OpenCL Benchmarks
These benchmarks are like the ones we ran under DirectX in that we tested both workstation and consumer cards (minus the defunct Quadro 5000, of course). The results themselves are fairly self-explanatory.
AMD's workstation and consumer cards do very well in our bitmining test, showing off the benefits of their GCN architecture.
We used Luxmark 2.0 since it contains a somewhat simpler scene. The slower cards’ performance barely even registers with more complex rendering tasks. As we add complexity, the FirePro W-series cards pull ahead even more.
PostFX uses both OpenGL and OpenCL to create a challenging benchmark. This used to be Nvidia’s strong suit, but the introduction of AMD’s GCN architecture reverses that trend.
AMD’s FirePro W-series cards don’t fare quite as well solving the NQueens problem. The differences are smaller than they used to be, but Fermi still wins.
