[jbaugh]

I've had my quad G5 now for about 2 weeks and I am concerned about its performance. I'm worried that I may have a hardware defect of some kind.
My configuration is:
Quad G5
GeForce 6600
Total RAM 6 Gigabytes
Two 500 Gigabyte internal hard drives.

When I run Xbench I get much lower scores than I have seen posted for this machine. For example my "CPU Test" score is 126 versus the 243 I've seen posted elsewhere. Also on xbench my "memory test" score is 135 versus the 517 score that has been reported. The Cinebench 2003 rendering score is 820 compared to the 1091 I see reported on another website. But on the other hand, I get between 29 and 32 Gigaflops on the Altivec Fractal test where 25 is the value on the test web site.

To summarize
My results: Results posted elsewhere for the
quad G5:
CPU Test 126 243
Memory Test 135 517
Cinebench 2003 820 1091
Altivec Fractal 29-32 GFLOPS 25 GFLOPS
Fill test 7736 9840

In all of these tests, with the exception of the Altivec test, my scores are much lower than elsewhere reported. What should I do? Should I be alarmed?
Thanks
John

[shell]

What kind of memory do you have, did you install it yourself?

[Luca]

Go to System Preferences -> Energy Saver -> Set processor performance to "Highest."

By default, the PowerMacs run at a lower speed until you max out the processor, at which point they switch to their higher speeds. This is quick and usually not noticeable, but it may be affecting benchmark scores.

Let us know if that works.

[jbaugh]

Originally Posted by Luca Go to System Preferences -> Energy Saver -> Set processor performance to "Highest." By default, the PowerMacs run at a lower speed until you max out the processor, at which point they switch to their higher speeds. This is quick and usually not noticeable, but it may be affecting benchmark scores. Let us know if that works.

I've had the processor performance set to "Highest" from day one. I just checked to make sure and it still is set to "Highest".

My memory consists of 2 x 1Gig stock RAM from Apple and 4 X 1Gig RAM I installed myself that I bought from Other World computing. Repeated hardware tests with TechTool Pro suggest that RAM and other components are OK.

Here are the Cinebench 2003 test results. The results for single and multiple CPU rending are much lower than others are reporting for the quad.

CINEBENCH 2003 v1
************************************************** **

Tester : Jb

Processor : Quad G5
MHz : 2.5 GHz
Number of CPUs : 4
Operating System : OS 10.4.3

Graphics Card : GeForce 6600
Resolution : <fill this out>
Color Depth : <fill this out>

************************************************** **

Rendering (Single CPU): 273 CB-CPU
Rendering (Multiple CPU): 835 CB-CPU

Multiprocessor Speedup: 3.06

Shading (CINEMA 4D) : 345 CB-GFX
Shading (OpenGL Software Lighting) : 918 CB-GFX
Shading (OpenGL Hardware Lighting) : 1827 CB-GFX

OpenGL Speedup: 5.30

************************************************** **
If bad RAM is a consideration I suppose I could temporarily remove the 3rd party RAM and retest?
What do you think?
John

[Luca]

Well, go ahead and try it, but I am pretty sure most benchmarks specifically try to isolate a single component to get the most objective rating possible. There shouldn't be much difference in the CPU test on a G5 with 128 MB of RAM vs. 2048 MB of RAM (theoretically of course). If your RAM is slow, it should show up in Xbench's RAM test, not in the CPU test. But like I said, it's always a possibility so maybe it's a good idea to try that anyway.

I don't know what else would be causing it, unfortunately.

[MacRonin]

Best way to benchmark your Quad G5...?!?

By sending it to my house...!?!

;^p

[v.noir]

Damn, I was about to say that. Curses, foiled again!

[jbaugh]

Simple solution. There is no problem with my quad. The problem is with my feeble brain. Mike, at XLR8yourmac.com pointed out to me that there are two versions of Cinebench, one optimized for the G5. Unfortunately, I was using the non-optimized version to obtain my test results.

I downloaded the G5-optimized version and now my test results are in line with those others are posting:
CINEBENCH 2003 v1
************************************************** **

Tester : jb

Processor : Quad G5
MHz : 2.5 GHz
Number of CPUs : 4
Operating System : 10.4.3

Graphics Card : GeForce 6600
Resolution : <fill this out>
Color Depth : <fill this out>

************************************************** **

Rendering (Single CPU): 360 CB-CPU
Rendering (Multiple CPU): 1012 CB-CPU

Multiprocessor Speedup: 2.82

Shading (CINEMA 4D) : 356 CB-GFX
Shading (OpenGL Software Lighting) : 1063 CB-GFX
Shading (OpenGL Hardware Lighting) : 1898 CB-GFX

OpenGL Speedup: 5.33

************************************************** **
Thank you all for your input. Yes, I'm embarrassed. But it will pass. I'm just happy all is OK with my machine.
John

[Moogs]

By using it for working / making money, then if money comes in faster than it did before, you have an efficient machine.

[Powerdoc]

With Xbench I have an overall level of performance of 140 (2 GB ram). That's mean that the quad is 40 % faster than a dual 2 ghz with a geforce 6800 ultra.

BTW Xbench sucks : some people have better performance with a dual 2,5 than my quad (perhaps they where tweaking the bench ...)

[Gargoyle]

Forget all those crappy number crunchers. Think of something useful and perform it on two machines and see which one wins!

eg. Get some raw uncompressed video and turn it into H.264 (Make sure you get the same settings on both machines).

Disclaimer: I have absolutely no idea of quicktime uses multithreading when encoding video! You might want to try encoding 4 streams at the same time!