Radeon Graphics cards Black Screen bug is haunting AMD

There's been polls recently too from the lead in the Vanguard program on this issue. (AMD's beta program of sorts for upcoming drivers.)
https://www.reddit.com/r/Amd/comments/f3dabe/looking_for_a_quick_survey_response_for_navi_gpu/

https://www.reddit.com/r/Amd/comments/eo6om7/looking_for_some_information_regarding_potential/

Been having the problems since the early 19.8 drivers or when I picked up the GPU but AMD has fixed several conditions for this in earlier driver releases but from the Linux drivers there's also comments on disabled stuff for workarounds and the GPU hardware fixes in the later released 5500 and 5600 GPU's plus the issue with the performance fluctuations and framerate drops took almost half a year to be acknowledged but at least things are happening.

Mainly only happens in a few games now for how things are on my own end but I've found no common pattern to explain what it is or how it's happening though the issues with Vega and Polaris would be a driver regression from 19.12.2 to 20.2.1 or Adrenaline 2020 but for Navi it could be anything or possibly a combination of several errors with the same end result, GPU shuts down.


EDIT: As to replicating the issue actually from what I read yesterday it sounds like there's this area in Witcher 3 with a gear piece and that's where it'll happen almost always.
https://guides.gamepressure.com/thewitcher3/gfx/word/527300430.jpg

Seems especially likely with shadow quality increased beyond high so v.high and ultra though why that is I'm not sure about, should increase the draw distance of the cascades but the game is still very conservative about shadow resolution so it should be 3072 or 4096 not the tweaked values of kill-framerate like 8196

EDIT: Hmm fillrate related well that's kinda vague really, buffers...pixel shaders and yep still vague and not really a answer to anything.
(Not like they'd program the GPU to terminate itself out of shame of failing to handle some demanding effects is it. )

 

Witcher III issue has been replicated, until it's fixed, DXVK can be used

 

For a while now I've had a random green/purple flash on the screen if my monitor is set at 75Hz only in web browsers and the Windows desktop using an RX 570. It doesn't seem to happen in games, and the problem goes away if I set the refresh rate to 60Hz (monitor is an LG 23MP68VQ)

 

Hi amd marketting dude, did you get permission from AMD before posting here?

Nvidia has to declare themselves when they post on behalf of the company, why doesn't AMD?

I have secret unreleased pictures to prove KissSh0t is a intel igp toting spy.

atleast its not crashing anymore

 

That's good,replicating the driver crash issues tend to be the toughest part for this error due to the randomness so hearing that at least one condition can be tested against and fixed is very positive.
Could be game specific but it's still a nice fix and if it's something that affects D3D11 there's a chance of it fixing other things too and over time these D3D issues will be resolved.
(Or what they are, can only assume it's that since DXVK bypasses the issue.)

 

The DELL sourced Win 10 x64 driver for TERASCALE cards (8490 1gb's and various cards all the way back to 4xxx~5xxx or whatever Terascale was), works just fine with zero issues. Used the heck out of those - for non-gaming machines (well, in-browser games don't count, card games in Windows don't count). Can say that the Dell provided driver is pretty rock-solid stable here. (I have boxes of these cards from surplus, so they do get use, but they're cheap at about 7 bucks a card shipping included).
Now, if using a more recent Polaris or newer card for browsing/movies etc, yes, I would believe I could not recommend 100% without the possibility of failure in there somewhere.
AMD drivers are an absolute train wreck. Thing about it that is causing so much problems for retailers is, many of these cards will appear 'bad' 'defective' or 'dying' to the end user even if the user has a lot of hardware experience, as the symptoms do mimic dead/dying cards quite well.

The issues with AMD cards having buggy drivers is very real. I own AMD stock. It's not killing the stock as most (60~70%) of the value is from the CPU division's success - but it would be nice to have the GPU division helping things, vs hurting things (black screen bugs, 5600 series launch disaster).
If they don't get their act together soon, it's going to cause even more retailers to stop carrying AMD GPU's.

If folks want to test if it's core clocks causing things to go POOF, down-clock the core by 100~250mhz (from stock clocks). If that solves it, consider keeping an eye on core clocks through MSI afterburner (or similar software like Riva Tuner), and log them periodically. See if the core is still going over what you set it. This can happen even if you NEVER overclocked the card since you opened the box, or never even installed OC software.
EX: My 1288 core clock Polaris-based card would frequently crash and have long 1 or 2 second long freezes in games, but not ALL games.
Installing monitoring software (which had not been installed on the PC before), showed me that when the core went above 1288 the freezes would occur, and when the card went above 1300mhz, sometimes as high as 1306~1320mhz (!), the application would hang due to a driver crash. Occasionally when the driver would crash, it would be a black screen and require a reset of the hardware. I have multiple samples on different Ryzen machines, so if you have Ryzen + Polaris or Navi and you're getting issues, mind that core clock! No one in our circle has any recent intel machines anymore (well, except one person, but he has a 1080Ti plus more money than sense), so I couldn't test this on intel.
Work-around: Set core clock 100~250mhz lower until the issues go away. If you're still getting some hitching/freezes but not black screen or driver crash, and they go away a second later and come back in a few minutes, you're right on the edge, just lower core mhz a bit more.
DirectX logs will report DEVICE REMOVED in the log files with these driver crashes while running DirectX applications and getting the above driver crashes. Honestly it seemed to me the reason this is happening on the 4xx cards more than 5xx cards (for the Polaris issue) is because the 4xx cards are being told by the driver to run at 5xx card core clocks, or otherwise over and above what they're stable at.
Some golden samples might never have an issue at all, as they can handle the higher clocks.

 

Separate cables confirmed.... I think there's a fault in my Silverstone 800W maybe I have dodgy parts or parts which passed QA at factory but are now dying an early death.... hopefully will test with new PS this weekend...

 

I've had black screen issues with 2 Vega 64s (1st was RMA'd for a separate failure) in Windows. Screen would go black, very rarely under load, often at idle, and the whole system would lock. The only way I could tell what happened was the NumLock/Caps on the keyboard would be stuck on/off. Only way I could find to fix it was a hard reset. I think it started in 2019 but it could have been 2018, first few months I had zero issues at least. In Linux I haven't had a single issue.

 

Any news on a new driver with the black screen fix ? They are slow as snails on the driver department at AMD , hardware is as good only as the software that makes it work. I would like to use newer drivers on my back up PC with a RX 590 but i cannot since i too suffer from the black screen issue , so is not only the 5700 video card issue as Polaris are affected too. The longer they take for a solution the more money they will bleed as people have read the horror stories all over the net and are not buying new AMD cards anymore due to all the driver bugs.

I am all into supporting AMD but if all this driver problems continue i will only buy nvidia cards from now on. AMD need to get on top of the game and instead of releasing heavy bloated software , give us the choice of what we want to install from the driver package as i hate their new software suit as i do not stream , do not need AMD to tell me to update , any advisors ...

 

https://community.amd.com/thread/176003

You can try this - some people are reporting it's working for them and they don't get blackscreens anymore:

https://www.reddit.com/r/Amd/comments/f6ccqo/a_tale_of_radeon_adrenalin_2020_ver_2021_optional/

No idea if it actually works though. Seems weird that AMD wouldn't catch it.

 

AMD never had better drivers, they had nothing to fix or improve performance in specific engines, so they decided to use their "free" time to revamp UI and turn customers into a product with all of the tracking features that can't be disabled.

Now, I believe to some people on this forums who used to tell me "drivers are rock solid, its faulty hardware". At least AMD did a good job this time exposing their "faulty hardware" with release, instead of years later...

 

Block RadeonSoftware.exe with windows firewall then it cannot send or receive data.

 

Polaris has base and boost clocks under the term Adaptive Frequency and Voltage Scaling (AVFS)

 

Hmm...err... I singled out the 5700 series on purpose...

 

If you use Wattman, the boost clocks are presented as another power state, yes. If you use GPU-Z, it shows "GPU Clock".....which is the base 3D clock and "Boost", which is the preset boost clock. The box labeled "Boost" in GPU-Z, has the same value as the last power state presented in Wattman. The Polaris cards won't exceed the clock speed set for the final power state without user action. The cards don't have the ability to "boost" like that. The drivers also won't change the value in that field "automatically". That field is populated by data stored in the vbios. It's a static figure, unlike NVidia's boost clock, that can vary based on operating conditions.

If you want a laugh, Wattman....or, to be more precise, the Performance->Tuning section of the new UI is showing Junction Temperature for my RX5700 as being 800C (1472F)..... Anyone got some marshmellows????