Ah, I see you upgraded finalyDade wrote: ↑Thu Jan 02, 2020 11:22 amMy RTX 2070 Super is "only" 2 times faster than my AMD 3900X. It is a huge difference compared my old PC where the GPU was several time faster than the CPU (AMD 290X Vs i7 3930K).
AMD/Intel need to put out a competitive GPU with NVIDIA: GPUs development is becoming more and more stagnant. It is the same situation of CPUs a couple of years ago.
This thread is about CUDA support under disguise: being able to compile the GPU kernel only once is one of the requirement to be able to support CUDA. But it is a long way.
I do believe gpu's might become more and more attractive as Vram capacities grow but corona is kinda proving that you can do a lot with cpus and very fast.
compilation isn't as big of an issue for a still frame, but i did some animation tests and when you want to render frame for 30 seconds total and 7 seconds are just compilation and session init. then you have a problem.
In comparison, cycles on optix compiles once (the first time you run new blender installation - if cycles was changed in some way) then it just starts, on top of that, changing any parameter is instant so you can tweak materials while viewport is rendering in realtime. You can do that with luxcore but only if you use cpu(slow) if you use gpu/opencl blender hangs until luxcore updates everything. This situation makes small tweaks on subtle material effects like low roughness variation extremely difficult as you can't see anything until: A. cpu clears up enough B: gpu starts rendering.