Did anyone try to render with the latest 3.0.0 Beta? I am only getting one thread and I think is just the GPU but when I switch to CPU I get the same problem. Previous 3.0.0 Alpha version I have installed when having both CPU+GPU I only get 32 threads and I don’t think the GPU is working. I guess I am missing something.
I now counted the threads in the 2.93.5 official release and again is just 32 threads. Shouldn’t it be 33 in total counting my GPU if I have them both active and GPU selected in the Render device option?
So it seems I read somewhere else multi device rendering isn’t yet implemented for Cycle X? is the 3.0.0 Beta using Cycle X? Anyone? What Cycles version is 2.93.5 using?
Renzatic
(Professor Emeritus Billy H. Wafflesmith XIV Esq.)
3
The 3.0 beta is using Cycles X, yes. 2.93 is still classic cycles.
Did you use both of them with GPU+CPU. Any idea why I don’t get 32 tiles from CPU + 1 from the GPU? Well 3.0.0 I only get 1 single tile even using CPU only. I think it is quite confusing now having both options available and changing anything in the Rendering options in the properties section does nothing. Not even in Preferences/System/CUDA
Renzatic
(Professor Emeritus Billy H. Wafflesmith XIV Esq.)
5
You do know that Cycles X doesn’t (currently) support bucket rendering, right? It behaves more like the viewport renderer, resolving the full image over time, rather than doing it tile by tile.
If you want to know if it’s taking full advantage of your CPU and GPU, render out your scene, see how long it takes, and compare the results to classic Cycles.
I heard something but why does it still have a tile option? So everything gets to work together in a single tile and depending on the size it could work better for certain devices? What about 2.93.5 that doesn’t use Cycle X and I still get 32 tiles only when using GPU+CPU. Selecting GPU in properties and having both enabled in preferences.
The difference in rendering time between 2.93.5 and 3.0.0 is massive to be honest.
Renzatic
(Professor Emeritus Billy H. Wafflesmith XIV Esq.)
7
For the first question, I believe they’re intending on reimplementing bucket rendering at some point, so that’s probably why they still have the tile options available.
For the second question, I have no idea. I’ve only used hybrid rendering on a couple of occasions, so it’s not something I’m well versed on.
Tile option is used in rendering hi-res images. The tile size is set to 2048x2048 I believe, but I didn’t check this recently. And all devices are rendering one big tile together.
This is a workaround to maintain low memory footprint during rendering images with very big resolution (like 8k or 16k, etc.) that otherwise would require huge amounts of RAM/VRAM to progressively render it entirely. This option kicks in automatically if render resolution is set to bigger sizes, so no user input is needed.
I think that’s the key point here. Old (pre 3.0) Cycles would allocate one device per tile but new Cycles (formerly called Cycles-X) now allocates all rendering device resources to one tile at a time.
The only option you have is whether you have Auto Tile on or off. If it’s off, then you get ONE big tile for the whole image and if that’s too much data for your memory/GPU then bad luck. So the default is for Auto Tile to be ON with the size set to 2048. If you render an image with neither dimension larger than 2048 pixels, then again you’ll just get one tile for the whole image.
If a dimension of your image is more than 2048, then you’ll get two or more tiles starting with the lower-left corner of the image, proceeding from left to right and then bottom to top (I wish there were still a way to get it to start centered in the middle). Each tile will be rendered (but not denoised until the entire image is complete I think) using all hardware resources configured and available and no work will be done on the second tile until the first is complete.
Each tile is always rendered progressively, with or without adaptive sampling. Cycles now defaults to Adaptive Sampling at 0.01, 4096 max samples, and OIDN denoising of the final image. I like to do 0.005 and 256 max samples for quick renders.