Intel Still Hoping To Have oneAPI/SYCL GPU Acceleration In Blender 3.3

[phoronix]

Phoronix: Intel Still Hoping To Have oneAPI/SYCL GPU Acceleration In Blender 3.3

To complement the AMD HIP, NVIDIA CUDA, and NVIDIA OptiX acceleration in Blender 3.2, Intel engineers are still hoping to have their Intel oneAPI SYCL support ready to premiere in Blender 3.3 for Intel GPU-based Cycles acceleration with Intel integrated and discrete graphics processors...

https://www.phoronix.com/scan.php?page=news_item&px=Intel-oneAPI-Blender-3.3-On

[JacekJagosz]

If this can be used on AMD too, and HIP implementation at one point matures enough to run it on Nvidia... It could be really interesting.
I hope if we get one implementation that works for 2 GPU vendors, one other can get removed, and we don't end up with too much fragmentation.

[piotrj3]

If this can be used on AMD too, and HIP implementation at one point matures enough to run it on Nvidia... It could be really interesting.
I hope if we get one implementation that works for 2 GPU vendors, one other can get removed, and we don't end up with too much fragmentation.

Rendering takes a lot of time. So if nvidia's cuda is faster on nvidia then OneAPI, it is absolutly stupid to remove CUDA. Rendering takes tons of time, and making it shorter results in a lot of workhours being dropped.

[JacekJagosz]

Rendering takes a lot of time. So if nvidia's cuda is faster on nvidia then OneAPI, it is absolutly stupid to remove CUDA. Rendering takes tons of time, and making it shorter results in a lot of workhours being dropped.

I was honestly thinking about HIP support. SYSCL is handled through the same ROCm stack, I wonder of AMD would drop maintaining HIP Blender backend? Will they work together on SYSCL or keep pushing HIP as their own?