

Nvidia Driver Crash
#1
Posted 20 November 2013 - 04:40 AM
symptoms: flickering boxy artifacts on frozen screen, serious sound lag/distortion
usually the drivers restart in time for me to continue playing, providing im not killed in the 30 seconds or so of it happening
#2
Posted 20 November 2013 - 04:54 AM
#3
Posted 20 November 2013 - 05:17 AM
#4
Posted 20 November 2013 - 09:24 AM

#6
Posted 20 November 2013 - 11:10 AM
#7
Posted 20 November 2013 - 04:39 PM
#8
Posted 20 November 2013 - 09:13 PM
This TDR (Timeout Detection and Recovery) error, is known by Nvidia, but mostly ignored when it comes to a solution, their official statement:
https://forums.gefor...ddmkm-stopped-/
If you find a "display" error in your windows event manager which refers to "nvlddmkm.sys"...welcome to my nightmare...
I have this Issue on two different Win 7 rigs...my old system with a Gigabyte GTX 560 Ti and my brand new rig with a Gainward GTX 760 Phantom...in both cases the cards are factory-overclocked...
Best workaround I found was to downclock the cards...

For me Nvidia lost a lot of credit, especially after going through that stuff for the second time...
Edited by SpOoKy777, 20 November 2013 - 09:16 PM.
#9
Posted 20 November 2013 - 10:25 PM
If you're having this problem with factory-overclocked cards and it's solved by down-clocking the cards (and you already have Prefer Maximum Performance set), then you have faulty graphics cards and you need to RMA them. Factory overclocks should be 100% stable in every game, but they sometimes have some that pass QA that shouldn't. It's not Nvidia's fault -- it's the specific manufacturers of your cards.
#10
Posted 21 November 2013 - 10:19 AM
The "Prefer Max Performance" thing is enabled on both systems...
The GTX 560 was RMA'd back then and came back with the error still happening...nowadays I use the old rig for video/internet only, reverted to the last stable driver (which was 314.07 for me)...and everything is fine...
With my new gaming/working system (GTX 760) I can't go back to that old driver, because its not supported anymore...so I test every new driver coming up from Nvidia...so far no luck...error is still there...
Weird is that I'm able to play hours and hours of Battlefield 4 (ultra settings), no errors...same for FarCry3 and Rage...I can run FurMark all over the place...no errors...
I start MWO or Crysis 2/Crysis 3...play for 10-30 minutes...BAM...driver timed out...same for Napoleon: Total War...thats what I've tested so far...
What I'm trying to say...if both cards are faulty...they shouldn't:
1) ...work fine with reverted drivers...
2) ...work fine randomly with some games (game engines)...
Edited by SpOoKy777, 21 November 2013 - 10:22 AM.
#11
Posted 21 November 2013 - 03:13 PM
Rolling back to a previous driver likely disables the newest functionality that SHOULD be enabled. Evidently it's this newer functionality that's causing your GPUs to not be getting enough voltage. That's why down-clocking works -- the GPU no longer needs as much voltage while running at a slower speed.
I have seen others on actual hardware forums have to RMA more than once (several times, in fact) to get a fully stable functional factory-overclocked GPU.
#12
Posted 21 November 2013 - 03:57 PM
#13
Posted 22 November 2013 - 10:32 AM
Quote
hmmm...thats so damn unconvenient, but I'll try it...I contacted my hardware dealer today...
Quote
Thats correct...as I wrote, at the moment the 560 system is at 314.07 and stable for what I use it for...so I could go up to the .22 but I prefer to let it where it is...
The new rig with the 760 is what bugs me the most...

But thx for the help so far, guys...I'll post an update if the RMA is done...
#14
Posted 22 November 2013 - 02:51 PM
What I tried after using those settings was reducing screen resolution and playing in a window. My desktop is currently set to 1600 x 900. The game is set to 1280 x 720.
I might make further tweaks later, but it's working right now for me (and in-game settings, V-Sync is left unchecked, everything is set to Very High and I have PostAA on).
So that could be another workaround to test out.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users