AMD Radeon Adrenalin Edition 20.5.1 driver download & Discussion

Would be interesting to have a whole bunch of 5700's and get down to "Yeah it's the hardware." in case of minor errors with the memory or whatever but there's still reports such as the AMD Reddit about people picking up the 5600 or 5700 and hitting black screen errors on various systems with the usual suggestions and workarounds and possible problems around system memory sensitivity and it's either very common or very infrequent and mostly random with some games being more prone to it (More RAM? VRAM? sensitive or just pushing them more.) but no confirmed fix that just works and then swapping to a Vega or 500 series or such from AMD or a 1000 or 2000 series from NVIDIA and it's just gone.

Somethings causing it but not enough to where it's 100% and just run this game like this and replicated and fixed and it was a driver issues but then AMD has also fixed almost a dozen or so different black screen errors in the drivers too unless those are all timing tweaks and memory related relaxations for not pushing the modules hard enough to where it hits errors and possible differences in Samsung compared to Hynix VRAM reliability or timing differences and overall error rates even on stock speeds.
(Being allowed to reduce the stock frequency or alter voltage more easily would really be interesting but I think the bios editing and tweaking stuff like timings/straps are where it's at currently with some smaller differences to overclock results and again very prone to instability.)


EDIT: Effectively test a more easily replicated black screen issue on my own hardware in some games where it happens more commonly and if 6/10 5700's don't hit the issue it could just be a higher overall fault in something in the hardware which even if my own card has improved as of the 20.2.2 drivers and works really well for D3D12 or Vulkan I can't really discard that as a possible explanation either.

Software issues being software should be more readily apparent for everyone on a driver known to be more prone to these crashes whether as the GPU is under load or just when idle but it's all just so random making it impossible to test effectively without a ton of trial and error or just chance even in some of the more sensitive games where this happens more often or at least is reported to happen more often.
(Popular game, more people playing said game and thus more issue reports or said game also being a factor.)


But yeah after everything else and going by AMD's fixes and randomness that's something I am thinking could explain it but then why would stuff like DXVK improve stability not decrease it and what exactly *is* the error itself with games like Overwatch, Witcher 3 and some other titles more prone to hitting errors.

Then again memory issues and what and how and how hard it can be to test for this even with more random reliability stress test modes and utilities not just more common or ordered reads and writes and it looking stable but maybe isn't 100% stable in actuality.



EDIT: Well I wouldn't know, feels like it's the only thing to explain the randomness of this issue but then the frequency of these reports even on newer manufactured hardware or the 5600's would be a problem for a higher than average (Possibly?) number of faulty cards and something possibly down to the memory as the problem unless the user can also resolve it if it's system memory related or stuff like that which might be stressed more with a 5700 than other cards too.

So many questions and no real answers I suppose, just a lot of theories and suggestions and more or less effective attempts at fixing it up and including replacing the card.

Really would like to know but if it's not hardware what else could it be, well unless the newer drivers do fully resolve it I guess we'll never really know.

 

Initially I was also wondering if it was shader related or possibly shader compiler related, Gujian 3 worked just fine until I got to the next major city and there's a pond with water behind it, just go there and look out over this and *boop* and the card and system powers down. (Or used to, DXVK and once again stability just kinda works for whatever reason.)

But then there's Overwatch or Witcher 3 and for that game suggestions such as "Go to Novigrad" or just doing some testing in Battlefield 5 if that's stable or not at which point it's mainly back to memory, system or video can't quite say which really.
And other titles or even just when idle on the desktop and not using the browser with hardware acceleration or doing media playback.

It's all just incredibly random and very hard to narrow down to any one thing.

 

i get you point infact in the "big fix release" AMD fixed a lot of stuttering/blackscreens/fpsdrops in very specifig listed games: point is that they haven't found a generic card/driver related issue so they placed something like a workaround to any specific game that they know was affected, or some game that was reported to them via their website report o in this forums/subreddits. So the question is if they haven't found a generic all-around fix to the blackscreens, they have to test any new game and make a workaround in a new release if it crashes at some point? if many black screens were fixed in the past by updating drivers it means that it was not a fault in DDR4 XMP memory or 1 vs 2 pci-x cables or pcix 3.0 vs 4.0 in bios etc

 

I recall back in the day where new drivers would change gram subtiming. Back in the day tighter timing could net you up to 8-10 fps boost using the same oc.

However, the down side was long term stability. You could game for days/weeks without a problem. Then get a hiccup.

Which is what those black screen reports remind me of.

 

You could be getting memory errors and not even know it. Performance drops and you get hiccups like you said maybe even a crash thinking whats happening.

I had my 580 at 2200mhz and very tight timings on top of it and until i realized whats going on. Hwmonitor helped in that case having monitoring for it on polaris cards.

 

The memory errors could explain why the clock rates would drop.

 

Yeah for Navi at least there's ECC support so assuming it can catch the more common errors the generally noticeably difference is that performance lowers from increasing memory speed.
Memory itself is also pretty sensitive so the default speed might already be pushing it pretty close and 19.8.1 already did something to try and stabilize it as increasing it during game runtime frequently caused a driver crash or system reset.

Other GPU's and you can hit near or above 1800 or even the max 1900 plus the various bios editors for modifying timings (The straps for various clock frequencies.) though from what I can tell in the bios data through TechPowerUp both vendors use different values here so there's a further chance Samsung and Hynix behaves differently under a overclock (Can't underclock or adjust voltage sadly as these parameters are locked.) which may be why Hynix seems to be able to hit higher speeds but that doesn't mean it's more robust it could be using laxer timings above the default speed or it could hit errors in different scenarios due to differences against Samsung modules.

And again with how it's pretty random and hard to test for, it's not like you'll get these little green glitches or other display artifacts and the higher speeds could also act up during lower GPU load situations or when the card does it's little clock fluctuations or spike to higher values even for a instant plus how different software stresses the GPU and components in various ways and also for memory as seen with RAM benchmarks and different stress test "patterns" for getting at least some sort of reliability value from testing.


EDIT: Error correction and some other additions are coming as a standard in DDR5 too aren't they, although RAM modules also have a lot of differences and then binning for what the kits should take coupled with what profiles and tweaks the actual motherboards apply as for some reason the little cache or storage for that XMP profile seems kinda insufficient so a lot of kits lack some of the finer timing values specified.
(Though newer AMD firmware / AGESA updates and the overall robustness of Intel when using XMP has seen this work reasonable well even if not entirely perfect.)

Can't say much for the Navi memory controller either, might be sensitive but it's probably just memory module binning variance and the overall differences from card to card at least for when users are applying overclocks here.

Lowering the stock speeds would be interesting to test although it'd pretty much confirm a memory issue if things were to improve far as some users stability experience goes or if going back to earlier drivers without whatever tweaks applied here would somehow be stable suddenly.
(Well that would require being able to hard edit these values through bios.)

Voltage editing could also be interesting although how sensitive are these and how much extra is AMD already using here, it might work with higher but degrade or hit errors even pretty bad ones quickly if the default is already close to pushing it.

 

Petty, if that's true. I would have thought, by now, that the timing would be more relaxed the higher in frequency you went.
I never understood why AMD/Nvidia keep vram timing a "secret". Last I heard about it is that timing can actually fluctuate. But I've not read up on that in some time now and can't confirm if that's the case in Navi or not.

Hmm, perhaps testing is need at 1650Mhz, 1750Mhz and 1850Mhz just to see what the outcomes would provide.

 

@JonasBeckman : I am kind of sure that Blackscreen issue is not between GPU and display. Running 240Hz and no issues.
My card is even overclocked and undervolted at same time. May be power issue for some. My cousin got old FuryX from me long time ago. And some time ago he got my old 144Hz screen. Basically setup that worked for me has black screen for few seconds before image gets restored in his system. And once he even had it permanent till he turned monitor OFF and ON again.
He has like worst electro-installation at home you can imagine with extension cord plugged into extension cord. Cables going across/over/loops with other cables...
At times he is issue free for a day then he has it every 5 minutes.

It is hard for me to blame HW that was issue free with me and has issues with someone else.

 

Yeah there's a lot of different factors and unknowns and a ton of possibilities which includes not just Navi but Vega and earlier.

Looking forward to seeing what 20.6.1 and later can do plus re-testing things more extensively once the new motherboard bios and possibly further updates to the chipset driver are available though the current heat wave has seen mostly work on less demanding titles or looking up playthroughs of some of the console exclusives. (The amount of details Vagrant Story managed for a PSX game was really impressive from a technical showcase and expertise in getting so much out of this system.)

As to bios and AGESA well this here is a Gigabyte board and they are apparently playing it safe although I had much the same issues with the prior system before mid-December and the new build and after that February and the 20.2.2 drivers which also seem to have resolved much of the idle black screen issues and other crashes although I still can't say whether it's hardware, software or a mix of the two at least for my own situation.
There's at least some fixes in newer AGESA updates I'd like to give a test and see if some of the stability issues are gone but until a newer bios update is available incorporating those it's 1.0.0.4 and no further hah.


DXVK also continues to improve and is giving good results for performance scaling and also stability though the randomness of the latter makes even more crash prone titles difficult to test easily nor is it easy to say exactly what or why the drivers or the GPU itself acted up.

There's the known bugs and issues too but they're a bit different such as video playback problems or browser glitches not quite the full system instability or driver crashes but they can happen.
Software like the overlay but that's easily renamed and then it's simple to just test without it loaded and compare.


Going to be really interesting to read up on experiences with Navi20 once those are out plus what the later driver updates could still do when AMD begins to fix these existing issues whatever the priority and current pending fixes are.
Plus whatever might change with the driver team and possibly expanding this if AMD continues to do well overall, pretty sure that's one of the main areas of improvement for getting more done but it's not exactly easy.
(Getting good programmers and such plus the pandemic since that's still a pretty big hindrance.)


EDIT: Well the new bios is available that was kinda unexpected.

 

I remember a while back that faulty power stripes were to blame for odd PC issues a few years back. It would be one of the 1st things I recall troubleshooting. For some reason people had cheap Kmart power strips that cost them $5 or so, 10 years ago. Plugging the PC directly into the wall socket elevated those issues until they got a more reliable power strip.

 

I would advise to check the PSU.... there are PSUs and "PSUs".... Check on a certified silver+ or gold one if the problem remains.... I agree with @Eastcoasthandle on that --- might be cheap power cord, Power Supply, or sometimes just properly re-seating the card is enough.

 

Did you report the issue you have specifying the software/game and your full hardware specs to AMD? http://www.amd.com/report

 

WinX v.2004 Drv. 20.5.1 (6.10.2020) WHQL Installed (Clean install with preserved settings).
Everything is fine.
FRT Limiter from AMD-Catalyst-15.11.1Beta-Nov14 not working anymore (no biggy cuz' i have RTSS & Chill). Presets for Gamma working fine.
OverdriveNTool.0.2.9 is needed for Win 2004.
AMD Memory Tweak is working fine.

Tested so far:
FC5 = OK
BFV = OK
CIV VI DX12 = OK
Forza H4 = OK
DOOM Eternal = OK

Better latency on this one ~low 62ns, consuming less RAM, snapier.
Overall im happy with the Update.

 

You should try WSL DX12 driver it's better in everything compared to 20.5.1 whql at least at my system.

 

THX for the Tip but i will wait for normal one.
I have done update today + cleaning + setting up etc.
Im done , now i wanna play some....

Note:
Give me a link, maby some other day i'll try this one out.

 

Yeah I know the feeling. I play always before new drivers, I give priority to entertainment, I must feel bored after long sessions of gaming before turning into changing drivers.

20.5.1 whql they are a week old now 10jun vs 17jun, new is better with latest windows!

https://www.amd.com/en/support/kb/release-notes/rn-rad-win-wsl-support

https://drivers.amd.com/drivers/dx12-wsl-radeon-software-adrenalin-20.20.01.05-june17.exe

 

Getting black screens (black screen and fans @ full speed (at least in Post Scriptum) with my vega 56 and the latest driver. Back to 20.3.1 and zero problems with that driver. I did seem that with the latest driver my fans don't go as fast when in game, not sure if related though.

 

New Era of Radeon drivers ! (Warning! You'll be Brutalized by this )

(20.5.1 WHQL 10.6) WDDM 2.7 hits 28K+ Easy (Tess x8 as usuall, default Tess at low 28k)
Comparing with the last 2 best drivers.

-> https://www.3dmark.com/compare/fs/22930498/fs/21955959/fs/21656229

Spoiler: Settings used, always the same BTW

Note:
For this kind of score before i needed 1795MHz 1.162v !! and HBM2 at 1200MHz
Fine Wine?

 

Hmm that's more than my Radeon VII gets at stock. O_O

https://www.3dmark.com/3dm/47940139?