AMD Working On CUDA Source Translation Support To Execute On FirePro GPUs

[phoronix]

Phoronix: AMD Working On CUDA Source Translation Support To Execute On FirePro GPUs

Early this morning I wrote a brief article about AMD working on an LLVM-based Heterogeneous Compute Compiler and since then more details have come to light...

http://www.phoronix.com/scan.php?page=news_item&px=More-AMD-HCC-Details

[-MacNuke-]

Well... the AMD OpenCL compiler is not capable of compiling and/or running the valid Blender Cycles OpenCL Mega-Kernel code and now they want to compile and run CUDA code?

[schmidtbag]

Interesting that only FirePro is getting this. But, aside from drivers and ECC memory, there is otherwise no difference between Firepro and Radeon, so I guess AMD needs something else to give FP some edge. If AMD can get CUDA to work on Radeon though, that will give their IGPs some HUGE leverage. Using their IGP for physx could be really appealing to gamers.

[DarkFoss]

I posted the link I saw at Rage3d in the other thread will post it here as well breakdown by Anandtech .

[L_A_G]

So they're now trying to widen the appeal of their FirePro cards by having them run CUDA programs? Seeing how common CUDA has become in comparison to other GPGPU platforms like OpenCL and OpenACC, it does make sense. However this is giving Nvidia a further upper hand in the GPGPU market when they should instead focus on open platforms like OpenCL and OpenACC with equal playing fields.

[schmidtbag]

However this is giving Nvidia a further upper hand in the GPGPU market when they should instead focus on open platforms like OpenCL and OpenACC with equal playing fields.

Actually, I'd say this would be giving AMD the upper hand, because now people have a choice when it comes to CUDA. On top of that, nvidia has largely been the one funding CUDA, so this is a double-loss for them. AMD may be attracting more users to CUDA, but they could be driving sales away from nvidia.

CUDA has been getting less attention because openCL is supported on more platforms. But I get the impression CUDA is slighly more polished and a better performer, so if someone wants to use CUDA but don't want to pay the extra premium for nvidia, then AMD is a good choice.

[wodencafe]

Actually, I'd say this would be giving AMD the upper hand, because now people have a choice when it comes to CUDA. On top of that, nvidia has largely been the one funding CUDA, so this is a double-loss for them. AMD may be attracting more users to CUDA, but they could be driving sales away from nvidia.

CUDA has been getting less attention because openCL is supported on more platforms. But I get the impression CUDA is slighly more polished and a better performer, so if someone wants to use CUDA but don't want to pay the extra premium for nvidia, then AMD is a good choice.

It's a double edged sword. One the one hand, it eliminates the NVidia vendor lock-in for running CUDA, while allowing AMD to piggyback on NVidia's development of CUDA.

But it also implies AMD conceded that CUDA was a better technology than OpenCL, which raises questions about AMD's future plans for both of these technologies.

[profoundWHALE]

From what I was reading up on, the translation works both ways, so making it easier to port one to the other and back again will make AMD's cards more supported in general, which is really really good for AMD.



Added picture for the ones who don't want to click links.

[Marc Driftmeyer]

Interesting that only FirePro is getting this. But, aside from drivers and ECC memory, there is otherwise no difference between Firepro and Radeon, so I guess AMD needs something else to give FP some edge. If AMD can get CUDA to work on Radeon though, that will give their IGPs some HUGE leverage. Using their IGP for physx could be really appealing to gamers.

Same can be said for Nvidia's Professional lines. AMD is building a bridge for professional vendors using professional lines to move from CUDA to OpenCL. Gamers aren't part of that.

[Marc Driftmeyer]

It's a double edged sword. One the one hand, it eliminates the NVidia vendor lock-in for running CUDA, while allowing AMD to piggyback on NVidia's development of CUDA.

But it also implies AMD conceded that CUDA was a better technology than OpenCL, which raises questions about AMD's future plans for both of these technologies.

No it's not. It's building the bridge that Nvidia doesn't have. Nvidia has no professional OpenCL tools to touch AMDs. This will increase AMD sales and once in the tiering with OpenCL will show vendors they have a much better future strategy.