This is my first post to this wonderful forum. I mostly lurk here picking up things
I regularly use darktable on my debian system. Today I ran the phoronix benchmark on darktable with just cpu and then opencl. The results are surprising (to me). Darktable takes twice the time to complete the test with OpenCL as compared to just cpu. The results are here:
The GT610 only can do up to 155 gigaflop with all single precision FMA instructions, which is terribly optimistic. Realistically it’s half of that at the very best, and only in ideal conditions.
Your CPU is somewhere around 40 gigaflop. I suspect it’s closer to being fully utilized than the GPU.