[Sparkletron]

<http://www.macworld.com/2004/11/news...ench/index.php>

So a single 1.8 G5 has a higher average Unreal Tournament frame rate than a dual 1.8 G5?

-S

[Lateralus]

The video card does make a difference, you know.

[Sparkletron]

Originally posted by Lateralus:
The video card does make a difference, you know.

All 1.8 stock G5 configs use the same NVIDIA GeForce FX 5200 Ultra GPU.

-S

[jamil5454]

And I doubt UT takes advantage of dual procesors.

[Sparkletron]

Originally posted by jamil5454:
And I doubt UT takes advantage of dual procesors.

I agree. Which perhaps explains why UT is not faster on a DP, but does not explain why it's slower.

-S

[SSharon]

shouldn't dual proc. machines still run faster than their single procesor counterparts even with apps like UT? I think that even thought UT is limited to one proc. the system which is still running can use some of the second chip but must compete for cycles in a sincle chip G5.

[macaddict0001]

true it can also offload sound to a different processor but this is definitly not completely accurete anyway because for some reason the two ibooks are also getting different scores despite having the same specs.

[Pierre B.]

Originally posted by jamil5454:
And I doubt UT takes advantage of dual procesors.

The result is really strange. It is well known that UT2004 can use both processors: it offloads the sound on one processor and it calculates the rest on the other. And this makes a big difference. Look here (in the insights after the graphs).

I would say there is some error in the Macworld benchmarks (or they run the test with sound off).

[Mafia]

Originally posted by Pierre B.:
I would say there is some error in the Macworld benchmarks

i think this is the truth.

[robo74]

I think something is fishy there as well.

Look at barefeats... I tust this guy more then anything... I can match his tests often with my own machines.

Real benchmarks on G5 updated


Rob

[a2daj]

The problem could be with the botmatches. They don't always run the same sequence. It all depends on the seed number (the rand value in the benchmark results). I can run one day with the same seed number but the next day the benchmark might decide to use a different value, resulting in different results. Then if you run the timedemo on a different machine, you'll most likely be using a different seed number and thus different results.

While barefeats does have lots of tests, a lot of the time they're not nearly as comprehensive as I'd prefer. Plus some of the conclusions are a bit out there. I prefer Mike Breeden's Accelerate Your Mac site. Unfortunately, he's been really busy with real life matters and hasn't been able to do a lot of reviews lately but typically, he does a better job than bf.

[robo74]

Mike does do a good job, I just threw barefeats out there because macworld just didnt make sense.

Rob