Hey Logan!
You been saying you like OpenCL for awhile now.
But why? I don't think I heard you mention it before.
Hasn't been Premiere Pro supporting CUDA for a much more longer time?
Regards,
Leonard
Hey Logan!
You been saying you like OpenCL for awhile now.
But why? I don't think I heard you mention it before.
Hasn't been Premiere Pro supporting CUDA for a much more longer time?
Regards,
Leonard
I'm no expert on Premiere but Adobe only began with specific Cuda support in their software because that was available before OpenCL was sort of ready, now they're all about OpenCL because that's an actual standard. Today you can see Nvidia mention CUDA when what they're really talking about is OpenCL, but that's to be expected.
ouh okay. but what does that mean? Adobe is going to put more support into AMD cards now?
OpenCL is for everyone and Nvidia aren't bad at it either. It's the future and Nvidia will have to, and constantly are, looking for other things to lock users in with.
thats just so sad for people, I have been thinking of switching to AMD for awhile now but the support of CUDA just prevented me from doing so as I am still using PP CS5
I want to just clarify something.
CUDA is not Nvidia's OpenCL implementation. CUDA is a computing architecture, while OpenCL is a language.
There is a lot more to CUDA than just OpenCL.
As far as OpenCL vs CUDA with Adobe, they have supported both for a while now.
OpenCL support with Adobe products works with a wider range of GPUs than just Nvidia.
so how does OpenCL compared to CUDA? since their totally different (hardware vs software)?
This question doesn't really work.
Open Computing Language (OpenCL) is a framework for writing programs that execute across heterogeneous platforms consisting of central processing units (CPUs), graphics processing units (GPUs), digital signal processors (DSPs), field-programmable gate arrays (FPGAs) and other processors.
Where CUDA is something else.
CUDA, which stands for Compute Unified Device Architecture,[1] is a parallel computing platform and application programming interface (API) model created by NVIDIA.[2] It allows software developers to use a CUDA-enabled graphics processing unit (GPU) for general purpose processing – an approach known as GPGPU. The CUDA platform is a software layer that gives direct access to the GPU's virtual instruction set and parallel computational elements.
From Adobe's perspective, Premiere needs to render frames. It can either use the CUDA API and send the frames to the GPU for rendering, or it can use it's OpenCL version of their rendering engine to send frames to the GPU. In the end they basically do the same thing.
As far as performance and comparison between the two, You get substantial decrease in rendering times in 1080p, and less so and sometimes slower rendering times with 4k. (with your average consumer market GPU)
As far as performance and comparison between the two, You get substantial decrease in rendering times in 1080p, and less so and sometimes slower rendering times with 4k. (with your average consumer market GPU)
In that statement which one is faster? Is this possibly why Logan likes openCL more?
My guess would be Logan likes OpenCL more because it is Open. The standardisation allows for everyone to be able to use it, but still make different implementations. And being open source means that the manufacturers can optimize without paying for licensing and stuff, and Logan have a lot of love for open standard software.
In regard to which one is faster, it's very dependent on the GPU that is in your system.
Oh yeah totally!
I mean for a dollar to dollar ratio, which one would you be more happy with? like I have a gtx680 right now, what would be a equivalent of that from AMD?
In my benchmarks the Photoshop was much slower (2x) with OpenCL ("Use OpenCL" checked in performance setting) but I did not test yet if it makes difference in Premiere. This is on Quadro K5000.
Premiere does not even support second CPU for rendering because of a "bug" which they could not fix for the last 2 versions, bad OpenCL support, so I am not surprised that they are lazy about implementing CUDA.
No wonder OpenCL is way worse than CUDA on Quadro...