With this being complementary data to yesterday's extensive raw OpenGL tests and this round only using a subset of the exposed graphics cards, it's just a one-page article today. For the Radeon HD 6770, HD 7950, and R9 270X I ran these additional tests from the Phoronix Test Suite while our open-source benchmarking software was monitoring the AC system power consumption using a WattsUp Pro power monitor, also PTS was calculating the performance-per-Watt, and the GPU temperatures were being monitored. However, on the Catalyst 14.6 Beta used for testing, the thermal monitoring seemed to be borked so the thermal results were limited to just the open-source driver.
For the open-source driver tests the Linux 3.16 kernel was used with Mesa 10.3-devel and the updated user-space stack on Ubuntu 14.04 LTS via the Oibaf PPA. For the latest Catalyst 14.6 Beta benchmarked, we had to downgrade to Linux 3.14 for kernel compatibility with the binary blob. The system used was the same as yesterday's large comparison.
On a performance-per-Watt basis, the HD 6770 riding on the R600 Gallium3D driver is nearly on-par in performance-per-Watt with the latest Catalyst driver -- the performance in Warsow was slightly higher with the open-source driver while its power consumption was also up slightly.
The GCN graphics cards using the RadeonSI Gallium3D driver have a ways to go still in catching up in raw performance and efficiency to Catalyst.
View all of today's AMD Catalyst vs. Open-Source Radeon graphics benchmark results on OpenBenchmarking.org where you can also run your own Linux graphics tests via the Phoronix Test Suite.