Not everyone is comfortable flashing their GPU BIOSes, and I don't think this is something the end users should have to deal with in the first place as it's the responsibility of the IHS drivers itself. Not everyone is comfortable having to use CRU's Restart64.exe 10 times a day just to reset the clocks when they get stuck. It's this train of thought is what made nvidia get away with scandals like the 3.5 GB of VRAM on the GTX 970 and the promised Asynchronous Shader support on the Maxwell series of cards which never came and now is disabled from within their drivers simply because users were "okay" with it.

That UVD bug looks like is more visible on old 79xx series than "new" 200 series. Maybe they change something in driver to make you upgrade,kidding. 😀 Anyway,Firefox or Chrome + hardware acceleration should not be used.Also uninstall useless Flash and java,because security reason.

I've explained a few times in previous threads

not enough, mr. Chrome/Chromium doesn't like Radeon GPUs even on Linux :bang: what about trying if the bug is still there with A DIFFERENT (and overall better 🤓 ) browser and video player? e.g. Pale Moon (install Adblock Edge + Ghostery 5.x as main extensions 😉 ) + Potplayer ?

Guys, the GPUs are old, but they are still supported and this should be fixed. Also flashing your bios won't solve the issue that makes the card never to use the proper 2D clock and always burn more power. My card is flashed, but I still want it fixed.

Maybe you don't care, but the rest of us do. We clearly aren't overreacting. You flashed a custom BIOS that sets the UVD state 2 to your 24/7 OC clocks, but the rest of us who're using a stock BIOS can't apply OC clocks once the bug shows up. Since it doesn't bother you, kudos to you. But don't tell others not to complain because they have a right to. And that's the purpose of this forum to begin with.

once you run 3dmark and see that clocks are stuck and card is not boosting to it's proper clocks, you are more comfortable with that then adjusting card's bios? Hmmm.... interesting. Flashing the bios was the first thing I did when I got my 280x...Albeit, being AMD's fault, it doesn't mean I should wait on them to fix it, since it's been like...what... 3-4 years now? Keep being uncomfortable with bios flashing if you want, but that's the only way to fix the card(s).

Still getting major DX9 issues in multiple games even after the 27/10 Windows patch and 16.10.3... Over-riding with old driver files doesn't work either :/

once you run 3dmark and see that clocks are stuck and card is not boosting to it's proper clocks, you are more comfortable with that then adjusting card's bios? Hmmm.... interesting. Flashing the bios was the first thing I did when I got my 280x...Albeit, being AMD's fault, it doesn't mean I should wait on them to fix it, since it's been like...what... 3-4 years now? Keep being uncomfortable with bios flashing if you want, but that's the only way to fix the card(s).

The Bios flash doesn't solve the clocks getting stuck at UVD State 01.

Has anyone thought of trying a 3xx vBIOS on their 2xx card and see if the bug goes away?

Anyway,Firefox or Chrome + hardware acceleration should not be used.Also uninstall useless Flash and java,because security reason.

On older CPUs you simply cant play a 60fps YT video without awful skipping, not to mention that color space will be wrong and colors washed out. For a matter of comfort with the bug, it's more practical to leave it disable instead of restarting the driver most times before launching a game as I resort to, but it has shortcomings like that one.

Tahiti is ancient at this point, you guys are overreacting. Who cares if idle clocks are stuck, performance clocks works just fine and thats what matters. My card is boosting to 1165mhz without any problems and im gaming every day on this system.

As others commented, it's different when you manually flashed UVD #2 clocks/voltage to match your desired OC, but the UVD #1 will still be too high for idle. Something similar can be done via AB in unofficial OC mode with PowerPlay concerning UVD #2 clocks (Edit: Not at all for a 280X, that's only for non boost cards where 3D and the highest UVD state are both #0 on the vbios). Btw, as UVD #2 is different (as a state) to a manually applied OC over the highest 3D clock, unless you also change base TDP on bios, +20% won't be applied but I guess you have covered that on VBE7.

Chrome/Chromium doesn't like Radeon GPUs even on Linux :bang: what about trying if the bug is still there with A DIFFERENT (and overall better 🤓 ) browser and video player? e.g. Pale Moon (install Adblock Edge + Ghostery 5.x as main extensions 😉 ) + Potplayer ?

My post was a direct response to @passenger for a way to trigger the problem step by step with a common software combination, not a question about software alternatives as there can be many other conditions that trigger it. Thank you.

once you run 3dmark and see that clocks are stuck and card is not boosting to it's proper clocks, you are more comfortable with that then adjusting card's bios? Hmmm.... interesting.

At the time I've seen that on a 7970 (I rarely have 3dmark on my system) before a patch fixed it, it was exactly the same condition that happened when Chrome sets clock at UVD #1 in my step "1" example some posts back, due to video playing on 3DMark's GUI, only UVD #2 is used for 3D. Still, it was possible to bench with OC applied simply by setting MSI AB to unnoficial OC mode, no need for a bios flash. Also, let me be clear: Using AB in "unofficial overclocking mode with PowerPlay support" mode to set an OC will also allow the card to use manually defined 3D/OC clocks while the bug happens, as they seem to affect UVD #2, because my non-boost 7870/7970 will also use those OC clocks when playing video in a non-bug situation (unlike official mode). At least it seems to work currently with AB 4.3.0 and 16.10.3. But there's a chance that "+20%" will not be applied, I guess 7950 users will find it easier to test for throttling. The problem that remains are idle clocks, as with a bios flash. Edit: I've had a friend to test it on a 280X, but it doesn't seem to work, as UVD#2 isn't replaced by the AB OC (under or outside bug). So it might only work on older non-boost cards as the Asus 7870 DC2 V2 I'm testing in. On these non-boost cards, the highest UVD clock and 3D clock are both listed as #0 in VB7, unlike boost cards that use separate #0 and #6 ids.

Has anyone thought of trying a 3xx vBIOS on their 2xx card and see if the bug goes away?

I guess only the R7 370 would be eligible over a R9 265, as probably older 7850 would have unsupported memory chips.

No problem here on Hawaii.

No problem here on Hawaii.

If it's in response to the UVD bug, it only affects Tahiti/Pitcairn on some systems. 🙂

Omega has been released at 11.14, Crimsons - 11.15. Should we expected another big update at 11.16? AMD bought HiAlgo's technology a few months ago so They could do really good stuff.

It's circumstance where the card gets locked at UVD clocks, and only uses the lowest UVD in idle (without anything 3D accelerated running), and only its highest UVD clock for higher loads such as 3D (meaning no boost clock for models that have it). It can be triggered for some users while having something running that sets the card in UVD clocks such as a browser and then playing a video with hw accel. After closing everything, the card will be stuck in UVD mode. This probably means idle at 450/1200 on your 7870 and 501/1250 on your HD7950 instead of 300/150 which can mean >30W, and no applied OC via soft (if they are non-boost models, the highest UVD clock will match the stock 3D one).

I have this issue with my R7 265, and until i read this i thought it was my VGA dying. So it is a driver problem? ok. Sometimes my video card stays at 450/1400 clock even when the computer is in idle. And when i enter a game it goes up to 900/1400 but does not reach card boost clock that is 925/1400. Also if i try using Afterburner to overclock it, vga clocks stays at 900/1400 no matter that I apply overclock using MSI afterburner. I have to turn off the computer (sometimes more than once, does not goes back to normal 300/150 clocks at the first time sometimes). I tried using windows power saving options to "balanced" instead of "maximum performance" but no result.

Yes, that's exactly the problem. Your R7 265 is a CGN 1.0 Pitcairn such as the 7870 in my example. It'll ignore your OC because AB applies it to your boost clock id#, but on load it'll be stuck using the highest UVD clock instead (different id on vbios). Using 16.7.2 shouldn't give you that problem, unless there's something already loading the GPU (or using 2 displays on desktop use, resulting in higher idle clocks). If after a reboot you aren't getting back to 300/150, it might be due to Windows Fast Boot on (tbh, I find it strange). Other than that, you have the option to use CRU and run restart64.exe to reset the driver. Then you'll need to relaunch hwmonitor if monitoring with it, as it'll get stuck reporting the previous clock before the driver went down. Disabling HW acceleration on your browser or flash might help too. A more permanent fix is to edit your bios to set your permanent OC on both the boost clock and highest UVD clock so 900/1400 won't be used when stuck in UVD mode. You'll still be stuck at 450/1400 on idle, tough. Edit: @geko95gek for the post bellow: First time someone reports with a Thames GPU (same as 7970m), but as with Pitcairn and Tahiti, it's another CGN 1.0 card so it's seems to affect the whole family as Romulus_ut3 pointed. That really sucks on a mobile GPU for autonomy.

Ok, weird thing. The UVD clock bug is back for me in 16.10.3. However, in 16.10.2 I dont have this bug. So, so weird?! Is AMD trying to test us, to see if people would notice lol

The UVD clock bug only affects GCN 1.0 it seems. I have it. I watch a couple of Youtube videos, close the browser, and it's there.

Im lucker, because this bug is not annoying for me since 16.9.2. When im using Mozilla Firefox to surf on youtube and run the game, clocks is stuck on UVD, but if i close Firefox (game can still running), clocks back to normal. The Fastest way to instant fix clocks without restart PC is restart Video driver. You can force it in RadeonMod or restart.exe delivered with CRU.

Omega has been released at 11.14, Crimsons - 11.15. Should we expected another big update at 11.16? AMD bought HiAlgo's technology a few months ago so They could do really good stuff.

Thats i am waiting for 🙂

Ok i go for total reinstall (clean approach) Uninstalled AMD/ATI Drivers by AMD Manager then -> Display Driver Uninstaller (DDU) V17.0.2.0 in Safe Mode and Uninstalled leftovers. In CRU 1.2.99-p2 folder -> reset-all Installed new 16.10.3 Beta, then imported my Gamma Profiles etc. and Imported my CRU Profile with FreeSync. Hmm, every Game runs OK And most important -> YES Mantle is acually working NP 🙂 Now i can Play BF:4/H | DA:I | Thief and Civ BE in my Favorite API 🤓 Sometimes we need to Refresh Drivers, it solved my Mantle Problem in any Game....

I can confirm 100% that the Linux open source driver doesn't have the bug.