Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by Hilbert Hagedoorn, Apr 14, 2021.
Ok thanks anyway, wait and see
I'm sure they are just ramdisk issues, all of the dump files i have analyzed list lmdisk.sys
I did uninstall the 04.21 updates but it didn't change a thing. Was getting 45 FPS tops in WoW: Classic and Elders Scroll Online on my RTX 3090.. Re-added the 04.21 update and downgraded to 465.89 and my performance is back to normal.
You have to be on either 19042.867 (20H2) or 19043.867 (21H1) builds to restore performance. Also make sure you use WUMT to hide the update.
I uninstalled this driver and returned to the previous one from nvidia. No more crashes or lockups. I also uninstalled the KB from ms after that but reinstalled it again since it doesn't crash anymore. To me this driver causes unexpected crashes, and I don't feel like trying it again with this KB5001330 uninstalled plus this driver. Performance isn't better compared to the previous one, and perhaps even worse in certain game scene's. 3DMark did give improved results, but I prefer real games over benches to play.
Anyone getting something like this on their 3080?
I noticed the issue happen quite frequently in BDO,sometimes in warhammer totalwar 2 and VERY rarely in wow.Usually happen when fps is close to max for some reason.Isse seems to be prevelant to happen more frequently on 1440p.
I tried another monitor that was 1080p 144hz and the issue happened again when I upscaled to 1440p
Rig is 3080/9900k/970evo/16gb ram/750 psu/rog maximus. (driver 457.51)
Monitor is S2721dgf Nvidia settings V-sync on/off ingame with -5 max frame limit(so 160)
I tried sending a trace log but nvidia removed the easier way to make it https://nvidia.custhelp.com/app/error/error_id/1
Instead were stuck with this https://nvidia.custhelp.com/app/ans...5cE1OX05UV0t5SmoxNjdIblNTYW5zSWwxb2clMjElMjE=
And I keep getting stuck at the CMD part
From the Command Prompt, navigate to the directory: C:\Program Files (x86)\Windows Kits\10\Windows Performance Toolkit\gpuview
If you chose a different installation path, navigate to that directory and drill down to the gpuview folder.
3. To start logging, type the command (the first time you run the command, you may need to repeat the process one time to start recording).
but nothing happens.
Getting annoyed now.Support is slow as hell too.It took them 30 days just to reply and all they could say is updated to latest driver or update to a beta bios.
Realy you still on that wow!
This is what I got from nvidia support after a month.
My colleague escalated your case to me for further review. From your feedback and troubleshooting, its too early to tell if your experience is related some perf issues we are investigating. At the moment, we can't rule out issue on any hardware in your system. Are you still hitting the problem with the latest 466.11 driver?
There is a newer SBIOS for your mobo, version 1901. Does the update make a difference?
So yes I'm still here.The fact they took a literally month to tell me update my bios (to a beta bdw,I have the bios prior) and updating to the latest driver (I crap you) is partly why I'm still here.
Log collector still available
Yep uninstalled that.
Il see if I can catch the issue with the log collector
Ok downloaded latest driver and at first I was getting the issue again.Tried playing in fullscreen and this time they stopped.(fullscreen didn't work on previous 457.51)
Don't get me wrong the game still has its micro stutters/stutters when running around with a horse really fast in the world.
These stopped from what I can tell.
Another issue I noticed tho is when running around on a horse ingame and gpu us close to max gpu usage (even seen a 100%) the gpu clock was dropping below boost.
My boost for my 3080 1785mhz and I have it the game set to maximum performance.I saw it occasionally drop to 1755mhz quite frequently while effects where going on which shouldn't be possible?Temps where below 70c too so it's not a temp issue from what I can tell.
So it was a driver issue?
What do you mean? clock adjustment is the base function of Nvidia GPU Boost, the lowest clock a 3080 can operate at is 1440 Mhz as the base clock (though would only happen under significant thermal problems).
Thermals are not the only metric for clock adjustment, Power util as a percentage of TDP is also a factor, as well as device perception of safe operating voltage.
As in my gpu was going below its boost speed while it had maximum performance enabled in the nvidia control panel for said game.
In all my games usually if I set maximum performance it never goes below 1785mhz which is the boost speed.I could be idling with 20% gpu usage and it would stick to 1785mhz
What I noticed was the gpu was going lower then 1785mhz when it shouldn't.Alteast I've never seen it happen in any other game.
Tested vermintide 2.Kinda dissapointed in the performance but its vermintide.Heard it could be the game.
the advertised boost speed is an average that sits at the middle of the boost table, your base clock sits at the bottom of that table and is the only guaranteed clockrate under max performance, everything above is dependent on thermal, power and util factors. The advertised boost clock is certainly not a guaranteed speed and never has been.
some sites used to put up charts of boost steps corresponding to vid steps, these days they don't bother, with pascal they were steps of approximately 13mhz, turing and ampere appear to be 15mhz steps
that would mean between base and advertised theres 23 15mhz steps, and likely 22 more above it for a total of 45 steps and a maximum possible boost of 2115mhz - im guessing your 3080 is reference clocks?
Its a 3080 tuf oc edition.The card actually boost horribly.The card barely breaks 65c and its quite erratic going from 1965mhz (or 1950mhz) to the 1800s at the slighest gpu flactuation.Its simply can't hold it's clock when crap get real which is wierd because my old 1080ti could hold 1950mhz until 70c and then boost down slowly to 1924mhz regardless of whatever happened.If that's not a sign of instability I don't know what it.
That's why I'm thinking there's something wrong with the card itself or the bios of the card atleast.
After playing some more games it seems the issue from black desert is not 100% completely gone after trying warhammer.Its just greatly reduced.I just can't accept such erratic frametime to be "normal"
You have them mixed. 19043 = 21H1, 19042 = 20H2. I'm running 19043/21H1 and I was back at 19043.867 but my performance running 466.11 was still terrible. Rolling back to 465.89 and my performance is back to normal running both 19043.867 and 19043.962.
I did also try 470.25 but it was just as bad as 466.11 when running both 19043.867 and 19043.962. I don't know what is causing the issues, but for me running 465.89 is working great no matter what Windows 10 Update I have installed so I'll just keep running the old driver for now.
hmmm tried 465.89 while playing a game and got this
https://drive.google.com/file/d/11fX51DYwH1vmn-Yp__74W8-sc9R0jaI6/view?usp=sharing stats sigh.
3080/9900k w10 20h2
can't consider dpc results while gaming, the pcie bus and cpu cores can go into low power states (parking, lspm etc) and then a delay cycle is triggered when the bus and chips are returned to high performance in which the device won't process any commands.
you can confirm this with hwinfo64 sensor logging, and the generic log viewer tool.
Would be nice if hwinfo64 logged dpc's at the same time and showed them on a graph
Even if it doesnt happen with certain drivers using the same settings?even if c-states/speedshift/step and maximum power is chosen globally?
I tried going as back to 457.30 and tried to display the erratic gpu clock when effects are going around but the game straight up crashed.
Also another possible issue I noticed on my 165hz monitor.It seems to periodically and randomly cap itself to lower fps.Don't know if its just a coincidence.