Skip to main content
L
LJ_Designer
Participant
August 22, 2012
Question

OpenCL vs Cuda for CS6

  • August 22, 2012
  • 2 replies
  • 33823 views

First of all, My English is not very well please bear with me. I do like to know how comparison with a GTX 6xx and an AMD 7xxx. This discussion is about OpenCL vs Cuda for CS6 programs and general for PS, video-editing and 3D rendering.

I already know about performance in gaming benchmark, value, performance ratio, and etc. I'm from eggxpert in chat room/forums. I'm helping to people who needs help with PC parts and build. Any people usually come to eggxpert ask for PC custom building for a gaming, basic, and needs. Also asking for issue in his or her PC hardware or software. Sometimes asking for help about workstation heavy using rendering (cad, lightwave, autodesk, etc.), more rams, more Hyper-Threading/Cores, and professional graphic cards. We might not best expert in that parts. I always ask them go to adobe forums then I linked for them. However, there are only few people who are graphic design and video-editing such as 2600/3770 w/o "K" if need overclocking or about close price, any GPUs GTX cards or AMD 7750/7770 or higher if buget fit. 2600/3700 8 HT great for "Intermediate" use with any programs like CS6, Cinema4d, and etc. I learned from Harm Millaard. Basic, Intermediate, and Difficult. Budget, Economical, and Warrior. Best bang for the buck for CPU.That is where I googled it and saw in forums about his wrote. Most of us (eggxperts) like this guy and some people too. This forum is great expert with good wise ratio, value, performance, and etc. just like we did in eggxperts but just issue, gaming, general, and PC parts.

I would like to learn more about OpenCL and Cuda. Some forums said OpenCL is faster then Cuda. Some people said CS6 is using OpenCL only. I checked Nvidia Site updated CUDA 5.0 to work with CS6 (GTX 6xx.) Also I checked adobe features for OpenCL there are many use features. Nvidia only some feature like color corrector, warp stabilizer, uninterrupted playback, and mutli cam support from Premiere Pro CS6 and some program. CPU would do work with other features anyway. Can you explain to me about OpenCL vs Cuda in performance/benchmark for those programs?

Thank you

This topic has been closed for replies.

2 replies

Harm_Millaard
Harm_Millaard
Inspiring
December 8, 2012

For Windows it is generally speaking like this:

CS5+ PR uses DirectX and only with nVidia cards with more than 896 MB of VRAM can CUDA acceleration be turned on.

CS5+ AE uses OpenGL and only with nVidia cards with more that 896 MB VRAM can GPU assisted Ray Tracing be turned on.

At this moment AMD/ATI cards can not use hardware acceleration at all. OpenCL is not yet supported for hardware acceleration. There are plug-ins for AE and PR that have real time rendering that can use both DirectX and OpenGL, even in PR.

The most decisive factor for hardware acceleration performance is the memory bandwidth, more so than number of CUDA cores.

On our http://ppbm5.com/DB-PPBM5-2.php benchmark results, the average gain from using a CUDA capable video card over software is around a factor 12 for rendering.

If you want some more background on building a system, look here: http://ppbm7.com/index.php/intro-part-1 including the rest of the pages.

R
RjL190365
Legend
August 22, 2012

Which type of system are you going to run? Windows? OSX?

If you are running Windows, keep in mind that the Windows version of Premiere Pro CS6 still does not support OpenCL at all for GPU acceleration. Therefore, all AMD GPUs will force Premiere to run only in MPE software-only mode. The same goes for any non-NVidia GPU, or an NVidia GPU that either has less than 1GB of total video RAM or is too old to support CUDA. MPE GPU acceleration is still NVidia CUDA only in Windows.

If you are running a Mac (OSX), it does support OpenCL for MPE GPU acceleration - but keep in mind that this support is officially limited to certain HD 67xxM GPUs inside newer MacBook Pros running on newer versions of OSX.

Hope this helps.

W
woodybrando
Participant
December 8, 2012

which is faster cuda or opencl acceleration? are there any benchmarks?

thx,

Jayson

AWDEfilms.com

ECBowen
ECBowen
Inspiring
December 10, 2012

The Question of which GPU processing language is faster is really not answerable. You have to be careful on what you read here. If an Application is developed with CUDA as the primary acceleration language and OPENCL is added later, the CUDA acceleration seems to be faster. If the application is developed in Open CL then the ATI cards seem to be faster than the Nvidia cards with Open CL. That is really about as much of a break down as you can get because the ram optimizations really decide so much of the GPU acceleration performance. So the actual deciding factor between the 2 is caching models versus available ram.

Eric

ADK

Terms & ConditionsAccessibility statement
Using the community Terms of use Privacy Cookie preferences Do not sell or share my personal information ad choicesAdChoices