Hi
Can anyone tell me if there is a ‘global’ switch to enable OpenCL?
On the Flipsolver inside my AutoDOP I see one option for enabling OpenCL, but then if I dive in there are dozens of nodes, each with their OWN OpenCL switch.
Doesn't make any sense to me….

I have noticed this as well. I think OpenCL is only partially implemented. When I took the time to dive into an asset and enable all the OpenCL switches, Houdini crashed. So I would trust that the developers have already been down this path as well and the provided configuration for any given asset is probably optimized for stability and speed.

FLIP work is just slow, there is no way around that AFAIK.

FLIP was the first thing I dived into when learning Houdini and it was torture on a 2.7Ghhz 6 core AMD with only 8GB of ram. A faster computer, and more RAM can help with this. If you have an understanding of overclocking you may want to consider that. Some processes inside Houdini are still single core and having the fastest possible CPU speed can help give a performance boost. But bad overclocking can also introduce system instability so you must walk a fine line there.

Thanks. Yes hopefully OpenCL will be better implemented in 16.
I don't have a ‘slow’ system, with an i7 6900K and TitanXP GPU (hence my preference for OpenCL) but should have gone with older CPU architecture and multiple CPU's

eco_bach
Thanks. Yes hopefully OpenCL will be better implemented in 16.
I don't have a ‘slow’ system, with an i7 6900K and TitanXP GPU (hence my preference for OpenCL) but should have gone with older CPU architecture and multiple CPU's

Can you elaborate please?