G5 - is it really that fast?

[Charles Martin]

> From: Derek Roff <derek at unm.edu>
>
> Greg Joswiak, vice president of hardware product marketing at Apple,
> said:  "The scores were higher under Linux than under Windows, and in
> the rate test, the scores were higher with hyperthreading disabled
> than enabled.  He also said they would be happy to do the tests on
> Windows and with hyperthreading enabled, if people wanted it, as it
> would only make the G5 look better."
>
> I find this curious.  Joswiak says that Apple chose Linux as the
> faster PC OS, in order to have a valid G5 vs. Intel comparison.  From
> a marketing standpoint, I would have used Windows, since that is what
> most customers will use.
>
 From a marketing standpoint, you're obviously right -- and it would 
have made the G5 look even better too!

But then we'd get endless complaints from people that Apple "hobbled" 
the Intel by deliberately using a "bloated" OS. The Linux geeks around 
the world would howl loudly about the G5 being a fraud, and that it 
couldn't stand up to a "lean" OS.

As Joswiak said, in EVERY situation where a variance could be used, the 
one that FAVOURED the competition was used.

Mark my words -- these benchmarks will hold up under the scrutiny.

> As to the importance of floating point speed, I have always read that
> all graphics, image and sound processing, such as video, games,
> PhotoShop, CAD, ProTools, etc, need floating point for all the
> serious number crunching.  If true, this would make floating point
> performance more important for all the work I do, except email and
> word processing, for which the slowness of my typing and thinking
> speed is the bottleneck.

What people should really pay attention to here is the clear DIFFERENCE 
between the SPEC results (which show the P4 or Xeon and the G5 
more-or-less evenish, with a slight edge going to the G5) and the 
Real-World results (which consistently showed the dual-G5 machines 
doing things about TWICE as fast as a comparable dual-Xeon setup).

The reason for this discrepancy? IMO, this is partially the superior fp 
ability on the G5, but *mostly* due to the addition of cutting-edge 
technologies in every other part of the machine -- serial ATA, 
hypertransport, 1GHz frontside bus, very very fast RAM, improved ASIC, 
etc.

This clearly shows that the G5 has to "fight" less with the rest of the 
system to do its work than the P4 or Xeon does. I predict that when the 
real-world benchmarks come out for the lower (single-proc) G5 models, 
they will still be fast -- very comparable to P4s with 150% the GHz 
rating -- but more like MATCHING them in speed overall rather than 
SURPASSING them in speed the way the dual G5 does with the dual-Xeon.

_Chas_

"To use the Mac is to be confronted, over and over, with the idea that 
the most mundane task can be done artfully and compassionately, 
beautifully and invitingly. " -- Glenn McDonald