Attached find the screenshot file of CPU GPU utilisation during export from Darktable and capture one 12. Capture one is much more faster in exporting and better at utilising the GPU. Is there a scope for fine tuning OpenCL performance? Attached find my darktablerc filedarktablerc.zip (6.5 KB)
There is several stuff on tweaking opencl at https://darktable.gitlab.io/doc/en/darktable_and_opencl.html
I’m not sure it’s relevant to compare the raw CPU perfs from 2 different softwares (different algos, different implementations, different budget). Not all modules in dt are OpenCL-enabled, so there are some memory flip-flops between RAM and vRAM that could explain the difference. Also OpenCL support in dt could probably be updated for modern GPU.
my purpose was to get the best performance from Darktable will try to tweak my setting as per MartinSt suggestions .