Hi,
do you know if Simulation CFD can take advantage of a TESLA gpu and parallel computing?
Regards
Solved! Go to Solution.
Solved by apolo_vanderberg. Go to Solution.
SimCFD does not do GPU computing
We do parallel computing by leveraging multiple CPU cores. This can even span multiple machines using HPC [where all machines are on an Infiniband (10Gb/s+) network]
Is this on the radar screen?
Any discussion of ..
CUDA
OpenCL
Xeon Phi
Ansys and OpenFOAM have GPU/Accelerator code with mixed results. My understanding is that the only thing that makes Xeon Phi cost efficient on Ansys is that Xeon Phi has a licensing cost advantage when using only 2 cores.
Realistically, nobody will use only 2 cores unless very limited by license budget. Obviously, Autodesk's licensing (today) is much more liberal than Ansys (but OpenFOAM is open source so 'nuff said there).
OpenFOAM has seen some great examples of CUDA and XeonPhi acceleration in particular but to date these are for a narrow set of problems.
Sim360 should give Autodesk a very compelling case for accelerating code via OpenCL/CUDA/Xeon Phi but obviously this is a coding challenge.