Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Vidik, Feb 5, 2022.
Nvidia Studio Driver v472.84 WHQL
Seem to be ok!
the io issues were fixed in the later studio 472's.
People just noticed huge difference (in number of registry requests) between two versions of NVIDIA drivers in that process. Since it is responsible for display things it has to depend on videocard drivers.
Yes, it is also usable from videocard drivers (as part of the UI). Windows are a graphical thing and they can be accelerated by Hardware. This Hardware has also Impact on color depth, refresh rates and so on.
What I wanted to say is, that as far as I know (but I am not sure), dwm.exe is an important part of the UI, so you will find a lot of entries in the process manager. Some of them are okay and some of them should not be there (like that registry writings). Most of us - including me - hardly ever watch the process monitor. So we do not know what is constantly going on.
Recent driver sets have given me the ole G-Sync flicker in WoW Classic again, so I've downgraded to 472 branch for now where everything was working fine.
Looking at this thread, I'd reckon I've made the right decision anyway.
Guys, you can't just compare CPU usage percentage with each other and decide what's normal. Everyone has different CPUs and 1% of someone's 5950X might be 5% on some low-end CPU (just an example). Also that dwm.exe is a windows process and has other jobs to do as well so usual spikes is fine.
On the other hand the amount of registry queries is definitely abnormal and seems to be a leftover from some testing/debugging.
To some extent, you can. DWM should just never consume more than perhaps ~3% at most for a brief moment, apart from some really slow CPUs. Sure this isn't exact performance analysis, but it's not entirely useless either.
Just did a recheck with 511.72, it's 0.1-0,6% on my end, must have been mixing it up with GPU usage which does go a bit higher dependent if anything requires hardware acceleration.
472.84 gets the monitors info once and is done with it, so definitely a bug that 511.xx continuously requests it.
Indeed, and it seems to have been left enabled for quite a while to.
I just went back to 472.84 after having the lastest all the time.
Been playing for 2 hours now and it's definitely the smoothest of the bunch.
472.84 the smoothest?
Atleast with my current configuration.
I was on all the latest drivers since then, and it never really felt perfect to be honest having tried the 472 branch again.
For me dwm.exe isn't that bad, it doesn't produce exactly those registry actions, intead it's HKU\...\SOFTWARE\Microsoft\Windows\DWM\ColorPrevalence and HKU\...\SOFTWARE\Microsoft\Windows\DWM\AccentColor being accessed all the time. Also HKLM\SOFTWARE\Microsoft\Windows\Dwm\AccentColorInactive. Anybody else has this?
I had ColorPrevalence being spammed, but don't recall seeing any of the other mentioned keys.
Increased i/o (writes to registry DB which is a large file).
Increased cpu utilisation of DWM process.
Do I need to worry about this guys?
I dunno what people are reporting but DWM.exe itself is a system's desktop manager process which does a lot of things to compose what you see on the screen. When it doesn't do these things it's less then 0.1%. When it does then it's not the issue which results in higher load.
And you'd need to run them on a i386 with an HDD from 1985 to actually notice either. Which is impossible because these don't have PCIE buses.
The issue should be fixed but it doesn't affect regular system usage in any way.
Panicking isn't mandatory, but highly recommended. The alternative is to don't mind and live happily ever after. How boring this would be though.
On my laptop 4/5 of the writes are the registry, far more than the logs, so NAND usage would be noticeable with constant registry updates.
Sounds like me, I live dangerously playing Serious Sam not in real life
Please. Are you seriously suggesting that some KB/s writes could be "noticeable" on an SSD?