Multi CPU Comparison

[wickedw]

Hi All,

Newbie here, Loving the new Perf 7 software, I am thinking of / using to help me choose our new high performance web server, I am mainly interested in the non graphics Marks areas of testing, but think the overall score should be useful also.

Does the CPU Mark reliably meaure the differences between single, dual, quad CPU (not core) multi-core setups in its tests?

I assume it does but just saving myself an expensive mistake.

Also, Can i say a score CPUMark 2000 is 2x faster than 1000?

I also noticed that as my 2d/3d tests fail on our servers (rightly so) BUT then they get stupid nos like 77,000 which could be seriously skewing their overall mark, are you aware of this? is it a bug?

Thanks for your help,
Matt

[David (PassMark)]

Does the CPU Mark reliably meaure the differences between single, dual, quad CPU (not core) multi-core setups in its tests?

The test maxes out all the CPUs and Cores available. See also the graph notes.
http://www.cpubenchmark.net/graph_notes.html

Can i say a score CPUMark 2000 is 2x faster than 1000?

This is true, but only for certain CPU intensive tasks that use all Cores / CPUs.

If the task involves disk or internet access then a machine with a CPUMark of 2000 might not seem any quicker than a machine with a CPUMark of 1000.

I also noticed that as my 2d/3d tests fail on our servers (rightly so)

Why should the 2D test fail? Even the most basic graphics card should support the 2D test. It might be normal for the 3D test to fail if you have a card that isn't a 3D card.

BUT then they get stupid nos like 77,000

Can you give more detail about this. What hardware & what tests return these scores? Is there any error message for the 2D/3D failure.

[wickedw]

Mr

Thanks for reply.

Yeh, I think its just 3d tests as you say, heres the messages in order, I can send you pt file if that will helps.

http://tinypic.com/r/2z845fl/3

[David (PassMark)]

I think the warning messages about 3D is normal. Just looks like you don't have a 3D card installed. Which is normal for a web server.

The unrelated warning about SMBIOS access is more in interesting. If you want to investigate further run PerformanceTest in debugmode. And E-Mail or post the log.

None of the screen shots you linked to cover the 77,000 number you mentioned, so can't really comment on that.