with the new G5 PowerMacs. I do want to question the SPEC numbers posted here

Now compare to the SPEC numbers posted here

Lets not concern ouselves with the AMD v. G5 just yet. Look at the Xeon numbers on the two pages. Apple's posts a: 11.1 for the Xeon 3.06 Ghz, while AMD's posts a: 19.6 for a 2.8 Ghz. Maybe I am a noob, and not reading the fine print, but I think something fishy is going on here. I have not seen AMD "fudge" numbers on benchmarks, but I have seen Apple do some pretty strange things with benchmarks.

Again, this is not to say that the G5 is a door stop, in fact I'd say it's a pretty kick-@$$ machine. I just hope somebody can enlighten me on why the difference here.

Thanks

i bet apple is using its own compilers just like amd and intel to make thier numbers look better

 

[Cubeboy]

Most CPU makers do use their own compilers or at least the best performing compiler (Take AMD for example). However, since official SPEC submissions don't allow for radical compiler optizations for the peak score and only very conservative or no compiler optimizations for the base score, it's very difficult to really boost your SPEC score by a significant amount over other comparable compilers.

As is apparent by the SPEC score the PPC970 seems to do significantly better on a IBM compiler (Visual Age?) than the GCC compiler.

The SPEC scores seem to match the predictions done by Ars Technica, at 2 GHz, the PPC970 is be slower than the Pentium 4b at 3.06 GHz with a 533 Mhz fsb in integer operations but faster at floating point operations. Not sure how it will do against a Pentium 4c 3.00/3.20 GHz with a 800 Mhz fsb.

But your right about the thoroughput of the Xeon, I've seen dual 2.4-3.06 Xeon SPEC rates from many compilers but never such a low score from a Dual 3.06 GHz.

 

[yzedf]

Apple's numbers are based on using gcc (compiler) for both G5 and Xeon rigs. The other Xeon scores are based on Intel's compiler.

In the real world, gcc for the PC only really matters for the linux crowd. Windows programs seem to mainly use Intel's compiler.

The difference in scores is a nice way of showing what junk the gcc is in comaprison to other compilers (in this case Intel).

 

[sturm375]

AMD used an Intel compiler for their SPEC here

SPECfp_rate2000 = 22.7
SPECfp_rate_base2000 = 21.0

Again an Intel compiler here, for the same Xeon competing against the G5.

Dell
Precision WorkStation 650 (3.06 GHz Xeon)
here

SPECfp_rate2000 = 15.8
SPECfp_rate_base2000 = 15.7


I guess this benchmarking is just a bunch of smoke and mirrors. G5 is still a very nice looking 'puter.