AMD has this reputation for “FineWine” as in some of their hardware aging like Fine Wine, with their drivers getting the most out of their GPUs over time. The GCN lineup of GPUs like the HD 7000 Series, The R9 200/R9 300 Series, The RX 400/RX 500 series are some examples of AMD’s “FineWine.”
Today I’d like to introduce you all to a different flavor of “FineWine” and that is the power of Open Source Community Drivers. This happens to be the Mesa RADV Radeon Drivers in particular on Linux. To keep things fairly plain and simple, through the combination of continuous hard work from the community, and with constant improvements to the likes of VKD3D-Proton and DXVK, Linux has reached to the point where it is overtaking the performance of older Radeon cards on Windows, at least in Raster only scenarios. The aforementioned benchmark comparison is one such example.
It looks to me that AMD’s older GPUs on Linux, with the latest mesa-git and Proton Experimental “Bleeding-Edge” can outperform their Windows counterparts. This applies to RDNA1, and older GPUs, such as Radeon VII, VEGA and Polaris.
Creating this comparison and sharing this in a video (which I can’t link to, probably due to my account being too new) has also brought a lot of hostility aimed at me, with people making all sorts of assumptions about my PC, my ability to install driver, etc. which is all fine and normal, as anything and everything on should be taken with some grain of salt. However, I can assure you that:
- These tests were done with as much apples to apples scenario as possible, as in same CPU clocks, same GPU clocks, same board power and GPU voltage across both operating systems, the very same in game Graphics Settings.
- The Windows installation and the GPU driver installation on Windows is absolutely fine. AMD’s “Cyberpunk 2077: Phantom Liberty” supported driver on Windows, Adrenaline 23.9.3 actually has performance regression compared to its slightly earlier revision, which is the 23.9.2, which is a bit of a shame really as the latest version of their Windows drivers were used to make this comparison which doesn’t represent the best case scenario for AMD’s Adrenaline Drivers. However, with the regression accounted for, the difference doesn’t change that much.
If you still have any questions, any doubts, I’ll be more than happy to address them. This isn’t to say AMD’s Windows Drivers are bad, they are the best that they can be, but to raise awareness that it is possible to level up your hardware thanks to the continuous efforts of the open source community.
Cheers!
