Can simulation use multiple gpu to speed up?

James

Can simulation use multiple gpu to speed up?

Richard Haseltine

No, as far as I know.

skyeshots

Daz's dforce simulation leans heavy on the CPU (vs IRAY which takes full advantage of multiple GPUs) and it is not parrallel threaded. Check out CPU Perfromance in Task Manager during simulation. There is a request in to enable more cores/threads for the simulations. In the current state though,max simulation speed will be coupled to your max single-core CPU speed.

crosswind

skyeshots said:

Daz's dforce simulation leans heavy on the CPU (vs IRAY which takes full advantage of multiple GPUs) and it is not parrallel threaded. Check out CPU Perfromance in Task Manager during simulation. There is a request in to enable more cores/threads for the simulations. In the current state though,max simulation speed will be coupled to your max single-core CPU speed.

On the contrary, dForce simulation per se mainly utilizes GPU. GPU load can be 30% ~ 99%+ depending on the numbers / complexity of the dForce items in the scene while there'll be 3% ~ 5% increase of CPU load because of multi-threaded co-operation in between GPU and CPU with OpenCL, as well as the utilization from DS...

That's what I've been seeing for years ~~

Richard Haseltine

crosswind said:

skyeshots said:

Daz's dforce simulation leans heavy on the CPU (vs IRAY which takes full advantage of multiple GPUs) and it is not parrallel threaded. Check out CPU Perfromance in Task Manager during simulation. There is a request in to enable more cores/threads for the simulations. In the current state though,max simulation speed will be coupled to your max single-core CPU speed.

On the contrary, dForce simulation per se mainly utilizes GPU. GPU load can be 30% ~ 99%+ depending on the numbers / complexity of the dForce items in the scene while there'll be 3% ~ 5% increase of CPU load because of multi-threaded co-operation in between GPU and CPU with OpenCL, as well as the utilization from DS...

That's what I've been seeing for years ~~

The single thread will be the management of the GPU simulation processes. Remember that dForce is using OpenCL, not CUDA, so it may not show on the same graphs in task manager if that is what you are using to check activity.

crosswind

Richard Haseltine said:

crosswind said:

skyeshots said:

Daz's dforce simulation leans heavy on the CPU (vs IRAY which takes full advantage of multiple GPUs) and it is not parrallel threaded. Check out CPU Perfromance in Task Manager during simulation. There is a request in to enable more cores/threads for the simulations. In the current state though,max simulation speed will be coupled to your max single-core CPU speed.

On the contrary, dForce simulation per se mainly utilizes GPU. GPU load can be 30% ~ 99%+ depending on the numbers / complexity of the dForce items in the scene while there'll be 3% ~ 5% increase of CPU load because of multi-threaded co-operation in between GPU and CPU with OpenCL, as well as the utilization from DS...

That's what I've been seeing for years ~~

The single thread will be the management of the GPU simulation processes. Remember that dForce is using OpenCL, not CUDA, so it may not show on the same graphs in task manager if that is what you are using to check activity.

I personally always check GPU Load with GPU-Z rather than Task Manager that may lead to unaccountable info.

TheMysteryIsThePoint

OpenCL is related to OpenGL, both are very old and neither supports multiple GPU devices. But unless you've got an EPYC processor, a midling GPU will still be faster and much more power efficient than even a high end CPU.