I am looking to get pointed in the right direction for documentation and support. I have been asked to look into K2 Grid Profiling. I have been told that it is possible to set up a profile where the Petrel Application can utilize the GPU in computational processing. Is this correct and where can I find information on how to do this?
If you use Pass Thru with the NV GPU on the K2 card to the Petrel app, they you can use the GPU for both graphics and CUDA (computational processing).
vGPU (Sharing the GPU among a number of users) does not support CUDA (computational Processing). We have CUDA support with vGPU (Sharing) on the roadmap so it will be coming down the road in a future update.
Am I to assume from your statement above that the Petrel Application will control the use of the GPU for both Graphics and Computational as it see fit?
Sorry I am not too familiar with CUDA yet.
Yes. Petrel makes heavy use of shaders to perform “computations”. It will use the GPU as it needs to because it is programmed to take advantage of the GPU. It can be used with our vGPU sharing and will run fine. Petrel does not explicitly use CUDA (I had to research).
Just as FYI, CUDA is a programming language for using the GPU for computation. If an application specifically needs CUDA for computation on the GPU, the GPU needs to be in Pass Thru to take advantage of the computational capabilities as well as the graphics capabilities in the Virtual Machine.