B. Mann Consulting

Yeah, I don't use any marketing slogans about Apple, considering my history with them. Making claims like "this is the fastest desktop PC ever" is pretty stupid considering the performance curve of the PC industry.

This is the article I will discuss, as it seems to have generated the most attention:

http://www.haxial.com/spls-soapbox/apple-powermac-G5/

This issue of manipulated benchmarks is turning out to me to be one guy's vengeful nitpicking - is he claiming that Apple tweaks benchmarks, and Intel or AMD doesn't tweak for their own processors? Ridiculous. What this guy did was take the Veritest results, and blow up every modification that was made as if this was a unique thing - meanwhile he is quoting vendor-supplied SPEC results from Intel as if they were infallable.

Since this can be done on any set of published results a vendor uses, I find the fact that he is doing it to try and spite Apple and Steve Jobs rather questionable - look at the minor things he claims to be significant, like how vendors price things at $1999 and $2999 rather than $2000 and $3000, he claims this is deceptive. Yeah, more like he hates it, and is trying to use this article as a platform for his own views. Further, he posts all the hate mail he is receiving from Apple users and responds to it all - as if these people were mature adults, and not idiotic teenagers.

What Apple has clearly done here is try to level the playing field for comparison - they used compiler software available for both platforms. Their decision to use GCC is actually very gutsy of them, as GCC for Intel is allegedly far more optimized than the PPC GCC - giving the Intel machines a definite advantage. The Fortran compiler they used is also cross-platform, but it's unlikely that the compiler is better at optimizing PPC code than x86 code.

Lastly, I am rather amazed at the results of the SPEC - particularly at how low the results are for the Apple machines. Then again, looking at the decisions Apple made, they pretty much shot themselves in the foot:

- The new Macs don't have good clock multipliers at all - they could have had 1:2 on all 3 systems if they kept an 800 MHz FSB and put DDR 400 on all three as well. But instead, on the one unit with an 800 MHz FSB, they went with DDR 333. Having studied a lot of memory tests on the Asus A7N8X, with its 333 MHz FSB and ability to work with DDR 333 and DDR 400, I can say that this is a huge difference - the Asus motherboard with DDR 333 significantly outperforms an identical setup with DDR 400 on the majority of tests, because the latencies of a 1:1 multiplier are substantially lower. This is a big loss.

- Another major decision on Apple's part was to use single-channel DDR on its motherboards, except for the high-end unit, which has one DDR channel per processor. Dual-channel DDR is the reason why the Asus A7N8X boards have been able to keep up with competition from Via and others. This would have doubled memory bandwidth and lowered latencies even further on the Apple motherboards.

So on these motherboards, Apple is really shooting itself in the foot. On the tests by the guy who criticizes Apple, the results show that the IBM Power4, upon which the 970 is based, is faster than anything else by a huge margin. How can those results be so large, and yet the 970 (which is designed for much higher clock rates) be so much slower?

Apple's SPEC results are on a prototype dual 2GHz motherboard using GCC, that Fortran compiler, and OSX. The results Dell publishes, which this guy does not dispute in the slightest, use Intel's C compiler, a different Fortran compiler, and Red Hat Linux.

So Apple has attempted to level the playing field with software, trying to make just a hardware-based comparison. Apple's to apples, so to speak. Whereas Dell and AMD have used the most optimized compilers they can find in their own SPEC results, and there is no knowing how many special libraries they used to boost the SPEC results. But this dissident assumes them to be canon, and assumes Apple's results are flawed.

Oh well, pobody's nerfect. Like smithdm3 says, it's great to see a new competitor in the arena, and regardless of the SPEC results, these new machines are incredibly faster than the G4 machines they replace. Give Apple time to make optimizations to GCC, OSX for 64-bit, and a second-generation motherboard with the 3 GHz chip Jobs said IBM would have out within 12 months, plus the potential move from 130 nm to 90 nm that an MPR analyst said IBM would be making, and this next year is going to quite clearly bring Apple back in the performance running, and promise a number of significant optimizations and performance boosts, to ensure Apple's hardware is competitive over at least the next year. And I can only hope IBM also makes a PowerPC derrivative of its POWER 5 processor