Hello, I have had a MBA 7900XT from Powercolor for the past few weeks now and been doing a lot of testing. Specs:
- 5800X3D
- Crosshair VIII Impact
- 32GB Trident Z @ 3733CL14 1.53V
- 7900XT
- HX1000
- M32Q 165Hz + S2721DGF 60Hz/165Hz (see below)
These past couple of days I've been watching the data during Youtube playback.
Regardless of the resolution of the video (1080p/1440p/4K), only when viewing a video in the default (small) player mode, it seems to cause crazy fluctuations in both core and memory clock, and consequently board power as well. As soon as the video goes to fullscreen (give it a couple seconds for the GPU to adjust), clocks calm down.
I have data for both 23.1.1 and 23.1.2. In all 4 graphs data is taken over the span of 60 seconds. The polling rate is 2sec for 23.1.1 results and 0.5sec for 23.1.2 but it does not affect the observations or conclusions.
- 23.1.1 is most pronounced, because with 165Hz+60Hz the VRAM downclocks properly and I can idle at the same power as single 165Hz monitor (~15-20W).
- 23.1.2 is more subtle, because VRAM is always 2487MHz unless I set both monitors to 60Hz, but core clock still shows the same oscillations so there is still a slight effect on power. Yes, I've already tried tweaking blanking in both Radeon Software and CRU at all refresh rates, it doesn't get better on 23.1.2.
AMD currently recognizes one unfixed issue in particular on RX7000 series on 23.1.1 and 23.1.2, "video stuttering or performance drop may be observed during gameplay plus video playback with some extended display configuration." I constantly experience this when trying to play a Youtube video on my secondary monitor while in-game (COD MW2 is most affected). The solution for me is to keep my secondary monitor clear of any content especially those with changing visual elements, to avoid incessant stuttering and routine frame drops.
As we all know, Radeon video playback power consumption is also high even when normal (40-50W). But clearly, this only applies to Youtube playback when the player is NOT fullscreen - as soon as the video is maximized, clocks drop and power dips to just above idle power (~30W).
Maybe a bit of a stretch, but possibly the clocks fluctuations for no good reason are causing both the high video playback power and the gaming-while-Youtube stuttering?
This issue remains in both single- and multi-monitor setup.
Anyone else observed this on their Navi31 card?
The data:
