Login

Up to date Cuda guide ?

Tips & tricks, working results, technical support

Re: Up to date Cuda guide ?

Postby kels » Sun Nov 17, 2013 3:48 pm

I'll receive the Tesla GPU in 2 weeks for a 24h testing :o

In the meantime, could you tell me if the CudaBridge can take advantage of some SLI configuration ? I know CUDA by itself will support it but it's up to applications to take advantage of multiple CUDA devices.

Think of dual GTX680 4Gb Edition for example.

thank you
kels
User Level XI
User Level XI
 
Posts: 166
Joined: Fri Oct 11, 2013 8:43 am

Re: Up to date Cuda guide ?

Postby kels » Fri Nov 22, 2013 10:02 am

Ok, I have now a fully working setup including both server side CPU and CUDA GPU (although I'm still waiting for another GPU ...)

I have very good results so far with --> MASTER page :

OPTTIMED ---> 15CUDA ac 2Mono
OPTFREQU ---> 14CUDA ch 2Mono

but is this correct ?

Because it looks like here you are doing only the FREQU on CUDA :
http://www.youtube.com/watch?v=pdlfjG7qOKc

thank you
kels
User Level XI
User Level XI
 
Posts: 166
Joined: Fri Oct 11, 2013 8:43 am

Re: Up to date Cuda guide ?

Postby enriquesilveti » Fri Nov 22, 2013 11:38 am

kels wrote:Ok, I have now a fully working setup including both server side CPU and CUDA GPU (although I'm still waiting for another GPU ...)

I have very good results so far with --> MASTER page :

OPTTIMED ---> 15CUDA ac 2Mono
OPTFREQU ---> 14CUDA ch 2Mono

but is this correct ?

Because it looks like here you are doing only the FREQU on CUDA :
http://www.youtube.com/watch?v=pdlfjG7qOKc

thank you


OPTTIMED ---> 15CUDA ac 2Mono
Enrique Silveti.
Acustica Audio customer and technical support.

MBP 11.5 (i7-4870 | 16 GB | 512 SDD)
SP4 (i5-6300 | 8 GB | 256 SDD)
UFX | Lyra2 | USBPre2
VM U15 | VM Win10 | VM OSX 10.12
N4/NAT4 | SPX3 | RX5 | LN2C | Smaart8 | R5 | PT12 | PX10 | NIK5
User avatar
enriquesilveti
Expert
Expert
 
Posts: 2663
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

Re: Up to date Cuda guide ?

Postby kels » Fri Nov 22, 2013 11:51 am

thx for OPTIMED but sorry, I don't get the whole point :roll:
should I leave the OPTFREQU to ---> 11IPPS ch 2mono ?
kels
User Level XI
User Level XI
 
Posts: 166
Joined: Fri Oct 11, 2013 8:43 am

Re: Up to date Cuda guide ?

Postby enriquesilveti » Fri Nov 22, 2013 12:14 pm

kels wrote:thx for OPTIMED but sorry, I don't get the whole point :roll:
should I leave the OPTFREQU to ---> 11IPPS ch 2mono ?


11 > CPU
15 > CUDA
Enrique Silveti.
Acustica Audio customer and technical support.

MBP 11.5 (i7-4870 | 16 GB | 512 SDD)
SP4 (i5-6300 | 8 GB | 256 SDD)
UFX | Lyra2 | USBPre2
VM U15 | VM Win10 | VM OSX 10.12
N4/NAT4 | SPX3 | RX5 | LN2C | Smaart8 | R5 | PT12 | PX10 | NIK5
User avatar
enriquesilveti
Expert
Expert
 
Posts: 2663
Joined: Sun Mar 28, 2010 9:00 pm
Location: Lodi | Madrid | Buenos Aires

Re: Up to date Cuda guide ?

Postby kels » Fri Nov 22, 2013 12:28 pm

so, I was ok with :

OPTTIMED ---> 15CUDA ac 2Mono
OPTFREQU ---> 14CUDA ch 2Mono

everything from CUDA per Nebula instance
thank you Enrique
kels
User Level XI
User Level XI
 
Posts: 166
Joined: Fri Oct 11, 2013 8:43 am

Re: Up to date Cuda guide ?

Postby jorismak » Fri Nov 22, 2013 3:49 pm

Did you get an answer kels, because your question is still unanswered in this thread I think.

11 is CPU, 15 is Cuda. Ok. But what is the difference between 14 cuda and 15 cuda?

And the Cuda FAQ states to only enable cuda for FREQD mode, and use the regular mode for OPTTIMED.

So I still see no answer to the question "does TIMED work on Cuda or not?".

BTW, I thought Cuda was only relevant if you used very long reverbs, and it would actually be a lot slower on 'simple' stuff like Pre's, EQ's, Tape, Dynamics, etc...

I tried enabling cuda on my (now sold) gtx660, and then tried it with the gtx760 I have now, but Reaper would render at 0.15 realtime with cuda, and 0.8 with pure CPU. It all depends on your libraries and settings you use ofcourse, but for me it seemed to be a lot slower instead of faster :).
jorismak
Member
Member
 
Posts: 344
Joined: Fri Nov 16, 2012 4:49 am

Re: Up to date Cuda guide ?

Postby kels » Fri Nov 22, 2013 4:32 pm

But what is the difference between 14 cuda and 15 cuda?


that was the point, indeed, thank you. It looks like Enrique doesn't like CUDA a lot :)

In the meantime, I did some samples rendered with various combinations -->

OPTTIMED : 5 (regular)
OPTFREQU : 11 (regular)

OPTTIMED : 15
OPTFREQU : 14

OPTTIMED : 15
OPTFREQU : 11 (regular)


From a sound point of view, all samples are the same (did the -180° test)
From a render point of view, speeds are similar (0.1 difference) in favour of cpu.

OPTTIMED : 5 (regular) /
OPTFREQU : 14 / ALL OF THOSE
OPTTIMED : 5 (regular) / DO NOT WORK IN MY CASE
OPTFREQU : 15 /

I have the GPU in a server config where CPU is 16% loaded and GPU 265% (funny but still realtime audio, no problem - must be a problem of loading evaluation!)

So, from what I see, OPTIMED do work with CUDA this config really unload the cpu :
OPTTIMED : 15
OPTFREQU : 14
kels
User Level XI
User Level XI
 
Posts: 166
Joined: Fri Oct 11, 2013 8:43 am

Previous

Return to Working with Nebula

Who is online

Users browsing this forum: Bing [Bot] and 1 guest