GPU computing leap forward? Cloud rendering agency claims to run CUDA on non-Nvidia GPUs

one of the important variations between AMD and Nvidia is their proportion of the expert photos marketplace. Nvidia dominates this space and its earnings margins, and at the same time as AMD has had a few excessive profile wins with Apple, it hasn’t reduce deeply into Nvidia’s marketplace percentage. part of the cause Nvidia has a lock on both notebook and high overall performance computing is CUDA, its programming language for GPU compute. Now one employer, Otoy, is claiming to have broken that lock.

Otoy is the owner and developer of Octane Render, a actual-time independent rendering engine that helps three-D rendering software suites like 3ds Max, Maya, Cinema4D, and Lightwave. It’s also to be had as its own standalone software program suite. It changed into the primary impartial rendering suite to assist GPU-best rendering and a high-profile early win for Nvidia’s CUDA — which is a part of why it’s sudden to see the organization branching out to aid different architectures on this fashion.

Didn’t AMD just announce this?

There’s a few timing oddities here that I’m no longer sure the way to provide an explanation for. ultimate 12 months, AMD introduced its Boltzman Initiative. a part of that initiative is a software layer that permits AMD GPUs to execute CUDA code through the use of a compatibility layer.

right here’s how VentureBeat describes Otoy’s new compatibility layer: “In a nutshell, Otoy opposite-engineered Nvidia’s popular motive portraits processing unit (GPGPU) software, referred to as CUDA, to run on non-Nvidia hardware. which means that packages written inside the CUDA language are now not distinct to Nvidia images chips.”

in step with Otoy’s CEO, Jules Urbach, the factor of growing this CUDA translation layer is in order that the organization’s high-quit Octane Render software can run as without difficulty on AMD GPUs as their Intel opposite numbers. “We had been able to do that with out changing a line of CUDA code, and it runs on AMD chips,” Urbach stated. “you could now software as soon as and take CUDA everywhere. AMD has never absolutely been capable of provide an alternative.”

AMD’s Boltzmann Initiative would seem to provide the alternative that Urbach is referencing, and it seems to accomplish the equal goal. It’s not clear how the two applications vary from every different, even though Otoy does point out wanting to run software program on a greater variety of platforms, working structures, and technology. AMD’s Boltzmann Initiative, of course, is designed totally for AMD’s very own GPUs.

As for overall performance, Urbach states that “It runs on the alternative cards on the equal speed because it runs on Nvidia playing cards.” but once more, that’s something AMD has implied about its own Boltzmann Initiative — when we asked the organization how AMD GPUs in comparison to NV cards going for walks CUDA, the RTG division implied that unless the CUDA code have been hand-optimized for a specific CUDA structure, it should run as quick on AMD hardware as on an Nvidia counterpart GPU.

Urbach claims that the lengthy-term goal is to permit CUDA to target Vulkan, DirectX, and OpenGL (together with Android, PS4, and WebGL 3), and that Otoy wants to be able to run CUDA programs on platforms like iOS, wherein Apple’s steel is the dominant low-overhead API.

Supposedly Otoy is operating on turning Octane Render into a plugin that the UE4 engine can utilize, but Octane Render isn’t used for real-time rendering. Adapting a version of it to work within a game engine might be extremely hard. It’s on no account clear why Otoy would need to translate native CUDA into a few of the APIs that Urbach lists — no games that I’m privy to leverage OpenCL or CUDA for any form of duties, and neither AMD nor Nvidia have mentioned using both language for this cause.

We’ve reached out to Octane Render and could replace this story if we pay attention greater info. at the least, it looks as if AMD’s push to transform CUDA code into some thing that can run on extra GPUs has stuck the attention and creativeness of other companies.