Radeon HD 7950 vs. GeForce GTX 680 On Linux
Should you be looking towards purchasing a new high-end graphics card from AMD or NVIDIA, here are some updated benchmarks of the AMD Radeon HD 7950 and NVIDIA GeForce GTX 680 when both graphics cards were tested under Ubuntu 12.04 with their proprietary Linux graphics drivers.
The open-source drivers for either of these latest generation GPUs are not ready for end-users:
- The Radeon HD 7000 "Southern Islands" series has its DRM support in place plus an emerging "RadeonSI" Gallium3D driver, but the Gallium3D driver isn't yet ready for proper use. There also is not any 2D/DDX support in place either as they are still trying to decide how to do it for this latest generation hardware. The Radeon HD 7000 series is now about six months old, and sure there were invasive architectural changes with GCN, but the support is still far from being ready, and that isn't an excuse.
- The reverse-engineered GeForce 600 "Kepler" series is actually in better shape than the latest-generation hardware. At launch, there was some reverse-engineered support available, even with Gallium3D support. This code is now mainline but the initial limitation that makes it not too useful is that for accelerated support you must first extract the firmware by running the proprietary driver. Hopefully for Linux 3.6 this will be fixed so that there can then be proper "out of the box" support without having to jump through some hurdles. There is also not any re-clocking support yet for Kepler so the performance is very poor.
Simply put, for the latest AMD and NVIDIA GPUs the current open-source support still leaves a lot to be desired. Using the proprietary drivers is likely the only real option for a majority of Linux desktop end-users for the next few months.
This benchmarking was done from the Intel Core i7 3770K "Ivy Bridge" test system. The NVIDIA GeForce GTX 680 was running at its stock speeds (there still is no overclocking support for Kepler/Fermi on Linux) while the Radeon HD 7950 was tested at its factory-overclocked speeds of 900/1375MHz and then again at the AMD reference clock speeds of 800/1250MHz.
The GPU temperature and system power consumption (the overall AC power use as measured by a USB-based WattsUp meter) was also recorded for each run via the Phoronix Test Suite. Aside from OpenGL, there are 2D and OpenCL benchmarks in this comparison too.
Latest Articles & Reviews
Latest Linux News
Most Viewed News This Week