So I'm planning on using the open source for my Fury X. Though I saw too packages on for arch: one called xf86-video-ati and one called xf86-video-amdgpu. I would normally assume that the amdgpu driver is the one based on the new open source driver from kernel 4.3. I've tried both of them and found them to be very similar in terms of performance from what I could tell. I also checked whether either of them was loading the amdgpu kernel module and both of them were while completely ignoring the radeon module I wanted to benchmark the two drivers but unigine benchmarks only support the proprietary drivers and I haven't found a good benchmark for the open source. So my two questions are:
1. What is the difference between the two drivers?
2. What are the best the best benchmark tools that support the open source drivers?
Here is the current status of open source drivers as an overview: http://xorg.freedesktop.org/wiki/RadeonFeature/
All new Fury and above series cards are unsupported by radeon, which is why amdgpu is loaded.
Not sure about benchmark tools, most common ones would be games but I'm pretty sure we don't have unified benchmarks in those except the in-game ones. Depending on your machine, you may wish to use ArmA III to benchmark as it would be able to fully load the graphics card(s) to capacity (however, bear in mind that it can also load your CPU to capacity due to it's quite incredibly high view distance).
If what you want to do is compare the open source drivers and the closed source drivers, you could also look into something like using Blender with OpenCL acceleration but then you run into the problem of OpenCL being kind of a work in progress kinda thing in the open source drivers :/. However, this will most likely give you the most consistent and easy to reproduce results concerning the performance of the card itself. For real world performance, you could make something like a very specific movement macro script for a game where you can rely on always being in the same starting position, and run that about ten times and take the average for each driver.
I remember a phronix article about how open source radeon drivers are getting much better.
Phronix also has benchmarking software but they tend to be huge and I have dsl.
Well yes I knew that the Fury cards aren't supported by the radeon driver but it confused me that the ati driver worked at all. As for blender I'll try it out though I don't know much about how it and for OpenCL I'll try agisoft photoscan because I still have it installed from a previous benchmark
Just be warned OP the drivers for the FURY with AMDGPU are lack luster at best a r9 280 beats it in benchmarks.
http://www.phoronix.com/scan.php?page=article&item=amd-fury-opensource&num=1
AMDGPU is new and the other one is old. Thats about it. If you didn't have a fury and you had a 250X like I did (emphasis on did) then you would use ati. You can use ATi if you want but you should use AMDGPU for future use. Look up the features and future intent on the driver.
The open source drivers in kernel 4.4 have improved my performance (at least in borderlands 2 as thats all I've been playing lately) by quite a bit . It's still only about half the performance you should expect but still much better than before. I'm gonna see how much the driver improves in kernel 4.5 but right now (I'll test this once I get a proper benchmark) the open source driver seems to be slightly faster than the (notably horrible) catalyst driver
EDIT: Did I say half? I just played some more games and I realized that it's getting way less than half the performance it should be getting