E L I T E
C H A N N E L PA R T N E R
simulation@rand.com | RandSIM.com | 888.483.0674
Test 2 – Internal Flow in Mixing Tank
This test doubles the cell count and adds a moving reference frame, which the GPU solver can handle today.
The GPU was equivalent to about 32 CPU cores in this test.
In the graph on the right, we see that the trends were relatively similar to test case number one. In this case,
the GPU single precision was just as fast as the CPU 32 single precision. We believe this is because there are
so many cores on a GPU for such a small grid that it needs more cells to see the upscale in speed due to the
thousands of cores of the GPU. Where we saw a GPU being the equivalent of about 25 cores in the previous
scenario, here we saw about 32 cores. Larger grids of tests three and four will see even faster speeds.
The graph on the left shows the mixing tank torque per 100 iterations. The green curve shows that the CPU 32
DP Coupled solver was very slow per iteration but required a lot fewer iterations. Regardless of solver settings,
we arrived at the same conclusion; the GPU solver is about as fast as 32 CPU cores.
Does Ansys Fluent with a Native GPU Solver Live Up to the Hype?
