A Look At The LLVMpipe OpenGL Performance On Mesa 19.0 With A 64C/128T Server

Given the proposed Libre RISC-V SoC that could function as a Vulkan accelerator by running the Kazan Vulkan implementation on it, I decided to have a fresh look at how the LLVMpipe performance is for running OpenGL on the CPU. Here are those tests done on a dual socket AMD EPYC server.

The proposed Libre RISC-V SoC at this point would be quad-core and clocked at least 800MHz. Their initial performance target is aiming for 25 FPS at 720p and capable of 5~6 GFLOPs. Quite low by today's standards and more so for a SoC that is likely more than a year or two out.

As Kazan is still maturing and not yet running Vulkan games, I wanted to see how the performance was for the mature LLVMpipe driver that provides Mesa OpenGL by running on the CPU and optimized via LLVM. For providing very high-end performance potential, I used the Dell PowerEdge server running Ubuntu 18.10 with Mesa 19.0-devel so LLVMpipe was running off the two AMD EPYC 7601 processors.

LLVMpipe presently exposes OpenGL 3.3 support.

The ET: Legacy open-source game derived from Wolfenstein Enemy Territory from the early 2000's could run at just 15 FPS on the CPU at 1280 x 1024...

Tesseract with LLVMpipe was only churning out 11 FPS at 1280 x 1024.



And the other OpenGL tests on the very low-end even with this very high-end x86_64 server setup. More data can be found via this OpenBenchmarking.org result file. Long story short, even with a high-end x86_64 CPU, the LLVMpipe performance is very slow. It will be interesting to see how well Kazan can mature for running Vulkan on the CPU in 2019.