AMD Software: Adrenalin Edition 23.8.2 - Driver Download and Discussion

That is not how it works in real world...
Every word can turn back onto you, so you never admit something first (unless it is either not dangerous or wasn't publicly discovered).
Otherwise they risk getting lawsuits from bunch of people, which won't be nice outcome.

 

Well there were some fixes that actually worked for some users so I doubt it is a hardware issue. But as said, they can't test every combination. I like that they work on fixes and while I am waiting for it to be fixed on my end I will just enjoy the card.
Also just went full AMD last month and I really enjoy it, no big issues that would make me return any parts. Only the high idle power consumption is quite annoying

 

It's been like this since the 5700XT launched. Or maybe before that but idk because from 2013 until 2019 I used Nvidia.
High idle power draw was the second thing that shocked me when I switched from Nvidia.(first thing being the absolutely garbage drivers when 5700XT launched lol)
I guess they don't care because almost nobody bothers them about it. If all people where like me they would have a full inbox only with emails asking them to fix this sh#t once and for all.

 

Yahh i know, sadly

Same, the card is awesome, as i said dont take my rant too serious it's just bit frustration i tried to lower the idle consumption as my room has no AC and i tried to heat it up at little as possible during summer

I guess as they had it in their known issues list etc that they will take care next-gen but yahh usually i dont care that much too was just during summer heat now

 

I am happy customer mails aren't forwarded to me as a dev at work haha

 

Well heat is an issue for me too, but in the end you won't probably notice a big difference with and without the PC on/being warm. In the end I am happy that they at least got good drivers now that work for my card

 

There is part software part hardware issue. Nvidia also encounters high power consumption but in much more rare cases.
Why? Well... AMD cards have less power steps for VRAM, basically only 0/900/1300/MAX if i remember correctly.
So if data requirements from monitors combined amount*resolution*refresh rate exceeds virtual passive transfer limit, VRAM must go on next step and it increases power consumption decently. This issue CANNOT be fixed via software patch, sadly. But it can be mitigated via FreeSync, or reducing combined specs under limit, for example if you have 2x4k*144 monitors it may be good idea to use 2x4k*120 as setup instead. Or do 120+144 for example, just to get under that threshold. But it is not that low.

On OTHER hand though, some monitors also using out of VESA spec blanking interval, which is less than normal. And AMD didn't account for such thing as it can be seen. Usually VRAM sleeps during that blanking period, but when this interval is out of spec, well... VRAM cannot sleep, and this part PROBABLY can be tied to scheduler. This issue is being gradually resolved via compatability updates on precedence basis.

Personally have not much issues with monitor causing high idle consumption, but i only have bleak 1x1080p*144 monitor. Instead i have issue with VRAM going into power down mode, which surprisingly increases power consumption to 21W from 7-9... Well, to each own problems.

Technically i would notice such thing, for example. My room is heating up really freaking easily. Without AC i would've suffered a lot. But impact of PC in idle during summer under direct sun beaming into small room... Nah, not comparable. PC being used though, on other hand, is still very noticeable.

 

Sheesh, just noticed that my edge to hotspot delta gained +5 degrees in Furmark on 265W load (tbh on every power level from -6 to +15% basically)
So now i get 25 degrees delta avg. And up to 30 in some cases (which are not THAT rare as there is specific hotspot delta behaviour on my card, but that is long and hard to explain, so let's just assume that game load types vary between games).
But worst part. It is not cooling (as it could've not changed), neither it is thermal paste/PTM. As i only was decently sure that it could've not been PTM7950, with paranoid mood i tried to replace it for thermal paste without change and than back to PTM... No change. And imprint on PTM was very good, no reason to suspect bad contact or pump out (not that it could've happened with PTM)
Meaning it clearly comes from somewhere else. Theoretically it can be fact that now i have NVME cover that blocks part of airflow? But then edge temps would've suffered more with hotspot meaning delta would've been same.
Ngl even with maxed 30 degree delta and slowest speed on frontal fan i barely go over 100 degrees on hotspot as edge will be around 71 (80/106 if i throttle GPU fans to 60%). So i am not in throttle range in any case. [Well, to be completely fair, at about 102-103 degree on hotspot mark it seems like REALLY soft throttle appears, like 10 mHz, which cuts basically 0.1-0.2% of FPS, while at 110 degrees it engages in real throttle which will only happen if i keep my frontal fans at 30% (min speed), and GPU fans locked to about 50% (1500 RPM). And it is not thing i do with my system]
But this still kinda sucks to experience. What is different?

 

uninstalling old drivers with "DDU" and install latest version drivers

 

Different card, new drivers, clean install. I've given up on it, its stuck and thats it lol

 

I pull 6W at idle on my 6950XT, only one 165Hz monitor though. Lot's of watts while gaming is the tradeoff, ha, ha!

 

7900XTX, 2x 34" Dell Ultrawide 1440p/144Hz and I have to set them both to 100Hz to get low memory idling.

 

it's the monitor not the card
I swapped from 144hz AOC to 165hz MSI and idle power went down
with the 144hz I could get low power only by tricking with CRU, it worked

 

It's the driver not the monitor.

With my old 290X and 6800XT cards, I had to use CRU to increase vblank time so the card lowered the power consumption in IDLE. Since the first driver for 7900XT, I don't need the CRU hack anymore. In both cases, I use the same monitor - 75Hz ViewSonic.

 

It's both monitor AND driver.
P.S. No issues with my 1080p*144hz AOC
P.P.S UPD:.. Decided to check how it behaves without FreeSync... Well, it was not good, surprisingly. Maxed out VRAM and 31W idle. Pretty sure it is also blanking issue.

 

So one driver works fine with my monitor but another one - no. Should I update the hardware of my monitor to work with the specific driver or it's more logical to ask for a driver that works with my hardware? It's not a "chicken or the egg" situation.

 

On other hand same driver can work fine with one monitor, but not another. As i said, issue is combination of both.

... Seriously? 20 lines by default? CVT-RB2 standard is 77 lines...

 

It's both. Some monitor manufacturers didn't comply with VESA standards and that's why this issue exists in the first place. AMD could fix it with the drivers if they wanted to but it's unrealistic to add a fix in the drivers for ALL monitors that are out there.
Probably they found a workaround using freesync but it's not working as intended because at first the idle power will be normal but after some time playing games, watching videos etc the memory clocks will max out again. You will see memory clocks at 0 but power draw at idle will be ~20W instead of 5W. A system restart will fix it for 30-60 mins and it will happen again. At least that's what happens on my monitor and that's why I still use CRU.

 

Oh f*ck... This is my exact issue. Now i see where root lies. Thank you very much for enlightening.
0 VRAM clock and 21W idle. And i actually told about it for a year, and only now i know reason... After alt-tab it randomly can fall to 0 and only can be returned via proper 3D load (aka game), or Win+Ctrl+Shift+B... And it is only temporary ofc.
When you turn into game, image will freeze for about 2 seconds before VRAM clocks go from 0 to max (as you are in game). It can also break stuff during screen capture.

Oh, and 120 Hz is suddenly fine... Aaand, blanking is 64

Also...Really? Is bandwith actually reason why they do this crap? To conserve tiny bit of bandwith?

 

And I ask you again - what do you prefer? To upgrade the hardware or just to update the software?
I agree that there should be some specifications and both sides - software and hardware to follow them. But it's easier to adapt the software to the hardware than the opposite.

When someone buys a new monitor - how he can be sure that the monitor implements all specifications precisely? Can you comply if you check it somehow later?

Another solution is if AMD provides a list of supported monitor models. Just like some motherboards have a similar list of compatible/tested memory modules.