… it seems to be the bucket size is fixed, for example I tried bucket size 10, but it’s still quite large.
(GPU mode)
Bucket size is automatically handled by the GPU engine based on the image frame. This affects buckets assigned not only to GPUs but to CPU when used as a CUDA device as well.
Due to GPU utilization during the rendering process, Bucket size cannot be changed and is automatically determined for your frame. However, the Bucket size is not dynamic (the size of the buckets do not change during rendering).
We will consider hiding the UI option while GPU engine toggle is enabled.
Hiding the UI could be useful. I tried to set a smaller bucket size, because it was looking so big. Maybe it could be possible to get back the dynamic bucket size at the end of the rendering like known from the CPU mode. This was a very nice feature.
It could be great if the user set a region render that the automatic bucket size is adapted to the region size. So, I render a small region with slow calculated metal reflections and the last few big buckets are very slow. I would be great if all GPU buckets could be used until the end.
Hi Micha,
The dynamic bucket size is only available in CPU rendering mode.
On the other hand, the bucket size itself can be changed but not all sizes are allowed.
There are some internal limits (there is a lower limit for example) that ensure that your GPU utilization is reasonable.
I’ll sync with the GPU team and possibly extend the parameter tooltip information.
I’m not surre you right understand me - the buckets are so large that for a region render often only a part of the available GPU power is used. A smaller bucket size would help a lot.
Thank you for mentioning this. Will look into it.
Kind regards,
Peter
I found a workaround - if needed for small regions I switch to progressive rendering and so I get the full power of all cards and avoid the large buckets.