CUDA on ARM

[Anwar]

Whether or not NVIDIA acquires ARM, they are adding CUDA support to ARM CPUs (https://nvidianews.nvidia.com/news/nvidia-brings-cuda-to-arm-enabling-new-path-to-exascale-supercomputing).  What does this mean for the future of V-Ray rendering on Windows 10 on ARM, or for that matter Apple Silicon?

[Justin Montoya]

You'd be better off asking over at the Chaos Group forums for that sort of question, but I have some opinions.

My guess is that it's definitely on Chaos Groups radar, given that the custom CUDA based RTX chips have incredible power in the optimized VRAY RTX GPU engine.  But I doubt it's anywhere near a priority.  ARM has a lot of promise, but it's still not a convincing solution for extreme power users who need high GHZ processors with multiple cores to get work done fast, and we don't care how much energy it uses while doing it.  ARM is more focused on efficiency, so think lightweight laptops and stupid, overly skinny devices like Apple thinks is a good idea.  

Then there's the CUDA problem.  

Apple has made it's bed on a sinking ship by continuing to cut off outside developers.  CUDA is no longer available on Apple, so I don't see why that would change once they move to ARM chips.  That means no VRay GPU or the newer RTX rendering on Apple.  However, VRAY CPU is still a fast, powerful solution that runs great on both X86 platforms and has many more features than the GPU engine, albeit, slower.  I would not expect the X86 based VRAY CPU rendering engine to continue on Apple after the ARM chip move but it's really up to Chaos to decide if it's worth supporting and how long that would take to implement.  Unless Apple backtracks on their CUDA stance, it wouldn't make a lot of sense for Chaos to offer the CPU engine on ARM Macs without CUDA, as it's only part of their VRay product and they are really pushing toward the CUDA products in the future.  CUDA is great, and just look at all the possibilities it has with ARM: https://www.nextplatform.com/2019/06/17/nvidia-makes-arm-a-peer-to-x86-and-power-for-gpu-acceleration/