LLVMpipe vs. OpenSWR Software Rendering On A 40 Core / 80 Thread Tyan Server
With OpenArena we see Mesa 17.3-dev is much faster than 17.0 for these software rasterizers, perhaps also due to LLVM 6.0 SVN being used as the back-end rather than LLVM 4.0. But between SWR and LLVMpipe there wasn't much of a difference in this game.
Xonotic was also much faster with Mesa 17.3-dev while the difference between SWR and LLVMpipe was about the same.
In the synthetic OpenGL tests via GpuTest, Mesa 17.3 built against LLVM 6.0 SVN remains superior to Mesa 17.0 with LLVM 4.0. But the SWR vs. LLVMpipe results remain relatively tight. It was a toss-up in many of the cases whether SWR was outperforming LLVMpipe or not.
I'll be back with more SWR/LLVMpipe tests soon, especially if getting the Skylake-X back-end working for SWR as well as seeing how these CPU-based renderers compare on AMD Eypc / Threadripper platforms for curiosity sake.
If you enjoyed this article consider joining Phoronix Premium to view this site ad-free, multi-page articles on a single page, and other benefits. PayPal or Stripe tips are also graciously accepted. Thanks for your support.