Radeon Crimson ReLive Edition 17.8.1 - Download and Discussion

Is it possible that I'm not affected by UVD bug on my R9 270X? I'm on 17.8.1, but same was with previous drivers. When I'm in game, it shows GPU max clocks (core: 1100, memory: 1450), but when I'm on desktop it's 300/150. Sometimes it shows 450/1450, like when I'm opening Chrome, but it reverts back to 300/150 after it opens. Does that mean I don't have UVD bug or what?

 

tested a few matches of Overwatch running 290x crossfire in AFR friendly mode. noticed a snappiness to windows 10 again. flicker is gone in Overwatch menus! so far best driver in awhile for me. I have the samsung CF791 monitor and the flickering seems improved as well. want to test more in 8 bit color/extreme freesync/DP 1.2 before confirming they resolved all the issues.

 

UVD bug must be triggered to happen, it doesn't kick right after you start Windows and depending on your usage, you might not even experience it on affected hw. This is what works for my 7870 (and did on a previous 7970 and a friend's 280X) but the "recipe" might vary for some users: Open gpu-z to see vram occupancy and clocks, under Chrome scroll down on a page like the FB wall for some time until vram keeps filling above >200MB and then clocks will be idling at 450/1450 without reverting to 300/150 after you stop and wait some seconds to confirm. At this point (if clocks are at 300/150 it'll fail to trigger the bug), launch a .264 video with mpc-hc and let it play for a while (probably something in hd under youtube will work too, provided its using a hw accelerated codec). This should trigger the UVD bug if after closing both Chrome and mpc-hc, clocks will be stuck at 450/1450 on idle, and then will only raise until the highest UVD clock for your card under 3D (and also ignore any soft OC set over the 3D clock).

 

Thats a shame

 

I have the opposite problem, that the clocks will be in low power state while watching a youtube video, and then doing something else at the same time will cause a major lag and cause tdr or bsod. using clock blocker and locking clock to full speed fixes the problem.

 

where is this option now??

 

The problem I mentioned was in CGN 1.0 cards. I guess reports with lower UVD clocks still stuck on 3D as you face seem something more recent with CGN 1.1/1.1+ cards and introduced in later drivers this year (I've never experienced it on a 7870 nor RX480).

 

on a samsung 24inch cfg70 using 144mhz ultimate engine @ rx470, freesync is a bightness strobe mess, can be reproduced in game but its very clear with the AMD windmill freesync demo, just for you guys can test with a samsung monitor.

 

Did what you said, UVD bug didn't trigger for me. Well, I don't know did I watch a HW accelerated videos, but tried with like 10 different videos on youtube, with facebook open and few more pages. Vram usage in GPU-Z was like 300 MB or so, and when I closed all of those, it reverted back to 300/150. Maybe I'm just lucky, or I wasn't able to trigger UVD bug. Don't know really.

 

Interesting... I've just tried it too with .264 accelerated YT clips started with the card already at 450/1200 (from FB) but the bug didn't trigger. OTOH, launching a clip from mpc-hc while already at 450/1200 from chrome browsing did it without fail.

 

What renderer are you using with MPC-HC?

 

Inbuilt EVR (CP) or Sync Renderer, nothing exotic or 3rd party. Even VRM9 (renderless) will trigger it and it doesn't even support DXVA. I can reproduced it another completely different system (p55 based) with a 280X under W10 x64 too.

Out of this OT, the only very strange thing I've found with the 17.8.1 is that on the gaming tab none of my installed games were showing (had to manually rescan).

 

My brother got FS monitor. It was occasionally going dark and starting again on random. That has been fixed with newer drivers.
But when I came to visit I took my short DP cable just in case. And 1st thing I noticed was minor changes in lightness of screen per frame (like backlight not keeping same value and changing 5% per frame).
He did not notice at 1st and took him a while on certain level of brightness to notice. It has been fixed by replacing his 2m cable for my ~80cm.

 

Do I understand it correctly that if I change voltage for any of the power states, the values actually remain on default?

 

Yep, atleast it's that way for me. I have a modified BIOS with changed RAM timings but that shouldn't be a problem.

 

OMG I've just noticed it. It's enought to simply Alt-Tab out of Wattman and the values are reset to the default ones. What the frack AMD?

 

So this driver is UVD bug free for CGN 1.0 cards ?

 

Not on my 7870 DCII v2 and a friend's 280x DCII using the chrome+mpc-hc method. OTH, YT alone vs other tabs doesn't seem to trigger it so easily as it did in the past, but that might not be completely driver related.

 

Well, I don't know. I wasn't been able to trigger it, so, dunno. Maybe I didn't do the right thing to trigger it. I have 17.8.2 now, working fine here too.