From FireUser they bring us an interesting comparative which explores the performance benefits of server computers multi-socket/node configurations that AMD FirePro professional video cards offer under the application of simulation : Simulia Abaqus Direct Solver.

Recent versions of Abaqus Direct Solver is compatible with the standard API-accelerated GPU computing (GPGPU) OpenCL, and by using the API in combination with a FirePro V9800 video card,  yield significantly higher results than quad-core microprocessor (2.9x), and even greater than that of 2 quad-core microprocessors (2.2X) running in parallel in a multisocket configuration.

In addition, test where quad socket configurations(16 cores) up to 12 sockets on multiple nodes (48 cores), in this configuration, the performance increases provided by the GPU was very little, because the work is shared between multiple CPU cores, while decreasing the use of the GPU.

Link: Video: 2.9x acceleration of direct Abaqus solver using AMD FirePro + OpenCL (FireUser)

Leave a comment

Your email address will not be published. Required fields are marked *