Old p3 toasts g4 performance in SPEC

[grant2]

This may interest people who wanna stick it to the Macintosh zealots =)

The Register Article

c't article

 

[AGodspeed]

MartyTheManiak beat you by one minute, meaning you're a repost. 🙂

And yeah, the G4's ain't worth shiat.

 

[grant2]

Damn I guess i have to be faster next time =)

 

[wepopfresh]

some people live by benchmarks, others don't.

So whats the point?

 

[AGodspeed]

<< some people live by benchmarks, others don't.

So whats the point? >>

The most practical reasons people buy a processor:

1. Performance (i.e. benchmarks)
2. Price
3. Reliability

Compared to what's on the market, the G4 is only practical for very few people. It's price/performance ratio is lackluster at best compared to what Intel and AMD offer, who account for 95% of the PC market in of itself.

Seriously though, I hope the G5 is a huge success, because that'll give me a reason to go play with a Mac. 🙂

 

[Barnaby W. Füi]

of course Mac's price/performance ratio is not as good as PC's, thats not news. But they are good for lots of things, namely OSX =)

 

[Rand]

SPEC is a highly regarded benchmark yes. BUT it's as much if not more-so a compiler benchmark then it is a processor benchmark.
And in this case that point is especially valid... look at the compiler used for the G4 in the Spec2000 benchmarks. The compiler they utilized is often cited among the PPC community as producing extremely poorly performing binaries.

It wouldnt take much effort for me to find Spec2000 benchmarks that position the P4 as scoring even lower then that depending upon the compiler used.
The most powerful rpocessor in the world can look sickening without a decent compiler in Spec, this has often been AMD's biggest problem with their relatively lackluster Spec scores relative to their excellent real world performance.

I'm not much of a Mac fan, but passing this one example off as being a valid indicator of potential G4 performance in the Spec2000 benchmarks is extremely poor journalism.
Someone at the Register evidently forgot just how much the compiler used can impact Spec results.
Am I the only one that remembers Spec2000 results suddenly being boosted by almost 500% in some cases due to variances in the compiler used?!

 

[grant2]

Rand, from reading the 2nd link (c't article) i get the impression they were "conservative" with the compilers.

People will never run software that has been compiled only on the latest, most optimized compilers, nor the oldest, least optimized. It seems that SPEC has chosen compilers which will represent the medium of the most common compilers that most users will encounter.

At worst, this spec benchmark is heavily biased to the same degree that the infamous photoshop benchmarks have been, so it ought to cancel out =)

 

[Rand]

<<
People will never run software that has been compiled only on the latest, most optimized compilers, nor the oldest, least optimized. It seems that SPEC has chosen compilers which will represent the medium of the most common compilers that most users will encounter.

At worst, this spec benchmark is heavily biased to the same degree that the infamous photoshop benchmarks have been, so it ought to cancel out =) >>



Spec doesnt choose what compiler is use, you can use any compiler as long as it retains the original code characteristics. Manufacturers typically update their Spec results constantly as Compilers improve. Remember the hoopla paid to Sun's Forte because it boosted performance in a specific sub bench so drastically as to equate to years worth of microprocessor architectural innovation.
My point is that Spec is a platform benchmark that is intended as much for a test of the compilers capabilities as of the actual memory hierarchy and processor, and to a lessser degree the motherboard.

You pair the P4 with a compiler as unsuited for it as the compiler used in this test is for the G4 and you'll see equally bad benchmarks.
I can easily point out sites wherein you can find a PIII 1GHz as destroying an AthlonXP 2000+ and the like simply because of a different compiler.
It shoud most definitely NOT be taken as an indication of real world desktop performance. Spec bears no relation to desktop application performance.