LLVMpipe Still Is Slow At Running OpenGL On The CPU

Two months ago we published our initial benchmarks of LLVMpipe, the Gallium3D driver that accelerated commands on the CPU rather than any GPU and unlike other Linux software rasterizers is much faster due to leveraging LLVM (the Low-Level Virtual Machine) on the back-end. Since then we have published new ATI Gallium3D driver benchmarks and yesterday put out Nouveau Gallium3D driver benchmarks, so today we are providing updated LLVMpipe driver results to show how well Gallium3D's LLVMpipe driver can accelerate your OpenGL games with a modern processor.

This round of LLVMpipe testing was done on the same system and software stack as yesterday's Nouveau Gallium3D results. This system consisted of an Intel Core i7 920 CPU (quad-core plus Hyper Threading) clocked at 3.60GHz, an ASRock X58 SuperComputer motherboard, 3GB of system memory, and a 320GB Seagate ST3320620AS SATA HDD. The software stack was an Ubuntu 10.10 daily snapshot with the Linux 2.6.35-5-generic (x86_64) kernel, GNOME 2.30.2 desktop, X.Org Server 1.8.2 RC2, Mesa 7.9-devel, LLVM 2.7, and an EXT4 file-system. For comparing the LLVMpipe numbers we have the Nouveau test results again for the GeForce 8500GT and GeForce 8800GT and we also tested a GeForce 8400GS atop Gallium 0.4 too as a lower-end NVIDIA graphics card for more comparable numbers to LLVMpipe.

The tests we ran with LLVMpipe were OpenArena, World of Padman, Urban Terror, and Warsow. All of this testing was done through the Phoronix Test Suite. For those interested in knowing about the CPU usage when using LLVMpipe, we have such numbers in our last LLVMpipe benchmarking article. That article happens to use the same Core i7 920 system for its LLVMpipe testing, albeit with an updated software stack. Unfortunately, LLVMpipe does not yet support compositing window managers so as such Compiz was disabled during the entire testing process.