Radeon RX 560/570/580 vs. GeForce GTX 1060/1650/1660 Linux Gaming Performance
If you are looking to soon upgrade your graphics card for Linux gaming -- especially with the increasing number of titles running well under Steam Play -- but only have a budget of around $200 USD for the graphics card, this comparison is for you. In this article we're looking at the AMD Radeon RX 560 / RX 570 / RX 580 against the NVIDIA GeForce GTX 1060 / GTX 1650 / GTX 1660 graphics cards. Not only are we looking at the OpenGL/Vulkan Linux gaming performance both for native titles and Steam Play but also the GPU power consumption and performance-per-dollar metrics to help guide your next budget GPU purchasing decision.
Given the interest from our recent GeForce GTX 1650 Linux review, this article offers more benchmarks of the current low-end AMD/NVIDIA offerings. In that earlier article there also wasn't any RX 570 reference point as I never had a review sample of that Polaris model, but given a recent online sale, I did pick up a Radeon RX 570 for this round (and future) Linux testing. So now with today's results is a complete look at the RX 560/570/580 Polaris cards against the NVIDIA Pascal/Turing GTX 1060/1650/1660 cards for a look at the roughly $200 USD and less spectrum.
Dozens of benchmarks were run both of native Linux game ports as well as benchmark-friendly Steam Play titles. The GPU power consumption was also monitored using the driver/hardware's exposed sensors (as opposed to the overall AC system power consumption that we usually do, since all my WattsUp Pro meters were busy in other systems) as well as looking at the GPU temperature for the cards tested. There are also performance-per-dollar metrics using the current US retail pricing from NewEgg.
The AMD Radeon graphics cards were tested using the Linux 5.0 kernel and Mesa 19.1-devel via the Padoka PPA. The NVIDIA cards were tested using the NVIDIA 430.09 driver release. All tests were done from Ubuntu 19.04. All of these benchmarks were driven via the Phoronix Test Suite.