V-Ray and using of processors???

[smail_cg]

Hi to all,
on my new comp with Core i7 - 720QM (4 cores, 8 threads - Win7 64bit, I try some my old scenes with V-Ray (trying on Rhino 4 SR8 and also on last Rhino 5 Beta and V-Ray for Rhino version 1.05.29). But I'm very surprised that the V-Ray couldn't use 95-100% of processor (cores) - for example for computing the Lights Cache V-Ray/Rhino use only 31% of processor, same it is in Irradiance Map, only in final computing of scene it can use about 95%...
Also processors (cores) dosn't work on the heightest frequency (only about 1150 MHz from 1600MHz)...
With my old notebook with Pentium M 1.5 the Rhino/V-Ray use for all computing 100% time of processor
Also that problem is why the computing times of scenes are not so better compare with my old comp...


thanks for your reply!
jhe

here is the print screen with V-Ray computing...

[ferrosoa_cg]

Re: V-Ray and using of processors???

Hi to all,
on my new comp with Core i7 - 720QM (4 cores, 8 threads - Win7 64bit, I try some my old scenes with V-Ray (trying on Rhino 4 SR8 and also on last Rhino 5 Beta and V-Ray for Rhino version 1.05.29). But I'm very surprised that the V-Ray couldn't use 95-100% of processor (cores) - for example for computing the Lights Cache V-Ray/Rhino use only 31% of processor, same it is in Irradiance Map, only in final computing of scene it can use about 95%...
Also processors (cores) dosn't work on the heightest frequency (only about 1150 MHz from 1600MHz)...
With my old notebook with Pentium M 1.5 the Rhino/V-Ray use for all computing 100% time of processor
Also that problem is why the computing times of scenes are not so better compare with my old comp...


thanks for your reply!
jhe

here is the print screen with V-Ray computing...

set in light cache passes to 8 ....

[LeoBonilla]

Re: V-Ray and using of processors???

Yes in the light cache Tab inside the V-Ray options you shoul equalize the number of Phases to the same number of threads.
Example:

i7 = 8 threads = 8 Num. Phases
Dual Xeón = 16 threads = 16 Num. Phases

[dkendig]

Re: V-Ray and using of processors???

Increasing the number of passes will increase the quality of the render, and it will use more of your processor, but it's not going to speed up the render any. Just to be clear, having more passes, means you will have more samples, and the output will be higher quality, but it shouldn't really speed anything up.

[smail_cg]

Re: V-Ray and using of processors???

Hi, thanks for reply...
yes I change the number of Phases to 8, the all processor (threads) are now using for 100% and the time speed up.
On very basic render is time with 2 Phases 3min 29s, with 8 Phases 2min 51s , and also with 8 Phases the render is a little bit more detailed.
It's also interesting this>
8 Phases > Processor using is 100% at 1.6 - 1.73 MHz - so using of Turbo Boost of i7 is very small
2 Phases > Processor using is about 40% at 2.0 -2.4 Mhz - Turbo Boost is high... It seems that the processor is trying to work on higher frequency with less usage...

[jonstinton]

Re: V-Ray and using of processors???

Yes in the light cache Tab inside the V-Ray options you shoul equalize the number of Phases to the same number of threads.
Example:

i7 = 8 threads = 8 Num. Phases
Dual Xeón = 16 threads = 16 Num. Phases

I have a quad core Xeon processor, so should I put in 32 Num. Phases?

[indielyness]

Re: V-Ray and using of processors???

what if you're using distributed rendering? do you equal the number of threads to the total amount of all the computers or just the main one

i.e main pc dual core + 3 slaves with dual core = 8 threads?

[dkendig]

Re: V-Ray and using of processors???

it's not phases guys... read the screen, it's passes. Here's the manual:
http://www.spot3d.com/vray/help/150S...s_lightmap.htm

Number of passes - the light cache is computed in several passes, which are then combined into the final light cache. Each pass is rendered in a separate thread independently of the other passes. This ensures that the light cache is consistent across computers with different number of CPUs. In general, a light cache computed with smaller number of passes may be less noisy than a light cache computed with more passes, for the same number of samples; however small number of passes cannot be distributed effectively across several threads. For single-processor non-hyperthreading machines, the number of passes can be set to 1 for best results.

Update: I'm an idiot for assuming we spelled this correctly. LEOBONILLA gets 10 points for putting me in my place!

[LeoBonilla]

Re: V-Ray and using of processors???

it's not phases guys... read the screen, it's passes. Here's the manual:
http://www.spot3d.com/vray/help/150S...s_lightmap.htm

Number of passes - the light cache is computed in several passes, which are then combined into the final light cache. Each pass is rendered in a separate thread independently of the other passes. This ensures that the light cache is consistent across computers with different number of CPUs. In general, a light cache computed with smaller number of passes may be less noisy than a light cache computed with more passes, for the same number of samples; however small number of passes cannot be distributed effectively across several threads. For single-processor non-hyperthreading machines, the number of passes can be set to 1 for best results.

I'm Sorry devin but in the VFR UI it says Num. Phases......so?

[LeoBonilla]

Re: V-Ray and using of processors???

Yes in the light cache Tab inside the V-Ray options you shoul equalize the number of Phases to the same number of threads.
Example:

i7 = 8 threads = 8 Num. Phases
Dual Xeón = 16 threads = 16 Num. Phases

I have a quad core Xeon processor, so should I put in 32 Num. Phases?

You have the same cores I have.

1 quad core = 4 cores = 8 Threads
Dual Quad core= 16 threads = 16 Phases

[dkendig]

Re: V-Ray and using of processors???

::smacks forehead:: pwned! Well done sir, well done. That'll teach me to assume we spelled it right.

[Micha_cg]

Re: V-Ray and using of processors???

My impression was, that the subdiv count set the sum sample count for all pases together and not for each. So, more passes/processors should cause, that the sum sample count is faster reached.