My dual xeon E5 2650-v2 slower than my old core i7 920 !

[ghiom3d]

Hello,

I need your help.
I'm using BB to render a scene with computer A and B

Computer A is the one on wich i built the scene and wich hold the scene and maps.
dual Xeon E5 2650 v2 ( 2x8 cores / 2x16Threads ) + 32 Gb DDR3
So, the computer A is 16 Cores / 32 Threads and run well at 2992mhz it's a young Pro HP with Factory watercooling

Computer B is where BB Manager is running and old the Vray dongle .
Single Old Core i7 920 ( 4 cores / 8 threads ) + 12 Gb DDR3 and run at 3200mhz it's an old Dell XPS for Gamer with Factory water colling
This computer B is theoricaly 3 or 4 or 5 times slower than Computer A.

the maps are saved with the scene when i export to BB job.
I check BB queue monitor and i see that computer A compute a frame in 8minutes and 19 secondes when computer B do the same in 8 minutes !
i look at the renders frames and i saw 8 buckets on computer B and 32 buckets on computer A. Wich is correct ..

So what is the problem

[Morbid Angel]

when computer a real renders does the cpu show 100% load?
also do you have dynamic memory and bitmap pager enabled? this could affect the performance

[vlado]

What Dmitry said; the most important one is to check that the Bitmap pager is disabled. It has terrible performance on multicore machines. If you are using ForestPro or RailClone in your scenes, make sure they are updated to the latest versions.

Best regards,
Vlado

[ghiom3d]

Thanks you Dmitry and Vlado

CPu are all at 99%-100%
I'm not using Bitmap Pager, it's disabled and i haven't Forest pack or railclone plugins .

[vlado]

Ok then; next questions: which V-Ray version is it? Can you get me the scene to vlado@chaosgroup.com? There might be other things that slow the multicore machine down; for example in a recent case I had, it was a GradientRamp texture that was causing it, but it can be any number of other things.

Best regards,
Vlado

[jstrob]

Ok then; next questions: which V-Ray version is it? Can you get me the scene to vlado@chaosgroup.com? There might be other things that slow the multicore machine down; for example in a recent case I had, it was a GradientRamp texture that was causing it, but it can be any number of other things.

Best regards,
Vlado

Wow! I didn't know about that. I guess you meant multi-cpu because i7 are also multi-cores. I have only multi-cpu dual xeon in my render farm. So It could be good I add an i7 just to know when multi-cpu problems happen. How can a gradientRamp texture slow down a multi-cpu machine?

[vlado]

Wow! I didn't know about that. I guess you meant multi-cpu because i7 are also multi-cores. I have only multi-cpu dual xeon in my render farm. So It could be good I add an i7 just to know when multi-cpu problems happen. How can a gradientRamp texture slow down a multi-cpu machine?

There is some very old code in it that scales poorly with many cores. But like I said, it can be due to other things as well, so best to take a look at the scene.

Best regards,
Vlado

[vance3d]

Any updates on this topic? I'm running into some odd slowdowns on multi-cpu machines too. It would be great to have a list of all the things in max that don't scale well like the gradient ramp mentioned earlier.

thanks,

V Miller

[vlado]

So far, the Bitmap pager, older versions of V-Ray/ForestPro, the Gradient Ramp have been the usual culprits. If you are in doubt, it would be best to get me a scene so that we can profile it.

Best regards,
Vlado

[vance3d]

I think I'll do that after I get through my current rendering blitz. Would using bercon gradient be a better choice than max's gradient ramp?

V Miller