Intel Xeon 5600-Series: Can Your PC Use 24 Processors?

[tsmithf]

[citation][nom]jecastej[/nom]I have the same 8 core Xeon since 2008 (yes a Mac Pro). In 3 years (2006-2008) I went from one dual-core 2.3 PPC machine to an 8 core Xeon and it was....[/citation]

Not strictly an 8 core Xeon, it's dual Quad Core, a lot of Mac owners seem to get confused over this one based on Apple's slightly deceptive marketing.

Plus I wouldn't pin your hopes on the CUDA rendering ability of your graphics chip just yet unless your quality expectations are quite low, it's good for quick rough work, but not ideal for final output, dependant on your use of course.

 

[luke904]

WHAT?!?!

NO GAMING TESTS?!?!


lol just kidding


would of been fun to see though.. just for the hell of it

 

[FiRem00]

He used 4 3Gb modules in this test? If it's triple channel with 12 slots total, he'd need to use 3 modules minium each per cpu to use triple channel, so surely it's using single channel speeds in this test?

 

[Draven35]

Wonder how it runs Inventor or Solidworks, I like responsive, but wonder if that would be over-kill.

I developed the tests and would like to get some Soldiworks tests in (and some more CAD-related tests in general) but there's issues with licensing.

Although Lightwave 3D is a good program, there's greater chance that the more expensive 3D modeling and Animation programs would have better support for dual Xeons. Would have preferred to see you test the dual xeons with the heavyweights such as Maya 2011, Softimage XSI, 3DS Max 2011. Movie companies and most game companies use the heavyweight programs instead of Lightwave.

Many television visual effects companies use Lightwave alot. It is what I know how to use, and was able to develop the tests in it. The Lightwave tests represent alot of work, centered around the 1.68 million polygon THG logo piece.

I'm working on making Max and Maya 2011 tests (the render tests will use the same logo animation) and getting comparable render quality results from mental ray is daunting... not to say it is 'hard', but it looks like the render times will be significantly longer than in Lightwave (hence why Digital Domain uses the Lightwave renderer for much of their commercial work and some of their feature film work.)

 

[Guest]

I am sure the correct threading settings are NOT set in Lightwave. It should work but only for rendering. The max. setting is 16. Unless you use a different engine which can go up to 512cores.

 

[Draven35]

Lightwave's thread settings were "Automatic"

 

[Guest]

Lightwave 9.6 only supports up to 16 threads, hence the not close to linear scaling. I think that with the new version that allows the needed 24 threads, the results will be quite much better and closer to a 2x improvement over the 980X.

Really appreciate you added the Lightwave benchmark. (and do totally agree on your comments on the other packages) Great review.

Now overclock them on the EVGA SR-2 and run the tests again? Please?

 

[sandsz]

My face just melted off

 

[Draven35]

Told you not to look into the ark.

 

[specialk90]

I have been helping to test Premiere CS5 with my Z800 and dual X5660 & 24GB ram. One thing we have noticed is that 4GB of ram per core more than doubles performance when exporting MPEG2. I can't believe you would test 12 cores/24 threads with ONLY 12GB of friggin ram. To make it a fair test between platforms, you need the same amount of ram per core. Redo the test with 24/32/48 for each setup respectively.

 

[Draven35]

We didn't do any MPEG-2 encoding from Premiere anyway... and you really shouldn't need 4GB per core for MPEG encoding.

 

[Draven35]

[citation][nom]kansur0[/nom]I have written in the forums to ask that workstation and consumer graphics cards be tested to see if performance differs between cards. My request has not been answered.[/citation]

The Quadro 4800 review and FirePro 8700 review did both touch on this...

 

[Draven35]

When the tests were developed, CS5 was not available. CS5 became available when this benchmarking was already in progress.