Radeon RX 480: RadeonSI Gallium3D vs. AMDGPU-PRO - Interesting CPU/Power Difference
Written by Michael Larabel in Radeon on 2 July 2016 at 08:01 AM EDT. 83 Comments
With the 12-page AMD Radeon RX 480 review under Ubuntu Linux, tests were carried out both with the AMDGPU-PRO hybrid driver and the pure open-source driver via the Linux 4.7 kernel and RadeonSI Gallium3D Mesa 12.1-dev. In this article are more tests of just the RX 480 when comparing the two Linux driver options for this first AMD Polaris card to make it to market. The CPU usage and power consumption difference are particularly interesting if you have been debating between these two drivers.

Besides just comparing the raw OpenGL performance between AMDGPU-PRO and RadeonSI Gallium3D for Polaris, the CPU usage and system power consumption were also measured in a fully-automated manner via the testing process through the Phoronix Test Suite.

Above is the system details for this Radeon RX 480 testing (yes, under AMDGPU-PRO it's reported as a HD 7700 right now). Below are the raw data metrics:

Particularly in the poorly ported games, the AMDGPU-PRO driver tends to do better than the fully open-source driver stack with Mesa being strict about OpenGL conformance. But in plenty of tests, the brand new Radeon RX 480 saw either the same level of performance as AMDGPU-PRO 16.30 or actually greater! By far the best discrete GPU launch we've ever seen when it comes strictly to open-source driver support!

Now some perf-per-Watt numbers between the two drivers for the RX 480:

The performance-per-Watt is comparable between these two AMD Linux drivers.

Very interesting though is that the CPU usage was constantly lower with the RadeonSI Gallium3D driver than under AMDGPU-PRO... About 5% less CPU utilization overall during all of these OpenGL benchmarks executed. Is Gallium3D/Mesa that much more efficient than AMD's closed-source OpenGL driver or is that binary blob just so old and dilapidated from handling so many workarounds, etc?

The overall system power draw also ended up being much lower with the fully open-source driver stack compared to AMDGPU-PRO. Is this due to power savings from the Intel Xeon E3 v5 Skylake CPU being in a lower power state due to the less CPU utilization? Or is the open-source driver doing something different with regards to PowerPlay / GPU power-savings? Unfortunately, I don't have any good answer right now. I did send these early Radeon RX 480 results to AMD developers earlier this week, but unfortunately they have yet to comment on these side-by-side driver numbers.
Related News
About The Author
Author picture

Michael Larabel is the principal author of Phoronix.com and founded the site in 2004 with a focus on enriching the Linux hardware experience. Michael has written more than 20,000 articles covering the state of Linux hardware support, Linux performance, graphics drivers, and other topics. Michael is also the lead developer of the Phoronix Test Suite, Phoromatic, and OpenBenchmarking.org automated benchmarking software. He can be followed via Twitter or contacted via MichaelLarabel.com.

Popular News This Week