Thanks for your feedback!
I've never noticed in that condition because I run windows with both Gamebar and Fullscreen Optimizations forced (via registry) off, as the new hybrid fullscreen mode can be problematic.
Beware that there are two different scenarios were UVD related problems can happen on CGN 1.0:
1) The one I mentioned above leaves you completely stuck at UVD clocks even without anything running (450/1200 on desktop and the highest UVD clock for 3D*) , and only re-initializing the driver via reboot or CRU's restart64.exe will make it return to regular clocks. This is the one that will happen on any driver after 16.7.2.
2) The other one, I assume it's expected behavior unlike the previous, but can also be problematic and hinder performance. There are cases where games or software use intros/elements that trigger the UVD states, and then when a 3D scene is started, if the previous UVD accelerated instance hasn't been properly finished in time, the card will then run at the highest uvd clock for 3D instead of using the correct 3D state. This is problematic on boost cards with lower high uvd clock than 3D clock. Usually these cards will only change to the proper 3D clock when a 3D scene starts and the card is resting at 300/150 instead of UVD clocks.
One very easy way to check it, is running a light clip in MPC-HC that makes the card go to 450/1200 while playing (just make it a small window), and then at the same time, run the 3D render test on GPU-Z (the "?" next to the pci-e bus speed). The card will jump to the highest UVD clock instead of the correct 3D clock state*.
But unlike in the first problem, after all software instances are closed, the card will go back to proper 300/150 clocks on idle. This problem will happen on any driver, including 16.7.2 and before, which are free from the first.
* On a older non-boost card were 3D and highest UVD clock are the similar, applying a 3D overclock is a easy way to distinguish between, because it isn't applied to the UVD state.
HI , i might found a quick fix you guys , at least it works for me , i made an account special to tell you this fix as i searched for 1 in a long time.
What i did its very simple , ive downloaded RadeonMod and on the Tweaks Tab i disabled ULPS then pressed restart drivers. (the last option there is UVD Clock FIX , and its on by default if its not just enable it!). Now i can be on youtube and twitch and play a game at the same time and the card uses the boost clock. I have an Asus r9 280x , before doing this if i had youtube/twitch or anything that used UVD and i was starting any game the clock was sitting at 970 now is sitting at 1070 as it should. Hope this will help you guys too. Best Regards.
Disabling UPLS isn't a solution for the problem I've listed above as "1)", which is the perma stuck UVD clocks. It might be helping you with "2)" (uvd to correct 3d clock transitions) under specific circumstances, but I've just tested with mpc-hc playing and then start gpu-z render test, and fails to apply the proper 3D clock.
In the past, the best way to solve "2)" was to use OC apps that applied clocks in "non-official mode with power play support" such as MSI AB and Trixx before it was removed from drivers. By that way, 3D clocks could be started from a UVD state, not requiring the card to be at 300/150.
How about problem with GCN 1.0 UVD clock&sleep? If you watch video in browser and send PC to sleep, after wake up videocard stuck on UVD clock, until you reboot PC. The last driver without mistake 16.7.2. Sorry my english is bad.
Hello, i just have the same problem but i was not aware what it was before some research, for me the bug is on since computer is turned on.
I got an Asus strix r7 370, with factory OC to 1050mhz but I was kinda worried about my idle temps being so high (56ºc) so I changed the thermal paste and saw no change at all, I've realized that on idle was very hot and on load it won't even do the factory OC, and was only at 925mhz (the stock clocks) and it supposed to run at 1050mhz, then I downloaded the GPU tweak II from the asus page and saw that even won't overclock at all.
After some days on google, I've used DDU to reinstall the drivers (I've tried with the last 3 driver versions available to date) and tried disabling ULPS and nothing, I saw a video about editing the bios so the UVD state has the same clocks as the performance state but i don't want to do that, I thought this was fixed and yet many people on the internet noticed this and the RX series has the same problem too and i guess some people has the bug going but won't even noticed because they don't know about computers.
for me bug is on since startup (win10 up to date)
ASUS STRIX R7 370 (is overclockable)
Load (1050mhz core / 1400mhz vram)
Idle (300mhz core / 150mhz vram)
Temps (idle 35ºC / Load 69ºC)
On the GPU Tweak II there is an option called "Overclocking range enhancement" that do what it's name says but has to disconnect/reconnect the gpu, that allows to the gpu be back to the normal operation on which drops the temps significantly.
ASUS STRIX R7 370 (is NOT overclockable, you can move the settings but it won't change anything on the HW Monitor)
Load (925mhz core / 1400mhz vram)
Idle (450mhz core / 1400 mhz vram)
Temps (Idle 56ºC / load 66ºC)
I took a peek on the bios editor and downloaded the bios knowing what's going on and yes those are my UVD settings and not the performance clocks.
Temps and power are a matter that concerns me, since the normal operation only reach 35ºC and with this bug and from the start reach 56ºC at idle!, I had to set a custom fan curve just to keep it at 50ºc idle all the time.
So please I hope this could get fix soon, I bought this card because the RX series is so expensive on my country right now, and knowing the bug affects the RX too I have to check on my friends to see if they had the bug too and are unaware of it, so i write here to let known to AMD people that there is many more people with this bug than you may think.
If you wan't an immediate solution without frequent manual driver resetting or bios editing, your only option ATM is to use 16.7.2 drivers.
Thanks! so... this is a driver fault then? do you think if do a clean reinstall of windows could I get rid of the bug?, anyway I'll install the 16.7.2 thanks again.
Don't bother reinstalling windows, providing you haven't something running in the background that causes those clocks (so then it's a different problem related to "2)", as I explained on a previous post on a post), this issue ("1)") will happen on any driver after the 16.7.2 set. Just test them and see what happens.
I have an msi r9 280 in my bedroom pc
and can confirm that the uvd bug has been present since 2016
how amd say they cant reproduce it is beyond me
its simple to reproduce
they clearly arnt using the right card or are not monitoring clocks properly