On 18.3.4 driver under windows 7 64 bit the memory speed of my GPU never returns back to 150Mhz from 1400Mhz once it hits full speed. Also the core never returns to 300Mhz but drops to what I think is the 2nd state of 450Mhz. This increases the idle temp by 10degC. Most drivers in the 17.xx series had the same problem. I am essentially stuck with very old drivers of early 16.xx or mid to late 15.xx
It also seems to ignore the global overdrive settings if I reduce the maximum clock speed and memory to keep the gpu cooler and quiet under gaming at the sacrifice of some speed. Essentially Global overdrive has been bugged for almost 12 months now on my card.
Report attached using the wizard.
Solved! Go to Solution.
If you look around, that's what most call the CGN 1.0 UVD bug (because your card will be locked into high/low UVD states instead of regular idle and highest 3D clock, causing soft OC to fail too as it's applied on the 3D clock state that is no longer available), and has been present in drivers after 16.7.2. It can kick often after playing video content in your browser, but there are other ways to trigger it. We've been said that the bug was acknowledged, but it was never solved.
To avoid/workaround it, you can either stay in 16.7.2, disable hw acceleration on a browser, or edit bios clocks to match highest UVD clocks to boosted 3D clock so you don't suffer performance loss in games.
One way to cure the problem without a reboot, is to download CRU and run its restart64.exe component, this will reinitialize the driver. Then you'll need to re-run your OC profile to apply it again.
hmm weird what things have you tried tweaking the wattman settings.
Well it does not enable and use the new wattman. Some older r9 200 cards like the 270 are still not supported by wattman so it uses the previous overdrive tool tab. So under overdrive you can alter core/mem/power. Makes no difference what these are tbh. Reboot the pc and the clocks go back to low settings but sometimes it gets stuck at high. I have read on the forums others have had clock stuck issues on newer r9 cards. I would buy a new up to date card but not atm due to prices almost 2x original cost due to mining craze.
Update : I disabled 2 scheduled amd tasks that got installed with the driver, drvcmd.exe and cncmd.exe, not sure what these do. I did not install relive. rebooted and the clocks seem to drop back to normal on idle. I will keep a check on them.
If you look around, that's what most call the CGN 1.0 UVD bug (because your card will be locked into high/low UVD states instead of regular idle and highest 3D clock, causing soft OC to fail too as it's applied on the 3D clock state that is no longer available), and has been present in drivers after 16.7.2. It can kick often after playing video content in your browser, but there are other ways to trigger it. We've been said that the bug was acknowledged, but it was never solved.
To avoid/workaround it, you can either stay in 16.7.2, disable hw acceleration on a browser, or edit bios clocks to match highest UVD clocks to boosted 3D clock so you don't suffer performance loss in games.
One way to cure the problem without a reboot, is to download CRU and run its restart64.exe component, this will reinitialize the driver. Then you'll need to re-run your OC profile to apply it again.
Thankyou for that info. I never had the issue in Crimson Edition 16.11.5 Hotfix (last driver before relive editions) which was the one I used for sometime so I guess it is a bit random in the 16.xx branch but I have been having other issues of crashing in a few games so was trying to eliminate the driver by using newer versions in 17.xx and 18.xx but all with the same bug and some with others much worse.
And yes, a few I tried in the 17.xx branch also had the clocks issue and one 17.xx version even locked/crashed the pc in ableton live 10 application which has never happened. I might even end up going all the way back to 15.7.1 which from distant memory seemed very stable and I don't remember having any issues at all.
A lot of the issues are coming from your crap windows 7 throwing issues with the 200 series cards & 300 series bug exists from 7. The bug actually goes away when running 10 as know a couple m8s with simular issues forced them to 10 was the best & cheapest fix & cards run like butter on newest drivers. Its not drivers at fault its 7 being so outdated & getting to the point AMD will be dropping support as well as Microsoft coz of bugs & issues. You work arounds are quick fixes on old drivers which wont help with newer tasks & gaming etc, your better of to bite the bullet upgrade you OS & get the most out of the newest batch of drivers & wattman works & runs way better with the 200 series cards & is supported as the Newest drivers are supported with 200 series cards still not legacy yet.
Windows 10 is a non starter for me atm. My pc is also an audio workstation (DAW) and windows 10 is a poor option when you need low latency with zero background hogs. I have read too many users having issues after upgrading when using it for pro-audio work. I would maybe consider w10 Enterprise since that is more streamlined, less bloat and zero telemetry / background hog but that is not available to buy as a single user so I am kind of stuck with 7 for now.
I use Solus linux as my main OS so win7 is purely a DAW and the odd game. I would just go back to an old driver if need be..
This bug seems to affect some win10 users also, I doubt it is purely a win7 only issue...
Upgrading to W10 won't solve it, unfortunately. I've had it on all my CGN 1.0 cards in different systems, same with a friend on a R9 280X. Really, if you don't need the latest driver, just stick to 16.7.2. It even supports vulkan.
2 years have passed, haven't corrected
Update. So far the bug has not reproduced itself since I changed (lowered) the max core and memory speed slightly. Exiting firefox or a game, it changes back to 2d idle state OK. I will keep an eye on it to see if it gets stuck again but so far it is behaving.
Update info to hep others with the same problem (May 1st 2018). So far the speed bug has not re-occurred but I only play 2 games. Running on 18.3.4 with a r9 270 and it seems stable so might be worth a try if you are having these clock stuck issues of the past driver versions and no you do not need Win10 as suggested. Running win7 x64 dual booting with Linux.
I had a similar kind of problem. I have 2x290 and 3 monitors.
After doing all kinds of testing, I found out that one of my display's refresh rate was at 59Hz instead of 60Hz. Putting it back to 60 fixed my problem. If one has multiple displays, check out every display's refresh rate...