The Low Down on SPEC
According to <a href="http://appleturns.com/"; target="_blank">As the Apple Turns</a>, German technology magazine c't just completed the SPECCPU 2000 test suite, finding the Dual 1GHz equivalent to an Intel Pentium III 1GHz in integer performance and lagging behind the Pentium III in FPU performance. (I'm assuming that the SPEC tests were single processor only.) While the results of this benchmark may put the G4 in a very negative light, there are a number of considerations one has to weigh before declaring the PowerPC a slouch:
1) It has been alleged that SPEC favors Intel processors;
2) A 1GHz Pentium 3 is said to be equivalent to a 1.4GHz Pentium 4;
3) A 500MHz G4 is said to be equivalent to a 1GHz Pentium 3 when alternate benchmarks are used;
4) Bare Feat's recent comparison between the Dual GHz and its Athlon and Pentium 4 rivals showed the G4 to be much closer in performance than SPEC would indicate.
With the aforementioned in mind, what reasonable conclusions would others draw about the validity of SPEC marks? I personally believe that the results in question are negated by the real world performance measurements used in other benchmarks. I believe SPEC is most likely fully optimized for Intel chips, whereas little if any optimization goes into PowerPC code. Comments?
1) It has been alleged that SPEC favors Intel processors;
2) A 1GHz Pentium 3 is said to be equivalent to a 1.4GHz Pentium 4;
3) A 500MHz G4 is said to be equivalent to a 1GHz Pentium 3 when alternate benchmarks are used;
4) Bare Feat's recent comparison between the Dual GHz and its Athlon and Pentium 4 rivals showed the G4 to be much closer in performance than SPEC would indicate.
With the aforementioned in mind, what reasonable conclusions would others draw about the validity of SPEC marks? I personally believe that the results in question are negated by the real world performance measurements used in other benchmarks. I believe SPEC is most likely fully optimized for Intel chips, whereas little if any optimization goes into PowerPC code. Comments?
Comments
1) It has been alleged that SPEC favors Intel processors;
<hr></blockquote>
it doesn't.
[quote]
2) A 1GHz Pentium 3 is said to be equivalent to a 1.4GHz Pentium 4;
<hr></blockquote>
dunno.
[quote]
3) A 500MHz G4 is said to be equivalent to a 1GHz Pentium 3 when alternate benchmarks are used;
<hr></blockquote>
uh. you mean altivec? ...
[quote]
4) Bare Feat's recent comparison between the Dual GHz and its Athlon and Pentium 4 rivals showed the G4 to be much closer in performance than SPEC would indicate.
<hr></blockquote>
Bare Feats is total crap. So is OSOpinion.
Many testing (especially to the latest Athlon test) are done 'out of site'....which means the third party can give bear feats some false info on the spec of the testing machine.
The latest result from the Athlon they mentioned must be DUAL not single.
Anyway you can get the dual athlon for almost half the cost of the Dual G4.....the 'Apple hardware sucks' fact remains.
As has been pointed out elsewhere, I think the results highlight the current state of Apple's compiler and math libraries: <a href="http://www.opensource.apple.com/bugs/X/Libraries/2692951.html"; target="_blank">(darwin bug)</a>. gcc is currently nowhere near optimal on the PowerPC architecture, but Apple is working on it. Had the benchmark been run under MacOS 9 with the Metrowerks compiler there would have been a significant speedup (about 50% - 60% in my experience).
This test has nothing to do with the G4, it is a comparison of compiler state, nothing more.
[ 03-09-2002: Message edited by: Fluffy ]</p>
<strong> Outside that, the sub-top-line FPU performance really won't make much difference.</strong><hr></blockquote>
except 3D rendering.....
Word is getting out. This is bad news for Apple, they better do something about it. Perhaps they could optimize SPEC benchmark apps by compiling them with Codewarrior and optimizing for Altivec? Otheriwise it's not a fair test, since SPEC is well known to be heavily optimized for Windows but not for OS X or OS 9.
[ 03-10-2002: Message edited by: Leonis ]</p>
I have been touching with Dual 500, Dual 800, Dual 1Ghz G4s and Dual Athlon 1900XP, 2 Ghz P4
On Cinebench and rendering in Lightwave and Cinema the Dual Athlon is the best that your money can buy.
In Photoshop the results between a Dual 1900XP and a Dual Ghz really depends on what filter you are using. Some are faster on a G4 and some are faster on the Athlon
After Effects 5.5. I tested the app with my friend's Dual Ghz G4 a Dual 1900XP Athlon. Using his project with has lots of layers and filters and stuff. Very interesting some happens here. AE rarely pushes both processor above 51% usage (OS X and Win 2K) with exception of some filters. So in other words it's like a single Ghz G4 vs a 1.6Ghz Athlon. In the test some part the G4 is over 50% faster than the Athlon and some part the Athlon is faster than the G4 at the same rate. Which means. Again depending on what kind of stuff is in your project......When working on a multi-million layers projects (especially to those using 3D layers) G4 is slightly better, but for projects with thousands of filters applied Athlon is slightly better.....
QT compression using Sorenson video 3.1 pro & Media Cleaner 5.1 . Dual Ghz G4 beats the Dual 1900XP Athlon hands down (30%). I guess it's the QuickTime trick.
I haven't used Premiere since the first day I touched FCP....so can't comment on this....
Don't know about DVD authoring
[ 03-10-2002: Message edited by: Leonis ]</p>