how did you get it to run on the gpu? I've found it to be pretty slow! Honestly I can't find any evidence whatsoever to support the claim that this is fast and runs on the gpu.
It runs on the GPU by default. It's entirely based on openCL. Second paragraph at the bottom: www.sidefx.com/docs/houdini//nodes/sop/mpmsolver.html "The solver is fully written in OpenCL to take advantage of the powerful GPUs available in most workstations."
@@PaulEsteves28 hmm I see. With the tests I've done though the solver seems to rely more heavily on system memory and uses very little vram (even on higher particle counts only about 2-3gb of vram). Not quite on the same level in terms of performance as something like dyverso in Realflow. I thought with your statement that "the solver runs completely on the gpu" that maybe I had missed a setting or two. Still an impressive system. I just wish there was a cuda option.
@@theshadow6273on GPUs you want to minimise your vram usage since memory transactions are what are usually the limiting factor for GPU calculations. Not saying that small vram usage necessarily means that the OpenCL implementation is good, but it could be an indication that it's been well optimised.
That jiggle from the burger is definitely going to be used differently by someone 😂😂
Bruh 😂
@@Shsj272 No idea what you could be implying.......
we need more info about meshing ;)
what is the creature that you use in the tests it looks cute
it's a test geo that comes with houdini called squab. He's cool! :)
How do you mesh the jelly geo?
All my meshing is either VDB workflow (like you'd do with fluids) or I used point deform on the original mesh.
looks like SPH or SPB Solver in Realflow.
I don't know specifics, but, it could be?
how did you get it to run on the gpu? I've found it to be pretty slow! Honestly I can't find any evidence whatsoever to support the claim that this is fast and runs on the gpu.
It runs on the GPU by default. It's entirely based on openCL. Second paragraph at the bottom: www.sidefx.com/docs/houdini//nodes/sop/mpmsolver.html
"The solver is fully written in OpenCL to take advantage of the powerful GPUs available in most workstations."
@@PaulEsteves28 hmm I see. With the tests I've done though the solver seems to rely more heavily on system memory and uses very little vram (even on higher particle counts only about 2-3gb of vram). Not quite on the same level in terms of performance as something like dyverso in Realflow.
I thought with your statement that "the solver runs completely on the gpu" that maybe I had missed a setting or two.
Still an impressive system. I just wish there was a cuda option.
@@theshadow6273on GPUs you want to minimise your vram usage since memory transactions are what are usually the limiting factor for GPU calculations. Not saying that small vram usage necessarily means that the OpenCL implementation is good, but it could be an indication that it's been well optimised.
@@theshadow6273 What kind of gpu do you have? If you have an amd gpu, it won't run very fast