The Performance-Per-Watt, Efficiency Of Intel/AMD/NVIDIA GPUs On Open-Source Drivers

Written by Michael Larabel in Graphics Cards on 5 June 2014. Page 1 of 6. 28 Comments

To complement the initial results yesterday of trying 60+ graphics cards on the open-source Linux GPU drivers -- with today being the ten year birthday of Phoronix -- here's the second round of our mass open-source graphics driver testing. While in Wednesday's article were the raw OpenGL results for the wide-range of graphics processors on the open-source Intel, Radeon, and Nouveau articles, in today's article are complementary results providing a brief look at the system power consumption, performance-per-Watt, CPU usage, and GPU thermal information when testing the hardware in the same configuration.

If you didn't already read Testing 60+ Intel/AMD/NVIDIA GPUs On Linux With Open-Source Drivers be sure to do so first to get up to speed as that article covered a plethora of in-depth OpenGL benchmarks, notes on the problematic graphics cards that reduced it to a 50-way GPU comparison, and other important information for open-source Linux users.

This second round of testing was done from the same Intel Core i7 4770K Haswell system (aside from when testing the HD Graphics 4400 of the Core i3 4130), Gigabyte Z97-HD3 motherboard, 16GB of DDR3 system memory, and 120GB Samsung 840 SSD. Ubuntu 14.04 64-bit was the base operating system while upgraded to the Linux 3.15 kernel, Mesa 10.3-devel, and other Git DDX drivers from the Oibaf PPA.

As always, the benchmarking was fully automated and monitored using the Phoronix Test Suite software. The GPU core temperatures were monitored by the Phoronix Test Suite using the exposed thermal drivers for the hardware. The system power consumption was monitored using a WattsUp USB AC power meter, which the Phoronix Test Suite is able to automatically read. The CPU usage was also monitored during testing to look at the overhead of the OpenGL drivers. This monitoring was setup for PTS by using the MONITOR=sys.power,gpu.temp,cpu.usage environment variable and by setting PERFORMANCE_PER_WATT=1 produces performance-per-Watt graphs for the data.

With all of the other details covered within the prior article, let's get straight to the results.

Related Articles