Discussion in 'Videocards - NVIDIA GeForce Drivers Section' started by PC-Tweaker, Dec 12, 2020.
not my experience, 10 series on 460.79 was set to legacy by default.
3 series might have a different set of rules?
potentially, To be honest, the reasons NOT to run msi's by default don't exist with Turing and Ampere as these cards can't run on Gen 1 controllers anyway.
2 series also legacy by default here.
I'll check what happens with the next driver release.
You never know.
Though Astyanax didnt have the same experience as me, it remains to be seen.
Thanks for sharing them!
Not only theoretically I think, pratically speaking not bad either (those pics are from my W8.1x64 install), all smooth for both sound and video for going all MSI, no (micro)stuttering here (except for those badly optimzed console ports of course, not much to do about'em)
That's the strange thing, no issue with that in W8.1 (see pic) as opposed to W10 (no go under 1544Mhz even when idling...)
At the rate NV are (paper)launching those 30x0 series, no worries, your tool is here to stay for quite a few years
@Astyanax , do you have a 3 series gpu ? (in your sig is a 1080Ti)
Would be interesting to see if any other owner of a 3 series card also have MSI enabled by default, or not...
I don't need hardware to check,
30 series uses Section 174 of the inf
Section 174 contains
AddReg = nv_msiSupport_addreg
[nv_msiSupport_addreg] does not exist except in the Reference inf, Mufflore should not have had MSI's enabled by default based on the Inf structure. No, wrong, I failed to look for its existence Above section 174
HKR,"Interrupt Management\MessageSignaledInterruptProperties",MSISupported,0x00010001, 1
HKR,"Interrupt Management\MessageSignaledInterruptProperties",MessageNumberLimit,0x00010001, 1
Turing uses Section002, Section002 does not have a [Section002.HW]
Pascal uses Section 090, also doesn't have .HW.
@Mufflore the 457 drivers shouldn't have required MSI Util either.
Indeed. Unlike cards from previous desktop series, RTX 30 series GPUs appear to be set to use MSI mode by default (my RTX 3090 does as well).
So, what is the solution and exact steps to resolve this issue? Thanks.
Nvidia Release Notes:
460.79 and 460.89
[GeForce GTX 1080 Ti]: The desktop screen may flicker. 
Use a different gfx card.
Use a previous driver
Use a lower refresh rate
Set max power in nvidia control panel
Wait for a fixed driver
Have a 1080Ti with no flickering, but with 3 monitors. My idle clock is 1468mhz, but thanks to nvidia inspector it idles at 202mhz with 3 monitors. The only flickering I have is when I login to desktop.
To Change Power, open Nvidia Control Panel, select the Manage 3D Settings on the left, Select the Global Settings Tab on the right, and find "Power Management Mode".
Here are my observations at idle (desktop tasks such as Outlook, Chrome, Firefox, etc.) with a GTX 1080 Ti:
Optimal Power: Core 139 Mhz, Memory 203 Mhz, Power Use 7%
Prefer Maximum Performance: Core 1493 Mhz, Memory 2754 Mhz, Power Use 27%
Adaptive: Varies depending on load - Core 139 to 1493 Mhz, Memory 203 to 2754 Mhz, Power Use 7% to 27%
Values obtained using Argus Monitor
Hello, I created an account just to answer here since I also had that problem appear all of a sudden since last patch. These days I play Cyberpunk and Hell Let Loose, Hell Let Loose config files are already pretty heavily tweaked so I didn't notice this issue, but in Cyberpunk, massive chunks of the roads and buildings were flickering just like you described.
A simple modification in the official Nvidia Control Panel fixed it for me (not NVInspector, the regular control panel), since I noticed there were kinda "2 instances" of the textures that were flickering, as if one said texture was rapidly switching back and forth between high def and low poly model, creating said flickering.
Change that option in the control panel :
Texture Filtering - Negative LOD Bias : Clamp
Fixed it for me on Cyberpunk specifically, can't say for other games but it's worth a shot and takes about 5 seconds to test.
I'm sorry to tell you this, but no, this setting didn't do anything for you. It has no affect on Kepler or newer, which have a completely seperate lod implementation to Fermi and older - and negative or positive lod have nothing to do with Z-fighting what so ever.
In case he did not touch anything else it did...
nope, without modding the driver kernel, the code is never touched on kepler and up.
If the only thing he touched was that option in NVCP then it fixed the issue. No matter if you correct about that option`s mechanics or not.