Discussion in 'Videocards - AMD Radeon Drivers Section' started by Hilbert Hagedoorn, Apr 23, 2020.
Probably when we get the hardware scheduler support.
AT Least its WHQL Driver mine wont take the newest yet for some reason
Been trying to play AOE 2 Definitive edition as well, it blue screens every time or crashes to the desktop, i also get the occasional bsod in some windows apps.
I have tested most games and they all run fine( shadow of the tomb raider, Rage 2, Deus ex;Mankind Divided ect ect) , the occasional stutter ( still annoying sometimes makes the game experience terrible.). I also tested another kit of DDR4 memory so it cant be ram causing the bsod, it all point to the gpu.
trying out different driver versions now.
Throw the overlay out and see what that does I would suggest, for RAM you can try without the XMP and if it still happens well it should at least reduce the chance of it being XMP/problematic timings related.
Also if it's running in the background close the web browser or disable the hardware acceleration but if you have a few games were you can easily replicate the issue I would definitively start by throwing out the overlay first since it has some known problems so it's a good starting point if you suspect it's a software issue.
EDIT: Anti lag and enhanced sync should be disabled too, anti lag might work reasonable well but enhanced sync is still not ideal.
VSync overall is kinda broken but a fix landed in the Linux drivers or rather kernel a short while back so maybe the next Windows driver will fix that. (~five months and counting reported multiple times yeah it's kinda silly.)
Can be worked around, introduces a few slight issues of it's own though.
Also it should cause stuttering together with moving the mouse due to skipping not blue or black screen tier errors that's kinda really bad but this is another longer standing issue that is awaiting a fix so the situation well it could be better.
EDIT: Throwing out the overlay should work by just opening the task manager and terminating Radeon Settings Software and the two AMD overlay services not the two other AMD services or stuff related to DVR just the overlay or server service and the desktop overlay one.
Might as well close the settings software while at it, opening it up will then restore the overlay unless the two files for these are also renamed so it's easy enough to get it back and the settings software doesn't have to be running in the background for this test if it does anything else which might as well be eliminated entirely ensuring nothing extra is loaded or running in the background.
EDIT: Really though some of these should be issues that get other smaller stuff pending and awaiting fixing not getting fixed while these remain pending especially after Vanguard testers or others can start replicating and figuring out what the issue is.
Well it might still happen and there's a chance the next driver or two might see some of these things resolved at last but the amount of time it's taken is pretty bad.
HI, i have actually installed the radeon drivers the old school method by installing it through device manager and not installing the radeon overlay software, it does not really make a difference because i still get BSOD and i tried all the above methods regarding enhanced sync , Vsync and radeon anti lag, it makes no difference with the micro stutters.
I have tested my memory without XMP and still the same results and i tried some other memory just to rule out bad memory but unfortunately i have the same results.
I did memtest 86 and Karhu ram test no errors, my pc runs cool and has good cooling in it and i am not running any overclock besides the Xmp on my memory.
I get BSOD in AOE 2 Definitive edition, when i work in my music production software and occasionally in windows.
In all three those instance it always comes up with this error.
"IRQL_NOT_LESS_OR_EQUAL" caused by driver ntoskrnl.exe when i get a blue screen.
I will try the above mentioned articles on the hardware acceleration.
Thanks again for your help, i truly hope AMD can get this sorted out as i dont ever see myself buying an AMD card again.
Hardware acceleration is two-fold for Navi with one being problematic VP9 video decoding performance and possible stability when doing video playback on the GPU though only Navi has this support and it looks like AMD has upgraded this on the newer upcoming cards whatever that might be updated with.
Open issues seem to be a lot of dropped frames especially at 4k and then Media Player Classic had updates specifically around driver bugs on video playback so it's not just Chrome or Edge or Firefox and browser related.
Second one is back and forth in regards to overall web browser hardware accelerated browsing stability although this can be a blank or glitched tab and not just a GPU crash or driver reset or just the browser crashing and it doesn't have to be from doing media playback or a image heavy site it can also be lighter workloads or just having the browser open but this goes back and forth although I thought the current status wasn't too bad currently but it's the same as many other reports and very varied and inconclusive.
Navi's sensitive to memory and possibly also something with just the I/O connectivity but it's a digital signal with HDMI or DP so 1's or 0's or it's either there or it's not so yeah it's tough to find anything definitive for answers.
Some displays might also have less robust scalars but this is a display image loss and not a blue-screen or such like what can happen with memory instability.
Rest is the usual on PSU being able to handle high transient power draws including spikes and how the GPU clock states can fluctuate but without the Radeon Settings Software Wattman issues wouldn't be a thing so it should mostly follow bios defaults here.
Though the card likes to try and boost and push as much as it can with the voltage and wattage available and AMD's and some vendors use a pretty hefty voltage and target power draw or wattage limit here so when it spikes you can go pretty high for what the hardware (PSU) might tolerate but these are more like shutdowns, bit of a timeout and possibly a automated boot up again like tripping the surge protection on the PSU.
(Not entirely though, earlier drivers had bugs that caused the same behavior especially with even moderate memory overclocks but again no software and no Wattman thus stock speeds including memory.)
(Shadow of the Tomb Raider had that going for a few benchmark sites including Gamer Nexus and some of their earlier 5700 / 5700 XT reviews.)
And then a dozen or so recommended practices short of swap to a 2000 series NVIDIA and see if that works better which well it still seems to help for users picking up a 5600 or 5700 model even now so I have no idea what the core issue here is.
(Just how sensitive are these cards and why and so many possible questions here.)
My own experience was kinda problematic with 19.8.x in particular being incredibly unstable and then eventually 20.2.2 stabilized much of the driver side of things (Hardware acceleration off though and no overlay and keeping enhanced sync off.) coupled with DXVK to remedy at least in part the stuttering and clock speed fluctuation and framerate variance with D3D9 and D3D11 plus somehow also improving stability in many cases although part of this appears to be the titles in question also violating some of the D3D practices and recommendations or doing other non-standard or entirely wrong implementation of some functions too.
(Native Vulkan or D3D12 and the card just goes without problems for hours, can't say the same for some of the D3D11 titles but it works through wrapping it with DXVK at least but it's still a bit unknown whatever it is that is causing that problem.)
Not much of a answer mostly just mentioning some of the listed issues again and a few of these have open tickets and tracked issues for the browser and media playback problem or in-place mostly working fixes or workarounds until AMD can fix the driver side or work around what hopefully isn't some more problematic hardware fault in VCN 2.0 then fixed in VCN 3.0 coming with Navi20 in regards to VP9 and 4k video playback and resulting issues.
It can work though, other than the initial 19.7.x drivers Media Player Classic and with MadVR has done video playback just fine although I can't say the same for browsers and stuff like Youtube but then hardware acceleration overall had problems across various browser agents like CEF found in various software such as Discord or some of the clients like Steam.
I would assume current drivers to be more reliable in regards to GPU hardware accelerated browsing but the media playback seems a bit less robust and I can't vouch for keeping the browser running in the background affecting stability or not but closing it down should quickly confirm if it helps or not.
Would hope it does but these issues can be a real problem to easily identify and troubleshoot unless you can narrow it down to I don't know memory instability or GPU hardware faults which also wouldn't be easy to do short of RMA's or something for the card and trying the new replacement I guess?
(Some stuffs easier to see but otherwise it's hard to say what's causing it.)
EDIT: Curious too what video memory testing might have given but it's not possible to adjust voltage easily here if that might help nor reduce clocks below the existing default and with error control in effect at best if it's not unstable you might just see a slight decrease in performance when using higher than default values instead of a small increase and then only maybe for some titles or very sensitive games.
(Witcher 3 pops up frequently, Battlefield V I think was another and Overwatch I believe probably a couple of others too I might be missing that otherwise crash fast.)
Two separate memory modules as well and no easy way to check through software, Samsung and Hynix and some difference in timings though clock speed is the same as is voltage AFAIK.
(5600's getting some bios updates to push it to near the 5700 speeds if supported but that's about it.)
Might be something to how AMD and some of the third parties handle these and hardware differences but since the stock AMD PCB is seen as really robust and over-engineered and reports exist on the various stock PCB models and then also AMD's own models including the 50th Anniversary card that probably doesn't mean much here.
There's a lot of suggestions and things to do but by now a lot of it has been tested and results are just mixed as always but I would hope there's a low chance of AMD just having a very high "dark" number of defect hardware on this card although if low-level API's are supposed to be more straining then cases like my own where these work best should be kinda non-existent and would quickly identify unstable hardware even after reducing AMD's driver software or other third party programs.
Plus this overall sensitivity thing and switching to a earlier AMD GPU model or NVIDIA and then it's all stable?
Somethings going on and it's not like I would know but I would really like to see it narrowed down and answered even if it takes the community a while to figure it out unless AMD makes some statement on what's going on here. (Probably unlikely.)
EDIT: Format the disk driver and reinstall, swap PSU, swap PSU cables, swap display, swap display cable or connectivity like DP->HDMI or HDMI->DP and then a lot of settings both in the OS and drivers and third party programs and much much else gets suggested.
Some of these problems and issues with the software like the display drivers can be confirmed like that VSync cursor issue and some of these have then been confirmed and replicated other stuff seems more coincidental although some cases have also been due to other hardware particularly sensitivity to memory if the GPU wasn't just problematic and needed a RMA.