AMD Radeon Adrenalin Edition 20.2.1 driver download & discussion

Let's hope AMD is hiring some good people...)

Don't have any black screen issues(aside from on wake), so I can see it being not super simple to pin down. Is it not possible to make a super light program and have people with the issue install it, if other ways are no good? Maybe bake it into the Settings and have it available as a toggle?

 

it has become most apparent recently that AMD is not testing their drivers in anything except the titles of the day, not good to have a radeon pro and having the driver break every opengl application on the market, whether game or workstation.

 

Here is horror story of mine:
My 2nd M.2 SSD does not remember RGB configuration for its LEDs through reboot/cold boot.
Therefore every time I boot my system that used to be colored as RED/Light Violet, I have Freaking Rainbow running in circles under my RX 5700 XT.

And I am pretty sure, that this Rainbow light takes away like 5% of my fps.

 

Rainbow is supposed to have 7 colors so your 5% performance loss is negligible. Hahah.

 

Won't be for long. Will replace it soon and then I might have to mess with the display timings for the new monitor.

 

Got zero problems w R9 380 everything running smooth.

w10 v1909 / Firefox.

 

Wolfenstein Youngblood plays smooth, FFXV also. Free games from this week and on through xbox game pass. Driver is nice, np.

 

BFV DX12 black screen and no signal error since 19.12.2.

I wonder when it will be fixed.

 

Had two black screens with these drivers and couldn't RDP in. I'm back to 20.1.4.

 

I am back to 19.9.2 to avoid the issue(or similar) you mention.This time I used the AMDCleanupUlility which is stored in bin64 folder.Any cons on using it for now on if I have a problem?
On the other side is it ok to just install on top if there aren't any problems?I am tired of the procedure.

 

No signal error in 20.1.4 and 20.2.1 with my 5700XT. Shame AMD can't get their drivers right because this card is a beast. Using 19.12.2 to avoid the no signal errors for now...

 

I had two black screens in total so far, with 2.1.2 and 2.1.3 drivers (one each), after the second someone in the forum here told me to turn off HDCP, which I did and I haven't had any more black screen yet (I've been using 2.2.1 drivers for a couple of days now). For the rest I don'y play many games, just D2 which works fine, very nice FPS, although I see the temp go very high in GPU junction, reaching 99°C under stress; I've got an XFX 5700 DD Ultra by the way.

 

I haven't experienced any black screens with my new setup, as 20.2.1 is the first driver package to touch this 1909 OS install. Using Chredge as my browser, card is a Sapphire Nitro+ 5700 XT. Motherboard is a x570 variant.

 

It's improved though some games are more sensitive than others and in my case since DXVK works around driver crashes including black screen issues it's some specific D3D11 error but I can't quite say what's happening and what else might cause it to occur.

Been experimenting with Ghost Recon Breakpoint as the driver will crash but not blackscreen around 15 minutes into the game but it can run for hours with DXVK (Left it overnight even and no issues there.) but no combination of overclocking or underclocking including increasing or decreasing the voltage affected D3D11's stability meanwhile via Vulkan wrapping I can hit low voltage and high clocks so something with D3D11 but what could it be.
(I do notice the fan speed does the typical 100% and gradually climb down and then hitting all over often overshooting the curve too while Vulkan keeps to the curve/stepping more accurately and doesn't ever go to that 100% thing and dropping down that the drivers really love to do now.)

Atelier just about any game via Koei and Gust induce the black screen unrecoverable crash but these have problems including bottlenecking the PCI Express slot (HOW!) sending checks four times - every frame. - (HOW!) resulting in GPU performance at best reaching 80% and numerous other issues. (...Yep...HOW!)

Overall the exterior of the ports or what to call it has improved (Mouse support, resolution support of sorts kinda depending on die roll or something and other stuff.) interior or the deeper code that's a bit of a flop-up.
Makes for good stability testing though and again wrapping it resolves the stability issues though without third party modding efforts the ports are inherently very broken so that's a thing too. (It runs but yeah the issues when looking into are really bad.)


Would be nice to point on something like oh it's a compute shader issue but well Atelier doesn't have those and Assassin's Creed Origins and Odyssey loves compute shaders and those are stable nor does it seem to be some general compiler bug as some older titles heavily reliant on the usual vertex and pixel shaders or tons of geometry and whether it's trying to be mostly single threaded or multi threaded can run just fine both D3D9 and D3D11 although performance can be just about anything since lower GPU usage really reduces clock speeds on Navi so it doesn't always cover well for frame pacing and resulting stuttering or small hitches as it boosts and downclocks constantly or even gets locked into running at low to near idle speeds because the GPU usage isn't 90% or higher ideally 98 - 99 percent for optimum boost clocks.

EDIT: OCP or over current protection could also be a thing as the GPU clocks up and down and spikes and drops but that's random too and it can happen for titles that are demanding enough to keep the GPU near constant performance levels and clocks/voltage levels and it can happen on older titles where it drops and boosts constantly.

I expect the release notes to mostly just be "Fixed a issue with the RX5000 that could lead to a black screen display signal loss." or something like that but it'd be really interesting to know what this actually was or if it's a variety of errors and some unexpected behavior although since it regressed and affected Vega and Polaris Watt-Terra-Flop did AMD even do with the drivers to cause that backwards breaking issue that should never have happened on earlier stable hardware.
(Or that too is a different issue same symptom thing but still a bad one whatever causes this.)



EDIT: Well to wrap up this somewhat long text it has improved and the driver situation far as stability goes is getting there but there's more work to be done and who knows how long it might take to sort out this remaining couple of issues just because of their randomness.

The clock speed fluctuations too but they're not as problematic though it does need some improving especially for when the GPU falls to near idle speeds in a situation in which this shouldn't really happen.

 

The downclocking behaviour is ridiculous. Not all games load the gpu 98%. Its been 7 months. Also a lot of people want to run multiplayer games with the lowest settings possible for high fps.

 

Yeah at least the Linux drivers are holding a higher minimum speed instead of fluctuating around 6 - 100 Mhz which improves overall responsiveness for applications using the GPU but when it comes to gaming it seems to be controlled almost solely by GPU load which would be nice to have some control over so it could boost as long as the other limiting factors of power draw, voltage and temperature are below where the card starts clocking down instead of immediately falling from the targeted boost speed (Or close to said target at least.) as soon as the GPU load goes down which depending on game optimizations and all kinds of factors could be briefly in the 70% - 80% but from monitoring the card drops to almost half speeds far before it really should and the resulting performance drop and how this affects frame times and how smooth or stutter free things feel it's pretty problematic.

Some power saving for idle and less demanding tasks isn't bad but the way the GPU operates just feels very overdone or how to describe it throttling down without any delay or user control.
Wattman wise just a target max allowed voltage and max boost speed seems like it'd do for Navi as the rest are almost ignored entirely or automatically generated the way it works.

Since the way the GPU scales differs between Linux and Windows it also means this can be affected in software so resolving the random clock fluctuations and problematic in-game drops to low or idle GPU clock speeds should be fixable it's just taking a bit too long but hopefully it will be resolved sooner rather than later whatever the next driver release will bring.
(February isn't over yet so there's room for a 20.2.2 driver whatever the next couple of fixes might be for.)

Well I don't expect it to be resolved quite that soon though before mid 2020 would be nice for how long it's already been now that AMD has taken to listing it under known issues which at least is some progress on this situation.
(It's not fixed nor discussed much from AMD but at least it's there now in the known issues section.)

 

Thats the only thing that bugs me, i only really need super high fps for esports games,, im a casual shooter player so i just lock at 60fps for those and action games ect

but fifa runs at 150fps with like 20-30 percent usage, the clocks do jump about causing the odd stutter too

 

Getting stuttering with this driver in BF1. It's annoying. Didn't have this problem with previous drivers as long as I increased the resolution scale to prevent downclocking. Not working anymore.

 

Which e-sport titles are you having issues with? Dota 2 (and PUBG) seem to run fine here.

 

i dont play any of those 2 games

recently i been playing fifa mainly and just finished MW campaign (borrowed from a friend) and no issues at all with that .. sp locked 60fps 3200x1800 absolutely fine