The AMD Driver Bug Thread

I cant be completely sure (cause i did change cases, but the new case should be just as high air flow as my past case) but my card when web-surfing idles at 41C with 11.3 drivers, where as before with 11.2 drivers it idled at 30C until i started doing some actual graphic intensive stuff then it would rise normally. I find that to be weird.

(Went from HAF 932 to HAF 922, but like i said...both have almost identical airflow so i really dont think its the case)
 
Then why do videos continue to play if you setup a profile that forces the card to stay at high clock speeds?
Er, what? The issue is not that videos require a specific clockspeed to play correctly, they'll play fine at almost any clockspeed you force the card to (with the exception of overly-slow clockspeeds). The problem is that the card glitches to an incorrect clockspeed when a hardware accelerated video starts playing, and WILL NOT scale up to 3D clocks until the video is closed.

Using the HD5850 as an example case, it is never supposed to run slower than 400MHz core / 1000MHz RAM (this is its idle clockspeed). It scales up to 725MHz core / 1000MHz RAM when you start a 3D application. If you have any kind of hardware accelerated video running, the card drops to 400MHz core / 900MHz RAM, which is LOWER than 2D idle clocks, and refuses to budge from that clockspeed until the video is closed (it wont scale up to 3D clocks even if you start a game).

Side effects of the incorrectly low clockspeed include lower 3D performance and display flickering (momentary or persistent). The RAM speed being adjusted is the cause of the flickering.

Still using the HD 5850 as an example, hardware accelerated video should run the card at 400MHz core / 1000MHz RAM, and should still allow clockspeed to scale up to full 3D clocks (725MHz core / 1000MHz RAM) if a 3D application is started. Currently, the BIOS on the card prevents this behavior, and overrides the driver with incorrect clockspeeds if UVD mode is requested by a hardware accelerated codec.
 
Last edited:
Er, what?

The issue is not that videos require a specific clockspeed to play correctly, they'll play fine at almost any clockspeed you force the card to (with the exception of overly-slow clockspeeds). The problem is that the card glitches to an incorrect clockspeed when a hardware accelerated video starts playing, and WILL NOT scale up to 3D clocks until the video is closed.

You sure about that, I think you should give it a go. Setup a CCC profile to lock the card in on stock 3D clocks and fire up a video. The video will play fine and the card will maintain the 3D clocks because that it all the profile you setup allows it to do.
 
You sure about that, I think you should give it a go. Setup a CCC profile to lock the card in on stock 3D clocks and fire up a video. The video will play fine and the card will maintain the 3D clocks because that it all the profile you setup allows it to do.
Does not happen, all CCC profiles are ignored and the card drops to 400MHz core / 900MHz RAM as soon as a hardware accelerated video is played. I've tested this extensively and AMD has verified the same behavior on their end.

Manually setting all three modes in the CCC profile to 725MHz core / 1000MHz RAM causes the card to run at those clockspeeds during 2D idle and during 3D mode, but the video BIOS overrides back down to 400MHz core / 900MHz RAM if you start said hardware accelerated video.

The only way I can get the card to maintain 3D clocks while playing hardware accelerated video is if I overwrite UVD mode's clockspeeds in the video BIOS with 725 / 1000, then re-flash it to the card. The card still sticks at that clockspeed while video is playing, but performance isn't impacted and flickering doesn't happen because it's glitching to a less troublesome set of values. I've provided two such modified BIOS in the very first post of this thread (one for the HD 5850, the other for the HD 5870).
 
Last edited:
The mouse cursor lag is still present in 11.5a. I haven't tried 11.5.
Ok, for future reference: When confirming the presence of an issue, state what card (Brand, model, reference or non-reference) you're using.
 
The only way I can get the card to maintain 3D clocks while playing hardware accelerated video is if I overwrite UVD mode's clockspeeds in the video BIOS with 725 / 1000, then re-flash it to the card. The card still sticks at that clockspeed while video is playing, but performance isn't impacted and flickering doesn't happen because it's glitching to a less troublesome set of values. I've provided two such modified BIOS in the very first post of this thread (one for the HD 5850, the other for the HD 5870).

Thanks thats good to know.
 
Alright, I'm marking cursor lag as unresolved again. Far too many people are reporting it, even with newer cards

Oddly, I can't get it to show up anymore on 11.5a... so they changed something, but obviously it hasn't worked for everyone.
 
So is the 11.4 still the latest CAP drivers out right now?

edit: got the 11.5a drivers installed along with the latest CAP and still having the issue where the mouse cursor lags in the bottom right corner when at the desktop or anything other then a game (in the time/date area and the show desktop area) I also tested some BFC2 and still having the vsync checkerboard issue there. Whenever I use vsync I get a checkerboard pattern on all the menu options and in game when I spin around.
 
Last edited:
Alright, I'm marking cursor lag as unresolved again. Far too many people are reporting it, even with newer cards

Oddly, I can't get it to show up anymore on 11.5a... so they changed something, but obviously it hasn't worked for everyone.

In my case, 11.5 (non-hotfix) still got laggy.
But disabling cursor shadow helps - everything working correctly.
No mouse laggines in upper right corner

cfg: win7x64, 6950 2GB, eyefinity 3x1
 
So is the 11.4 still the latest CAP drivers out right now?

edit: got the 11.5a drivers installed along with the latest CAP and still having the issue where the mouse cursor lags in the bottom right corner when at the desktop or anything other then a game (in the time/date area and the show desktop area) I also tested some BFC2 and still having the vsync checkerboard issue there. Whenever I use vsync I get a checkerboard pattern on all the menu options and in game when I spin around.

I'd recommend the 11.5a hotfix as those are the ones that are supposed to fix the cursor lag not the 11.5 also I notice 11.4 cap 2 is worse than 11.4 cap 1. Cap 2 breaks crossfire scaling in crysis 2 and also seems to perform worse in applications not mentioned. So it seems to be as far as Cap 2 is concerned that the added profiles work well but cap 2 borks all other profiles I have personally tested. So 11.5a hotfix with 11.4 Cap 1 FTW
 
I'd recommend the 11.5a hotfix as those are the ones that are supposed to fix the cursor lag not the 11.5 also I notice 11.4 cap 2 is worse than 11.4 cap 1. Cap 2 breaks crossfire scaling in crysis 2 and also seems to perform worse in applications not mentioned. So it seems to be as far as Cap 2 is concerned that the added profiles work well but cap 2 borks all other profiles I have personally tested. So 11.5a hotfix with 11.4 Cap 1 FTW

Just checked and I did install the 11.5a HOTFIX drivers. Can't seem to find the 11.4 cap 1 profiles though.
 
Still cursor lag in the top-right corner using the 11.5a hotfix drivers. XFX5870 running Win7 64.
 
I also tested some BFC2 and still having the vsync checkerboard issue there. Whenever I use vsync I get a checkerboard pattern on all the menu options and in game when I spin around.

I have this same problem under BC2 and Dirt 2 (and presumably a handful of other titles that I haven't checked yet) running 11.5a. Only seems to happen when running in Eyefinity, single display eliminates the problem.

Running 2x Gigabyte 5870 2GB running DP to 3x Dell U2410's
 
So what is the best driver for single 6950? I'm still using early 11.4 preview drivers... is it worth it to move to 11.5 or 11.5a hotfix?

Thanks :)
 
Im getting some weird shadow problems with Metro 2033 using the 11.5a rc2 drivers... I read that it was fixed in 11.4 previews but why am I still having it?
 
I'd like to submit a bug... if you want to call it that:

No matter *what* I do, from a FRESH install of WIn 7 using versions 11.3, 11.4 and now 11.5 - CrossFire disable/enabled, 1 card in/out, etc.

I can NOT get L4D or L4D2, OR Star Trek Online to NOT crash upon entering the game.

Both Left 4 Dead and Left 4 Dead 2 drop IMMEDIATELY to the desktop WITHOUT error message, and there is no driver error mentioned anywhere in Event Viewer.

Star Trek Online crashes with an error message and subsequently uploads debug data to Cryptic. I'll post the contents of the error message when I get home tonight.

If ANYONE has any fixes or is having the same trouble, I'd appreciate a response regarding this.

The specs to my rig are accurate in my signature: I'm running Windows 7 Pro Service Pack 1 (64-Bit), 11.5 drivers are currently installed with Crossfire enabled. I ONLY have trouble with these 3 games.
 
Last edited by a moderator:
I'd like to submit a bug... if you want to call it that:

No matter *what* I do, from a FRESH install of WIn 7 using versions 11.3, 11.4 and now 11.5 - CrossFire disable/enabled, 1 card in/out, etc.

I can NOT get L4D or L4D2, OR Star Trek Online to NOT crash upon entering the game.

Both Left 4 Dead and Left 4 Dead 2 drop IMMEDIATELY to the desktop WITHOUT error message, and there is no driver error mentioned anywhere in Event Viewer.

Star Trek Online crashes with an error message and subsequently uploads debug data to Cryptic. I'll post the contents of the error message when I get home tonight.

If ANYONE has any fixes or is having the same trouble, I'd appreciate a response regarding this.

The specs to my rig are accurate in my signature: I'm running Windows 7 Pro Service Pack 1 (64-Bit), 11.5 drivers are currently installed with Crossfire enabled. I ONLY have trouble with these 3 games.

L4D part one at least I know for sure is an application issue. The game seems to not support the scaling to fit into your eyefinity setup by default. To fix that game you need to right click the shortcut file and add - window to the end of the target file. once you do that launch the game and when it loads up set your resolutions in the options.

This very well could be the case for L4D2 but I got it for xmas on Xbox 360 so I didn't buy it for PC. Don't know anything all about star trek online.
 
Getting cursor corruption with my VisionTek HD 6990 with only one monitor at a time on eyefinity at 5760 x 1080. . Running windows 7 64bit SP1, and tried with latest ati drivers (11.4, 11.5, and 11.5a). Reboot resolves it, but after playing a game it will appear and only happen on one monitor at a time.
 
Getting cursor corruption with my VisionTek HD 6990 with only one monitor at a time on eyefinity at 5760 x 1080. . Running windows 7 64bit SP1, and tried with latest ati drivers (11.4, 11.5, and 11.5a). Reboot resolves it, but after playing a game it will appear and only happen on one monitor at a time.

Its a weird thing with the clocks changing... You don't need to reboot everytime, just do something that switches from 2D to 3D mode, like watching a youtube video... It goes away when you load one up...
 
L4D part one at least I know for sure is an application issue. The game seems to not support the scaling to fit into your eyefinity setup by default. To fix that game you need to right click the shortcut file and add - window to the end of the target file. once you do that launch the game and when it loads up set your resolutions in the options.

This very well could be the case for L4D2 but I got it for xmas on Xbox 360 so I didn't buy it for PC. Don't know anything all about star trek online.

Ah, I forgot to mention I'm not doing eyefinity... I have a single LG 26" @ 1920x1200.

Definitely not eyefinity related, and it seems to not matter whether CrossFireX is enabled or disabled... puzzling.
 
Ok, finally got an email back from AMD. Here's what we got:

- Issues with Overdrive and UVD clock have been reported to a software engineer for root cause analysis.
- They were unable to reproduce slow display mode switching.
- They were unable to reproduce mouse lag.

Personally, I'm ready to write-off the "Slow Display Mode Switching" bug. I haven't had it recur for quite some time now (Not since Catalyst 11.3). So lets call that one "fixed" unless we hear some new reports on it.

It troubles me that they weren't able to reproduce mouse lag, though... I'm going to set up another call with them so I can walk them through reproducing it, they probably just missed it.
 
Last edited:
Installed Crossfire, now I get terrible flickering Crysis 2, didn't have the issue with the single 6970. Tried the new 11.5a hotfix, now Mass Effect flickers as well. Tried all the hacks / workarounds I've seen around on the forums (RadeonPro forcing Crossfire profiles, adjusting brightness settings, color profiles...), nothing will work.

It isn't even so much as flickering as a dark layer over the frame. If I move my mouse / view extremely fast, everything looks fine actually - the dark blobs and "overlay" disappears. As I slow down, dark blobs start to appear and as I stop moving the entire frame then becomes dark again. Mass Effect 1 I don't have the darkness, just a constant and very annoying flickering.
 
Ah, I forgot to mention I'm not doing eyefinity... I have a single LG 26" @ 1920x1200.

Definitely not eyefinity related, and it seems to not matter whether CrossFireX is enabled or disabled... puzzling.

I was able to play l4d part one no problem with 11.4. Not sure what's up with your setup then? Are your cards overclocked?
 
I was able to play l4d part one no problem with 11.4. Not sure what's up with your setup then? Are your cards overclocked?

No sir... bone stock XFX 6970's in CrossFireX on a Rampage 3 Extreme - all system clocks are bone stock.
 
No sir... bone stock XFX 6970's in CrossFireX on a Rampage 3 Extreme - all system clocks are bone stock.

Strange, I mean I just played the game and uninstalled it about a month ago. I finished the new sacrifice campaign on the 11.4 drivers

I have 2 unlocked/very overclocked 6950s in crossfire x
i7 2600k @ 5Ghz
Win 7 Ult x64

I used the steam version of the game. Have you tried doing what I suggested before despite the fact that your not using eyefinity?

Go to the shortcut .exe file the game installs, right click it and go to the target tab. You'll see a long path there, e.g. c:/programfilesx86/valve/hl2 etc.. highlight but don't erase the target. leave a space and type in - window at the end of it so afterwards it'll say something like

c:/programfilesx86/valve/hl2 - window

Afterwards click apply and then try and re-launch the game.
 
I did not, however I'm willing to try anything... I miss playing Versus, heh.

Doesn't explain STO though, that's DAMN weird.
 
I'm having some issues related to Windows 7 event viewer that could slow down heavily boot/shutdown performances. Fortunately there are temporary solutions found in these threads:

http://forums.amd.com/game/messageview.cfm?catid=279&threadid=141688 (still present since 10.8 - I can reproduce it with 11.4 drivers)
http://forums.guru3d.com/showthread.php?t=343572 (full explanation in the first post with all details)

IMHO it is necessary to update the first post of this thread, to recall AMD attention when major driver problems as Powerplay downclocking and mouse lag will be officially solved.
 
Last edited:
Anyone playing Quake Live? After updating to 11.5 I kept getting stuttering ingame. Went back to 11.4 and works fine now.
 
I'm having some issues related to Windows 7 event viewer that could slow down heavily boot/shutdown performances. Fortunately there are temporary solutions found in these threads:

http://forums.amd.com/game/messageview.cfm?catid=279&threadid=141688 (still present since 10.8 - I can reproduce it with 11.4 drivers)
http://forums.guru3d.com/showthread.php?t=343572 (full explanation in the first post with all details)

IMHO it is necessary to update the first post of this thread, to recall AMD attention when major driver problems as Powerplay downclocking and mouse lag will be officially solved.
Bug confirmed, added to original post.
 
Bug confirmed, added to original post.

Sorry but the 'UVD information' spam in event viewer is not minimally connected with the 'DCOM error' spam in event viewer. In my previous post I reported two different bugs.

I checked carefully this thing because the only workaround for 'UVD information' is the following registry key fix:

Code:
[HKEY_LOCAL_MACHINE\SYSTEM\CurrentControlSet\Services\Atierecord]
"eRecordEnable"=dword:00000000

About 'DCOM error' workaround it is sufficient to set 'AMD External Events Utility' as manual, instead to disable it totally.

The 'UVD information' spam comes when using some sort of video like youtube, while the 'DCOM error' appears to only show up during shutdown process.
 
Last edited:
There seems to be a GPU usage bug present in select races in DiRT 3. On certain tracks the GPU usage on my cards (sig rig) drops from 95-100% per card to 50-55%. First good example is the second race on the DiRT Tour (American RallyX Tour, Smelter Houghton Sprint). As a result the frame rate drops quite a lot from maintaining an average of 65 to around 35.

This game is also effected by the Crossfire Eyefinity checker boarding problem while vsync is enabled this issue appears to be limited to DX11 games.

Edit: I should mention I'm running 11.5a RC2 with 11.5 CAP1
 
Last edited:
Thanks for clearing that up weah, modified the my first post.

Also, another day, another email back from AMD... and I don't like what I'm hearing. This is in reference to "2. PowerPlay Downclocking"
Regarding the UVD, it is the expected behaviour that our driver will use UVD status of 400/900 MHz while video playback even if a 3D application is runnung.

If you stop HD playback and run the 3DMark individually then the High Performace values set in the AMD Overdrive will kick in.
I've asked him to clarify what "expected behavior" means for us. Hopefully it means "we're confirming that it's reproducible" and not "this is the way we intended for our cards to behave"

If it's the latter...I'll be a little speechless, quite frankly. That would mean they intend for the card to get stuck below 2D idle speed, they intend for 3D performance to suffer greatly from playing a video, and they intended for displays to flicker (which they do on a lot of cards when memclock is adjusted). If that's REALLY what they're saying, then this HD 5850 will be my first and last AMD card.
 
Last edited:
Thanks for clearing that up weah, modified the my first post.

Also, another day, another email back from AMD... and I don't like what I'm hearing. This is in reference to "2. PowerPlay Downclocking"
I've asked him to clarify what "expected behavior" means for us. Hopefully it means "we're confirming that it's reproducible" and not "this is the way we intended for our cards to behave"

If it's the latter...I'll be a little speechless, quite frankly. That would mean they intend for the card to get stuck below 2D idle speed, they intend for 3D performance to suffer greatly from playing a video, and they intended for displays to flicker (which they do on a lot of cards when memclock is adjusted). If that's REALLY what they're saying, then this HD 5850 will be my first and last AMD card.

in terms of software development / testing / fixing. Expected behavior means as designed. So that is the way it is supposed to work.

I'm not sure why you are speechless :), do you really watch youtube while you play a game? at least not in full screen, or run any other flash / uvd app.

As far as mem causing screen flicker, that applies to ALL clock states then, even though I have never seen it. 2d clocks are 150/400? so what is so bad about 400/900 for UVD that would make it flicker, but not 825/1275 for 3D(for example).
 
Back
Top