AMD Radeon Adrenalin Edition 20.2.1 driver download & discussion

With every adrenalin 2020 driver I've tried I get this followed by a black screen sometimes that needs reset.Is this the black screen issue that some people mentioned or just me?

 

That seems like it's something hardware related.

 

Oh boy... They mentioned the same to me somewhere else.It started happening after installing adrenalin 2020.I ended rolling back to 19.10.1 (with DDU) but it happened again.Now I am with 19.9.2 and for now all good but it's still too early.

 

Have you checked other monitor(s) or try changing cable(s) and/or input/output? Perhaps it is not the GPU itself.

 

I have a TV connected with my monitor.If it happens it always happens late night while using the TV with Madvr.If I remember well everything was fine before changing to 19.12.2.
Another thing that was happening with 19.12.2 was that the picture was freezing sometimes when going fullscreen with Madvr(mpc-hc).Windowed and fullscreen again was fixing that.These two came together and made me suppose driver issue.

 

Every day or two I get a Black Screen when using Media Player Classic, since the first 20.x.x driver. Gaming is nice, though, it seem to be something with video playback. Vega64 here.

 

Vlc seems to work fine but also no hw acceleration.

 

Still experiencing exactly the same texture problems with my 5700 xt when playing Modern Warfare the only driver that appears to work correctly is the 19.12.1.
19.12.2, 19.12.3, 20.1.1, 20.1.2, 20.1.3, 20.1.4 and 20.2.1 all cause the problem of disappearing and reappearing textures in MW.

There's been 7 updates of this driver since the problem was introduced in 19.12.2
I see no mention or acknowledgement of this problem by AMD in known issues.

 

AMD driver team is aware of this, even if it's not mentioned under Known issues.

 

I've been told this before but there's a big difference between them being aware of it and actually acknowledging it as a problem and fixing it. Even if they said they were working on it that would be encouraging.

I know a beta tester might have inside knowledge but they can't actually confirm much, only AMD can do that.

 

Happy to see the Downclock listed also maybe there now trying to to do something about it or at least figure out wtfs causing it.

 

AMD has no driver team and they are aware of nothing like those in Vanguard. Issue report collecting employees know the absolute truth. Driver's quality assurance is lower than ever since Navi launch.

 

Why do you think that AMD driver team is not aware of what we know in Vanguard team?

 

Nice reading about reason of possible downclocking
https://old.reddit.com/r/Amd/comments/eykth3/clarification_downclocking_is_often_a_symptom_of/

 

Hm
have installed it now - just wanted to give some feedback as i complained a lot about black screens / strange behavior of my vega 56 with 19.12.2 and 20.1.1

It has been fixed since 20.1.3 (as they also mentioned in the changelog )

So everything seems stable for me (until yet )

 

I said i wont but i still installed it. Played NFS Heat an hour or so with a chrome on, 30+tabs, music in the background, no issues. I dont think Polaris has been effected with these black screens all of these 2020 drivers ware working fine here.

 

Yeah part of it seems related to being CPU bound which would shift the bottleneck from the GPU to CPU and then the GPU operates at a lower usage level thus reduced clock speeds.
Part of it might also be display related possibly tied to the 120+ Hz thing where the card operates at higher idle speeds at least for the Windows driver plus the overall data and bandwidth differences though also tied to the display resolution.
(Since a lower resolution is less demanding and might also see less GPU utilization thus it clocks down.)

Interesting to read about the difficulties in replicating the problem though since some of the frequently affected titles have been mentioned often and API's like D3D9 seem to be affected almost uniformly or how to say though that does tie into CPU usage and overall less demand on the GPU too if it's a older title though dropping below game clocks or even into idle such as with Minecraft still seems like it really shouldn't ever happen whatever causes that.

D3D11 there's a bit of everything from threading and overall optimization to lack thereof, Koei Tecmo's titles might be a fun little compilation since the PC ports are hit by any number of optimization errors though that would also be testing against a broken bit of code so a bit skewed (IE how Atelier Ryza runs several checks per frame making GPU usage above at best 80% completely impossible thus the card will always be running at low clocks since full usage is again impossible to actually hit.) though this doesn't explain stuff like Sekiro and how DXVK can increase CPU demand (Wrapper after all.) in addition to delivering both higher framerate and more stable frame times and thus being smoother although some part of it might be due to AMD's D3D11 driver code and certain bottlenecks too for all I know.


Outright driver crashes also ties into it where the card might first behave erratically and then crash but this is a whole different complex issue I've barely gotten any understanding on at all for how things are breaking and I doubt I ever will.
(Part of it but figuring out the whole thing well that's something else.)

EDIT: Oh and the GPU settings and Wattman and such also factor in.
Optimistic defaults and the way the driver is kind weird ignoring or setting it's own defaults and limits instead of the bios values so you get a clock speed of 2100 Mhz at 1.2v only the power draw could be 180w or it could be 190w or it could be even 230w from the bios.

Different games different demands so something like Outer Worlds, Witcher 3, Borderlands 3 or the new Call of Duty seem to demand more whereas other titles might require less.

So the card boosts within what it has and what it's limited to and in some games it'll hit power or thermal limits or both and clocks fluctuate from that too.


Increasing the power slider is a quick way to supply (If stable!) the power supply and remove that bottleneck without touching anything else although depending on the GPU you might push into the limits of what the card can take and the frequent higher spikes and overall transient draws can hit both the GPU and importantly also the PCI slot or the PSU triggering a over current protection fault and what would look like a odd shutdown and reboot.
(Hard to track and monitor without some hardware though.)


EDIT: It's kinda odd though that the defaults in Wattman are a bit varied but there might be some explanation for it, seems more conservative in the newest drivers however rather than the earlier sets (First in 19.9.x compared to earlier and then again in 19.12.2 and compared to earlier for the two I know about.) but then Wattman can be a bit inconsistent overall and not the easiest to dial in.
(Even if you just save the defaults as a XML and compare against without enabling or changing anything.)

Defaults here as I remember are something like 2050 Mhz at 1152 Mv for boost and it's voltage now.
Though looking at the bios it's just a range so how it calculates the actual default values well that might be tied to other parameters or even the "quality" or binning on the chip itself perhaps.


And the 5700 XT overall has a pretty wildly varied number of GPU's ranging from 6 to 8 pin and power draw targets and overall tolerance and binning does at least affect how much it requires to where even the stock speed might be hampered by this.
(IE hitting 1800 Mhz boost instead of 2000 Mhz until some additional power is made available by increasing that slider.)

 

Weird, VLC is working fine for me even with DXVA2.0 enabled

//just uploaded a video:

 

Don’t think it has anything to do with awareness. It seems that if a problem is not under known issues then AMD doesn’t know how to fix it (yet) and when it finally makes it to the list, AMD has at least an idea how to attack it. In other words, we only list what we know how to fix. If we list everything reported we would look bad when some problems would hang in the list for months.

 

don't think I can fix high memory clocks with CRU. Displays are different in both horizontal resolution and also refresh rate. I'm not concerned with idle power usage (~33 W for GPU) or memory temperature due to clocks. I was just having an urge to change something given I have options, hence my question.