Discussion in 'Videocards - AMD Radeon Drivers Section' started by Krteq, Feb 11, 2021.
Zero issues here with 21.2.2. Either video playback or gaming.
Which card are you using if I may ask? I just see Radeon Technologies in the GPU section. All is well here so I'm curious.
Is it just me and my memory playing tricks on me or gta 5 recently got some boost in performance? Today i launched the game to check few things and i feel like they improved performance a lot, fps is much higher than it used to be at the same settings. Can anyone confirm or deny? Anyone with similar or opposite experience? I'm talking about single player as multiplayer in gta has never been my cup of tea.
Also, as some people here are using dxvk, i decided to give it a go. I'm kinda surprised with results. In some games it gives some performance improvement, some games have worse performance with dxvk, other for some reason completely ignore dxvk's .dll files and many games just crash on launch when dxvk is applied to them. Definitely will keep it with games in which dxvk works and gives boost.
Can someone tell me what is expected behaviour with vsync on in Quake RTX game? I have 75Hz monitor. When I turn vsync on in the game - I get 25 or 38 fps constantly (it switches for some time to 25, for some time to 38 during the game). If I turn vsync off - then the game runs so fluently with 80~90 fps (1920x1080@75 mode).
(Of course, I lower the voltage - otherwise the game hangs on my system)
The GTA 5 part is placebo. For RDNA1 users, recent drivers fixed Reflection MSAA. Now that you're using some 3% more GPU power to properly render that (as opposed to less to render due to broken reflections), game is slightly more GPU bound now and performance can actually improve, depending on CPU/settings/location used. For every other uArch, GTA 5 should perform the same.
For the DXVK part ... definitely recommend DXVK Async vs regular DXVK. Having shader warming without stutters (even if those would've gone away after a while) is great. Performance in some games improves, in others, frametimes are consistent, in others just lower CPU usage/temps in general, in others ... simply worse performance. GTA V doesn't work with DXVK for me unfortunately
Dying Light for me, on 5700 XT, gives me some 20 fps less than DX11 (so I'd get 100 instead of 120 fps). Assassin's Creed games though, GTA IV and Borderlands 1/2/TPS benefit massively from DXVK
Thx for answer. Just to satisfy my curiosity i just checked gta 5 under 2 different drivers and indeed it is placebo as benchmark on 20.5.1 and 21.2.2 gave more or less identical results.
About dxvk, yes, for me as well gta v doesn't launch with it. Also i tried all borderlands games and all of them, except for 3, completely ignore dxvk .dll files. At least they work for you and give you nice boost, good for you. The most severe case i found in my yesterday test was ac3 remastered which has loss in performance as on top of that audio got messed up.
Installing the wrong version will do that. Example, putting x64 builds in a game requiring x86.
half refresh rate setting giving you 38fps? 25 fps i have now idea
Has something to do with double/triple buffering. On displays without F/G-Sync the vertical frequency can only be half/third of the set frequency. 75/2 = 37,5(38) or 75/3 = 25. To ensure vertical synchronization of the hertz(frames/sec) it locks you to one of these values. 75 fps on a 75 hertz is possible but only if it NEVER dips below that value or else it selects the next viable variable (37,5 or 25).
I lowered the resolution to 1280x720@75 and the game started with 75 fps. But at some moment the framerate dropped down to 38 again and never restored back to 75.
I miss the normal old vsync when it was just a limiter.
Do you have V-Sync activated ingame, in Radeon Settings or in AfterBurner?
38 FPS or 25 FPS referring to half (37.5*2) or 1/3 (25*3) refresh rate
BTW, V-Sync was never "just a framelimiter"
//EDIT: I just noticed that MattStiles already pointed that out
For Borderlands games (except 3) you have to use x32 DLLs (d3d9 a dxgi) and paste them to "..\BorderlandsXX\Binaries\Win32\" where main executables are located.
And don't use RTSS, it's showing some false readings there. Use DXVK HUD instead
RTSS might prevent some games from launching with DXVK (Assassin's Creed Odyssey/Origins for me), however Borderlands 1/2/TPS worked perfectly on my 5700 XT.
Driver 21.2.3 launched as well
Yes, i forgot about that. Windows xp x64 was released in 2004, it took over a decade to game developers to start noticing on bigger scale they should be making x64 executables for their games at least as optional executable file next to x32. Thx for reminding me games used to have x32 executables.
Unfortunately borderlands games get lower performance with dxvk on my 580. Out of curiosity i also checked physx effects. Unfortunately, despite having ryzen 3800x even this cpu is not enough to render physx at decent framerate and while game itself has over 100fps when physx clutter drops in fps can drop even to a bit above 30.
Oh well, enough off-topic, let's go to new driver 21.2.3!
Borderlands 2 at least has a visually impressive Physx implementation, even if optimization wise, it uses the old Physx 2.x library. On 5700 XT, there is clear performance improvement vs no DXVK for Physx. Performance still tanks drastically, but it's just slightly better than without DXVK.
Pre-Sequel has Physx 3.0? (or latest build of 2.x anyway) and Physx performance is slightly better.
And DXVK does wonders if you don't use Physx. Game is severely CPU bound with that "Ultra draw distance" at certain places. x86 vs x64 doesn't bother me that much, vs poor CPU optimization (just 3 threads, just DX9 in 2012)
XP x64 wasn't all that consumer orientated, to be fair. I'd say x64 broke mainstream only around 2007 with Vista, even if it was technically available since 2006. Mass adoption was around 2010 with Windows 7 if my memory serves me right (kinda everyone had 4 GBs of ram by then)
Unfortunately on rx580 dxvk doesn't improve performance in borderlands for me and i mean game without physx. My previous post may be confusing a bit on that matter
First i tested dxvk without physx, it degraded performance compared to dx9 then i launched dx9 with physx medium and general performance was ok but with physx clutter can severely drop down to 30-ies.
Yeah, I dunno what's the deal with that. I tested Assassin's Creed Odyssey and a friend did so with his RX 580. I was getting LESS CPU bound and GAINING performance on 5700 XT (R5 3600) (while being vastly smoother as well). He was LOSING performance while getting a smoother frametime with his 580 (R5 1600). He lost 5 fps from 47 or so with his settings.
I'm not sure why that's happening, given as far as I know, there is no extra feature set available in DX12 or Vulkan in GCN4 vs RDNA1 (aside from having Vega's Rapid Packed Math and Primitive Shaders available)
Had to say this driver suck in dx9. Like I have half less FPS. 21.2.1 works ok, not great but ok(because it utilize less than 90% GPU any given time).
Wonder if 21.2.3 is better in that regard.