I'll paste back what I've wrote on guru3D about the fixed notes on the driver:
Graphics and memory clock speeds may remain elevated or locked while gaming if video content is also playing on the system."
Well, I can't reproduce the CGN 1.0 bug anymore in the same way I've always consistently triggered it. This driver seems to have finally fixed it.
The behavior is back to pre 16.7.2, but this also means that the driver fixes statement above doesn't apply for these cards. On a CGN 1.0 HD7870 (pre OC'd Sapphire at 1050/1200 3D on stock vbios), while playing a video that causes clocks to switch to 450/1200 or 1050/1200 (both UVD modes), or by letting chrome cause the card to idle at 450/1200, then initiating 3D content such as the GPU-Z render test will cause the card to go to 1050/1200, but that is the highest UVD clock, not the 3D clock state (same stock values on this card).
This means that any OC applied via software over the 3D clock won't be used because the card is running in the upper UVD state instead. The same will happen by launching a video on mpc-hc, leaving it playing and launching a game in (true) FS mode -- OC won't be applied because the card is in the highest UVD mode, not 3D. Can't comment on a boost card, since I haven't any to test.
Just to add, since I can't edit. The driver solves issue "1)" (uvd clocks stuck until reboot) but not "2)" (card won't raise to highest 3D power state when starting a 3D app while low/high UVD is active and other content is still using it) that I listed in detail a few posts back in this thread. But 2) affected these cards since the beginning, and the only way to workaround it, was to use the unnoficial OC mode with powerplay support that was available in MSI AB, but drivers don't have that OC path available anymore since about a year ago. Thanks.
No.2. If iirc, not reaching full clock on the boost was expected behaviour if another app(e.g. video clip) was using graphics. Sapphire(at the time) had a small article explaining the bios boost function(can no longer find it).
The same bios boost applied to some rebadged 2xx models, I remember thinking 'why buy if boost is limited under certain conditions'.
The stuck clock bug may be fixed but
the clocks arnt
Play AVC video on Media Player Classic on monitor 2
Start Game on monitor 1
Results in the R9 280 running at UVD speeds even though a game is playing
e.g. 827mhz instead of 1000mhz
can we expect a fix for this quicker than 2 years?
Yes, this issue "2)" has been in CGN 1.0 cards since as far as I remember (late 2013', since didn't have one before).
They simply won't switch to 3D clocks if a 3D app is started at the same time something is already making the card use UVD clocks, 3D clocks will only be correctly applied if the card is resting at 300/150. This was tolerable with the first gen non-boost cards running stock, because the highest UVD clock is equal to the 3D clock as in my example above. The nuisance would be apparent when using a OC over the "true" 3D clock, as it would be ignored.
But it becomes a problem on boost cards where the highest UVD clock is considerably lower than the boosted 3D clock, like on your 280.
The driver fixes note seems to address this issue in the way it's phrased, but it only solves the stuck UVD until reboot problem (still, MUCH appreciated!).