Nvidia GTX 660 - Crashing Constantly

Toggle sidebar Toggle sidebar
J

Jack Moran

Honorable
Aug 2, 2013
5
0
10,510
First of all, thanks for taking the time to look.
My problem is that, recently, my GPU has been excessively crashing for no apparent reason.
The crash that occurs takes several different forms, either my monitors saying ' no signal', my monitors being solid colurs (such as white or pink) or my graphics card recovering with the TDR message 'Graphics drivers have recovered'.
This started happening fairly recently, and, after the first crash that happened about a week ago, it crashed again around an hour later as one of my RAM sticks failed. I'm not sure if this was coincidence as they weren't branded RAM sticks, and were quite old DDR3 ones.
Because of this i bought 8gb of corsair vengeance and have still been getting the crashes so I don't think they are linked.

My system specs are:
i5-4670k
MSI z87-g55 MoBo
2gb EVGA GTX 660 2gb
8Gb vengeance ddr3 Ram
120gb Evo SSD
Assorted mechanical hard drives
750w corsair PSU.

I am running 2 1920*1080 monitors @ @ 60hz. My cpu is stock clocked, and has never been overclocked. it is also cooled by a h100i and isnt having temperature spikes.

The current Nvidia driver I am on is 347.52 and my previous driver (when the first crash happened) was 344.48.
All of my drives are up to date across my system.
Before the crash happened the only thing that had changed was that I was using 3 monitors. 2 on my GPU, and 1 on my intel integrated on my motherboard. I have since uninstalled the Intel drivers and turned the option 2 use the integrated off via my MoBo, and the only graphics chip picked up in device manager is the 660.

I have memtested, CPU diagnosed. I have also monitored my GPU temps and after moving some fans around, it is levelling out at around 50c while playing counter strike global offensive (one of the more afflicted games). I have underclocked it to the base clock rates, to no avail (however it does seem to be running for longer between crashes, could be coincidence).

Any help would be appreciated.
 
Solution
T
".. I have underclocked it to the base clock rates, to no avail (however it does seem to be running for longer between crashes, could be coincidence)..." Try underclocking below normal 660 specs, both GPU and memory. If that stabilizes things then you know it's not software problems, it's less likely to be PSU problems, etc.

Agree with *not* using both integrated and GPU to drive monitors. Any chance the crashes only happen when running a particular game or program -- things aren't as well tested with multiple monitors. If underclocking doesn't work suggest dropping to a single monitor to see if that helps.

Your temps seem good to great (50C, my 7850 is closer to 75C), however things like VRMs on the card are not monitored...
Sort by date Sort by votes
".. I have underclocked it to the base clock rates, to no avail (however it does seem to be running for longer between crashes, could be coincidence)..." Try underclocking below normal 660 specs, both GPU and memory. If that stabilizes things then you know it's not software problems, it's less likely to be PSU problems, etc.

Agree with *not* using both integrated and GPU to drive monitors. Any chance the crashes only happen when running a particular game or program -- things aren't as well tested with multiple monitors. If underclocking doesn't work suggest dropping to a single monitor to see if that helps.

Your temps seem good to great (50C, my 7850 is closer to 75C), however things like VRMs on the card are not monitored. How long has the card been in use? Have you tried blowing dust out of the card, especially under the heatsink as well as in the part where the fan runs ?
 
Upvote 1 Downvote
Solution



First, thanks for responding!
I'm going to try underclocking below normal now, do you know what clock rates would be best for the memory? it's just when i edited them just it instantly crashed and i was wondering what a stable underclock would be.

Second, the game ive noticed it most frequently in is counter strike global offensive. I've ran this game while having dual monitors for over 2 years though and its a recent problem, ill still try one monitor though.

I also did a full case clean with compressed air around 2 weeks ago, and attempted to clean the gpu as much as possible. The problem is that the EVGA 660 is enclosed in the casing to blow air out the back of the case, so is difficult to clean properly. I blew air through the card and i might of dislodged dust and actually made the issue occur/happen if it's causing an unmonitored part to overheat? could that have happened?

I've also just, today, done a full reinstall of windows to remove any old intel drivers or nvidia drivers that may have conflicted and installed nvidia drivers 344.48 (some stable drivers that i use almost every windows rebuild, so i am confident it isn't a driver issue) and the issue is still persisting.

 
Upvote 0 Downvote
You've got a good plan. The reinstall of windows solves any problems with data corrupted by your old failed dimm and stored back to disk.

Also, see if you've set your MB to automatically overclock memory (I saw your CPU was not OC'd). If so the new memory might have an XMP profile that is OC'd more than it or your MB can handle. Maybe verify in bios the memory timings vs. supported.

I don't know a good value for the memory underclock. I do have one kids with an HD7850 running 10% slow on both clocks to get stability since the card was new. He's prefered to keep the slower card than RMA it. We started 20% slow, that worked. So we went to 10% slow, that worked. We then went back to stock, which failed in a few hours. 10% has worked for a year now of daily use. A few months ago on a reformat from vista to win8 we tried removing the 10% -- failed in a day, so it got underclocked again. Hope this helps.
 
Upvote 1 Downvote


Well after taking the GPU apart, cleaning the heatsink thoroughly and noticing quite a lot of dust, i have reseated it with new thermal paste and all was well until today. I managed games etc. yesterday and today it's crashing on a lot of games.

I took your advice and checked the BIOS, and noticed that on the auto dram frequency it was running at 1300 as opposed to the 2133 my ram runs at, so i have now upped this accordingly. I couldn;t find a corresponding voltage setting, and i'm not entirely sure what i would volt it to to sustain the faster clock speeds on the RAM so i will see how this goes.

In regards to using one monitor, the GPU still crashed with no apparent difference than if i was running 2/3.

If worst comes to worst, im going to get another GPU next month (i was thinking 970 unless you/anyone has any better suggestions) and try that. If that crashes, i'll do a differential voltage test on the PSU and see if it's that, and if it is replace it. My GPU is due an upgrade anyway, as long as i don't have to replace my MoBo or anything.

Any other components this sounds like it COULD be?
 
Upvote 0 Downvote
Just an update if any one has ideas:
I've ran GPU-Z and recorded it to a log file.

The second before the GPU crashes, the TDP spiked from 38%, 40%, 33%, 40% to 97.6% and the next recording is 18% when the gpu crashes.
Is the 97.6% TDP which follows after a 98% GPU load spike.
This '98%' load was while on a half life 2 mod, not sure if it's super unoptimised or my GPU is abnormally spiking.
 
Upvote 0 Downvote
There is GPU load and TDP. they are different. TDP is the max temp (very roughly). Saying TDP spiked from 38% to 97.6% says the card got hot suddenly. GPU load is an indication of how hard the card is working, it has nothing to do with temp.

The GPU load spiking from 38% to 97% max sense. The temp jumping up and then falling to 18% of max temp doesn't really make sense. Can you check the units and make sure you are reading the right data.
 
Upvote 0 Downvote
I hate to necro this post but I'm having the exact same issue with the 660. I've switched it over to a 630 (Much lower power consumption) and the regular crashes stopped, but now anytime I load a game (League of Legends, Diablo 3, Halo, etc) it will crash in the same manner as the 660. I'm thinking it's a bad PSU, but I'm unable to test it as I don't have a voltage meter.

The thing that bugs me is that I get driver error messages sometimes even though I have all the latest drivers and updates.

Haven't tried a clean install yet.
 
Upvote 0 Downvote


I have had the same issue over the past 2 years in various different systems, tested everything and when i switched back to my old card (gtx 460 SE) everything works perfectly. used to have my pc crash about 5 times a day doesn't happen at all for me without a 660. been through my 5th one through warranties and they still don't understand that the 660 is a broken GPU. if yours is still under warranty see if you can get it replaced with a different card. currently trying to convince the company i bought mine from that the components design is flawed.

ps. as a side note every new 660 has run for about 4 months before developing the same issue again and ive tested the same rig with a 780ti for months and no issue, sadly i have to send the 780ti back though since its not actually mine and was loaned to me by a friend.
 
Upvote 0 Downvote
for a few months after i put my new machine together... i was having this same issue. i find that if i update the driver for the 660 to anything newer than 344.75..... it will crash. take the time to read the update notes for the update prior and after this one i am using of 344.75... if u pay close attention to the update notes u will see there were some things changed after that will directly affect our card. i felt these changes are the issue. i have been running on 344.75 with no issues now. and everything is cranked up to the max on my box.

msi z87-gd65 gaming board
gtx 660
16gb ddr3 ram patriot black
i7-4790k
750w bronze supply
500gb SSD Evo samsung
all in a nice Antec 900 case

i used to get the same white screens. and a pop up at the bottom informing me of the driver failing and reloading up. a good half of the times were while i was streaming on twitch. the other half were just normal gameplay... game of choice. guild wars 2....
 
Upvote 0 Downvote


Ummm. ".. been through my 5th one through warranties and they still don't understand that the 660 is a broken GPU. .." After trying 5 of these did you consider that something OTHER than the video card might be broken ?

+1 HaxorTeamGo for mentioning that going to an earlier level video driver might help. A full system software rebuild might do wonders too. If the GTX 660 was failing any where near the rate needed for you to get 5 cards that all failed the same way then the internet would be full of stories about it.
 
Upvote 0 Downvote
Better late than never - maybe this will be helpful for those, who are googling it now:

Had similar issues with my GTX660 - GPU crashed (screen went black for few seconds, sound freeze, sometimes restart needed to resolve, sometimes it just went back to normal) - what I did was in MSI Afterburner just turn the GPU Power to 110%, thats it! Problem gone! Well, at least for me! Hope this helps!
 
Upvote 0 Downvote


 
Upvote 0 Downvote
I had the same problems, starting with the last 2 nvidia updates. I finally went back to the October 1st update and that solved all of my problems. I had crashes on start up from sleep or hibernation. When I would roboot it would take a few minutes and then crash with different colored squares and lines then solid brown or orange screen. Once in a while the recovered from video crash would pop up. Since I reverted back I have had no problems. 373.06 is I believe the correct driver to run. Hope that helps. I would hope that nvidia comes out with a fix soon.
 
Upvote 0 Downvote
I have solved the problem, i can't belive it! OMG OMG OMG i'm so happy i don't know what to do...
I had this crashes ever since i have bought this pc.
It works real nice i have a Amd fx 8350 8core, 16gb ram, gtx660 2gb gddr5 , 750 watt PSU .
If i was playing dota2 or cs go or gta 4, 5 , it runs real nice and smooth no lag but after sometime it randomly crashes.. sound freezes, monitor turns black sometimes pink sometimes white.. sometimes i have to switch off the psu so my pc can restart.
After reading in two years lots of oppinions and try this try that.. i tried a lot of things non of the things worked for me............
I can't belive it is so easy and nobody said this before, or i haven't read it on any forum by now.
I accidentaly changed my resolution from the NVIDIA settings. Open Nvidia Control Pannel , click on Change Resolution under the Display tab.
now you will see something like this.. Connector: HDMI - HDTV , Under connector you will see Resolution: Skip the HD - SD section, scroll down to the middle and click under the PC section your desired resolution.Click apply!

no reboot needed. now enter your game , play and expect to get crashes ( you won't )

thanks i hope i helped somebody in 2017 :)
 
Upvote 1 Downvote



Damnn!!! U save my ass... It works for me..
The problem is: On march 16 2017 while mining eth, an error occur "out of memory" and the error persist even if u re-benchmark.
After seing this post, I remembered that i remote my pc from other pc with a 32' monitor. Therefore resolution change..
Thanks Man!!!
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom