Radeon HD 5000 Series Gallium3D Performance vs. Catalyst

Earlier this week I delivered results showing the AMD Radeon HD 6000 open-source driver becoming more competitive against the proprietary AMD Catalyst driver. The Gallium3D driver is still far from reigning supreme in most aspects that concern desktop Linux users, but much progress has been made in recent months. For those HD 5000 series graphics card owners, here are similar open-source driver vs. AMD Catalyst OpenGL Linux performance tests.

The results in this article show roughly the same commendable performance of the Radeon HD 5000 series as the newer Radeon HD 6000 series graphics cards as compared to the Catalyst driver. The latest open-source driver with using Mesa 9.3-devel Git master and the Linux 3.11 kernel with Radeon Dynamic Power Management enabled generally yields 50% or more the performance of AMD's binary blob. For some workloads, the open-source OpenGL performance can even be around 70~80% that of Catalyst.

In this article, results from the Radeon HD 5770 and Radeon HD 5830 on both graphics cards are being compared between the open and closed-source drivers. Unfortunately, this comparison isn't as extensive as the Radeon HD 6000 series comparison due to not having as many HD 5000 series GPUs on hand for testing. Additionally, when trying to test one of the GPUs (a Radeon HD 5750), the system would lockup under any 3D workloads with the open-source driver stack.

All of this Linux OpenGL benchmarking was facilitated in a fully automated and repeatable manner using the open-source Phoronix Test Suite platform. As with the other tests, the latest kernel and Mesa Git were used while the Radeon DRM driver had DPM enabled. Swap buffers was disabled as always. The rest of the platform was left stock, including not forcing the R600 SB back-end, but separate Phoronix articles will deliver the shader optimization back-end benchmarks especially now that it looks like it could soon be enabled by default.