Hi there!
Is there any way to utilize dual core cpu + gpu in one task. And if there is, is there any considerable speedup of utilizing dual core cpu + single gpu instead of single core cpu + single gpu in calculation on Fenics?
FEniCS doesn’t currently support GPUs, so I don’t believe you will experience any speedup by considering GPUs.
However it has been fundamentally designed for parallel processing using MPI.
The actual speedup you can get is highly dependent on a combination of factors, including the problem in hand, the solver, and preconditioner etc.
3 Likes
I’m dealing with problems in different area of physics: plasma, cosmology, optics and so on. For this tasks I suppose to utilize fenics as “Swiss knife”. At a moment I’m trying to find out configuration of appropriate workstation to work with fenics. So, gpu is not an important element. And now i have two ways. The first way is to choose workstation based on dual old low speed xeon (e5 2680v3 12 cores 2.5 GHz freq per cpu). The second way is to use single but modern and fast Ryzen 3 3900x (12 cores 3.8 HGz freq). I would very appreciate for advise which way is to preferable. Or may be these ways give a similar performance (about tens percents difference not critical for me)? Thanks in advance!