Advanced Micro Devices (AMD), if they are so advanced, then why is CUDA> Openc CL?

Is OpenCL ready for practical use yet? How can I know it’s progress?

OpenCL is being used for a wide variety of things, just not Blender. Moreover, AMD is not OpenCL. AMD supports OpenCL.

To keep tabs on Blender’s progress, you can track this page:

To keep tabs on OpenCL’s overall progress, go straight to the source:

If you need Blender OpenCL support on AMD cards, you can buy Indigo Renderer, which is supposed to be fully AMD-ready. I don’t know how they are managing the memory problem that is plaguing other projects, but they did.

I suspect that AMD dragged its feet when it came to developing better tool sets and drivers for Radeon-based OpenCL processing for the same reason that Nvidia cripples its GeForce cards: they want people to buy their wildly more profitable workstation and compute cards. I think this is stupid, backwards thinking. I know neither company wants to admit it, but I think that workstation cards are going the way of the dodo.

AMD has been around since the 90’s at the least, and that was back when they were actually leapfrogging Intel from time to time (introducing concepts like 64 bit before they did).

So at one point, their devices actually were advanced, and even now they’re quietly working on new concepts like unifying the CPU and GPU along with a new ‘compute core’ type and new API’s like Mantle and HSA in an attempt to someday be the leader in innovation once again.

Actually OpenCL is pure Apple invention, back in time ATI was with strong relation with Apple and obviously they have some support, and looking how things work in 2014, ATI/AMD actually do not very interested with OpenCL, despite loud marketing claims (photoshop acceleration and bitcoin, that is all after so much fuss). I think they do not interested in HSA in same way, too many time passed, zero “big” application shown.

Let’s give AMD one more chance. The latest 14.6 beta Catalyst drivers do seem to have improved something with OpenCL rendering and cycles, on the compiler and GPU compatibility side at least. It’s a start.

It’s a start. :slight_smile: