[modcrash.com] Nvidia 320.18 WHQL Display Driver is Damaging GPUs

[aigomorla]

i assume none of the cards which have died were watercooled?

cuz if one was, it would point to a very big problem.
More on the lines of voltage being fed in excess to a IC which wasnt designed to handle it.

If all the cards were on stock air cooling, id think a IC heated up and died.

 

[Vesku]

Was playing a MMO and someone asked if the rest of us were getting artifacts in the mission we were running. Another person immediately asks "Nvidia card?" turns out the person artifacting has a brand new 660 TI, the guy who asked if it was Nvidia tells him his 650 was having problems with latest drivers and the 660 TI guy should rollback.

So yeah, pretty buggy driver if I am running into people with artifacting issues in a random collection of 5 gamers online.

 

[amenx]

Nvidia have been stumped on this driver since June 3. Dont see any updates to this statement. Although I have zero issues with this driver, I think they should pull it from their site. And reissue the 314.22 with updated INF for the new 700 series cards released since.

June 3rd Update: We are still unable to reproduce the general freezing and crash issues internally. We have tested our entire suite of QA systems, and an additional suite of systems featuring factory and user overclocked GPUs, but have not had a single occurrence of the reported problems. We need your assistance to make progress: if you are suffering from these issues provide as much information as possible via our feedback form: http://surveys.nvidia.com/index.jsp?pi=6e7ea6bb4a02641fa8f07694a40f8ac6

Furthermore, if you live near our Santa Clara headquarters please mention this in your submission - we can borrow your afflicted system, allowing us to instantly reproduce the issue and debug it immediately.

If you have extra info that you cannot add to the feedback form, please note your submission and add the extra details to a post in this thread: https://forums.geforce.com/default/...lay-driver-feedback-thread-released-5-23-13-/

We want to fix this problem, we're gamers just like you, but we need your help. There are millions of potential system configurations out there and we simply cannot build and test every single one prior to a driver release. Previous releases have been stable and our Quality Assurance processes have worked satisfactorily. Your info, and your systems, will help us reproduce and fix the issue.

If you continue to be afflicted by the issue I recommend rolling back to the last stable driver, be that a 320.xx beta, or 314.22 WHQL. If rolling back fixes your issue please tell us, as that info is helpful also.

Andrew Burnes
NVIDIA

 

[Stuka87]

Nvidia have been stumped on this driver since June 3. Dont see any updates to this statement. Although I have zero issues with this driver, I think they should pull it from their site. And reissue the 314.22 with updated INF for the new 700 series cards released since.

I work in QA, and I feel their pain. Trying to replicate an issue that is happening in the field now and it can be hard without knowing all the variables.

 

[Gikaseixas]

If they're aren't able to pinpoint the issue then there's good chances for the problem occurs on the next driver release??

 

[VirtualLarry]

If they're aren't able to pinpoint the issue then there's good chances for the problem occurs on the next driver release??

Seems likely.

 

[flexy]

I am running the 320.18 since they came out and have to admit I first ignored those "driver fries card" threads since it doesn't sound plausible to me.

Then again, with my 660 TI OC I see rarely and on occasion very similar artifacts as I have seen in some screen shots (distorted polygons), eg in FC3, TR3 but always attributed those occasional problems (also rarely green/pinkish checkerboards) to my high memory overclocks.

Now, on the other hand, those threads about the 320.xx problems make me curious whether its not really a driver problem.

My OVERALL impression of the GK-104 card that it's "finicky" in general and simply un-predictable, especially when overclocked.

People don't realize that, in terms of core overclock, Kepler most of the time has no headroom whatsoever, 40+ on the core or so is usually already enough to make the card(s) entirely unstable, Kepler is simply not a chip well suitable for overclocks, at least not for core clocks.

Some guy also had an interesting comment and said that many cards that come overclocked out of the box might already hit some critical threshold and I could see this well being the case. Add some higher temps like now and you get a pretty unstable system, together with the non-predictability of GPU Boost etc...and then of course the users who use 3rd party progs like AB/Precision etc. and keep adding some clocks to the core without really knowing what they're doing.

The same in regards to Bios mods, there is tons of users who simply get themselves a modded bios because 'some dude' somewhere says it's "better", but also lacking the understanding about consequences what the modded BIOS actually might do. Example: If you unlock your BIOS and 'simply' make the higher voltages available without any other changes , this alone makes Kepler boost higher in the table but who tells you it's stable? On my Gigabyte 660 TI OC, the max boost table clock in the BIOS is at 1320 (!). So..let's say I 'unlock' my bios and set both sliders to 1212.5mV, the card attempts to boost to 1320 now, needless to say it's not stable at that clock. And mind you, this only from unlocking the voltages WITHOUT any added core clocks and of course without giving added core clock via AB.

People don't realize that overclocking Kepler is a more complex and often paradox thing to do (because of how GPU Boost works)...also, add the difficulty to test memory overclocks as being "stable" since GGDR5 mem has error correction built in.

 

[3DVagabond]

So return your card to stock and see if it stops.

 

[nitromullet]

I had artifacts in BF3 with the 320.18 drivers with a GTX 770. Decided to see what version was on the disc, and it turns out EVGA put the 320.08's on there. So far so good with the 320.08 driver.

edit: just got the card, so it's running stock clocks.

edit 2: I guess I spoke too soon. Getting artifacts with the 320.08 driver too.

 

[toyota]

Far Cry 3 just went bananas with the artifacting. I had played around in the game off and on for weeks now on 320.18 with no issues until today. restarted game and it was fine. still thats odd to just happen out of the blue like that.

EDIT: apparently many people have had this randomly happen in the game long before the 320.18 drivers even came out though.

 

[3DVagabond]

Far Cry 3 seems to give both brands fits. Stuttering, artifacts, etc...

 

[toyota]

Far Cry 3 seems to give both brands fits. Stuttering, artifacts, etc...

game had ran super smooth for me with no issues in about 4 hours total with the game. the artifacts a few minutes ago freaked me out but yeah after searching its apparent lots of people have been randomly getting those for months.

 

[flexy]

Far Cry 3 just went bananas with the artifacting. I had played around in the game off and on for weeks now on 320.18 with no issues until today. restarted game and it was fine. still thats odd to just happen out of the blue like that.

EDIT: apparently many people have had this randomly happen in the game long before the 320.18 drivers even came out though.

This is exactly what I mean with "finicky" and "unpredictable". This is really why I hate this gen of GPUs.

You can run tests with Heaven and run a ton of benchmarks and testing suites, monitor your temps and check for errors for hours - but it might STILL artifact or distort polygons when you least expect it and even after you're convinced the card is stable. I have never, EVER had such issues with the old GTX 275. In terms of OC, I set my OCs after some testing and then run the card (275) with the OC set in BIOS literally for years without ONE issue. With THIS effin' card you literally never know what's around the corner, regardless how many hours you test it.

 

[thilanliyan]

This is exactly what I mean with "finicky" and "unpredictable". This is really why I hate this gen of GPUs.

You can run tests with Heaven and run a ton of benchmarks and testing suites, monitor your temps and check for errors for hours - but it might STILL artifact or distort polygons when you least expect it and even after you're convinced the card is stable. I have never, EVER had such issues with the old GTX 275. In terms of OC, I set my OCs after some testing and then run the card (275) with the OC set in BIOS literally for years without ONE issue. With THIS effin' card you literally never know what's around the corner, regardless how many hours you test it.

Different games/benchmarks stress differently so it is bound to happen. I remember a while back (not sure which card) my card was stable in games but would always crash in 3DMark11.

 

[Mem]

Update on my 320.18 to 314.22 drivers,as you probably know I was getting lockups out of gaming ,games freezing at launch ie ME3 and basically issues like that since I had the 320.18s installed for my 560Ti card,I'm happy to report that the good old 314.22s are still running great with no issues either in or out of gaming on my PC.

I stand by my statement that the 320.18s are probably the crappest Nvidia drivers I have ever used since TNT days.

Anyway there must be something to it since I'm not alone and many others are having similar issues.

My default running 560Ti card runs cool anyway especially in my LanCool K62 case with 4 case fans(smallest being 120mm) and only reason I can see for the lockups/crashes was the 320.18s drivers,funny how my Reliability History chart has gone from critical events(some were caused by Nvidia and others was Windows not shut done properly ie hard lockup) ie red crosses on graph chart to none since I went back to 314.22s about 10 days ago so rock solid ,(no hardware or software changes in my case apart from drivers).

I'm still a bit surprised no Nvidia statement on this matter.
For the record I don't consider myself anti-Nvidia ,just a user/owner like many of you that uses both AMD and Nvidia cards,funny how the AMD conspiracy always pops up when Nvidia owners like myself have issues,guess I was dreaming it all and I'm still asleep .

 

[Chiropteran]

There is a new driver out now without the bug-
text

 

[thilanliyan]

guess I was dreaming it all and I'm still asleep .

Must be, because nVidia drivers are perfect...absolutely nothing has ever gone wrong with nVidia's drivers...ever. :sneaky:

 

[stahlhart]

Infractions issued. Please keep the discussion on topic, that topic being the issues surrounding the 320.18 WHQL drivers.
--stahlhart

 

[Chiropteran]

I'm shocked. In dozens of threads about AMD products we have nvidia users coming in and suggesting to use an nvidia card or Intel users suggesting to use an Intel CPU, and I've never seen anyone get an infraction for such posts in the past.

It's kind of a scary precedent, from now on I guess you can't compare a product to it's competitor's equivalent without getting an infraction.

Anyway, to the point of the thread- I don't really understand why this is such a big issue for nvidia to fix. Don't they have a change-log in their code, so they can see exactly what was changed in these drivers? Doesn't nvida follow basic code version policies?

 

[Stuka87]

I'm shocked. In dozens of threads about AMD products we have nvidia users coming in and suggesting to use an nvidia card or Intel users suggesting to use an Intel CPU, and I've never seen anyone get an infraction for such posts in the past.

It's kind of a scary precedent, from now on I guess you can't compare a product to it's competitor's equivalent without getting an infraction.

Anyway, to the point of the thread- I don't really understand why this is such a big issue for nvidia to fix. Don't they have a change-log in their code, so they can see exactly what was changed in these drivers? Doesn't nvida follow basic code version policies?

Of course they do. But that only tells them what changed, no why a change has a given effect. I work in QA, and test a driver that we develop. Trying to track down obscure driver bugs is very difficult if you cannot replicate the issue.

 

[amenx]

Anyway, to the point of the thread- I don't really understand why this is such a big issue for nvidia to fix. Don't they have a change-log in their code, so they can see exactly what was changed in these drivers? Doesn't nvida follow basic code version policies?

I'm sure they do and that may be whats perplexing them and possibly why they're not removing the driver from their site. Didnt you read the Nvidia statement about this driver which I posted above on this page?

Another thing, all the reviews done on the 780 and 770 (43 reviews in May/June so far), have used the 320.18 and none of them reported any issues. Which makes you think, that with the sheer number of cards out there, just a small percentage of those with problems can be enough to set off what may appear to be a tidal wave of complaints. This is not to minimize the real problems many may be having, but it looks very unusual. Cant recall a precedent with Nvidia with something like this. The issue with infamous 196.75 driver was quickly identified and it was quickly pulled from their site. They havent done that yet on this one, though they have replicated some game issues with it (BF3, AC3) which they said will be fixed with next driver. But nothing yet that could explain damaged cards. Only thing I can think of that could damage cards is volts and heat, but it seems they're confident there is nothing in the driver that would cause a surge of volts or heat. What else could it be? Who knows, in the end they may indeed find something and pull the drivers from the site. But it better be resolved one way or another before the next driver is released, they wouldnt want anything bad from it to be carried over.

 

[Erenhardt]

Another thing, all the reviews done on the 780 and 770 (43 reviews in May/June so far), have used the 320.18 and none of them reported any issues.

Which makes you think, that review samples are not the same as retail cards. Differences in core quality may cause the issues.

 

[amenx]

Which makes you think, that review samples are not the same as retail cards. Differences in core quality may cause the issues.

Review cards (or CPUs) may be binned to look good in overclocking, but how does that translate to overall quality? By quality meaning or having relevancy to the driver problems, ie, free from crashes, BSODs, dead cards, etc? How do they separate or identify the good ones from the bad ones in that regard? And what do they do with the 'bad ones'? Sell them to the public?

 

[RampantAndroid]

Installed these on my SLIed 670s last night...no fires to report.

 

[96Firebird]

Has there been any user reports of problems on the 7xx series cards? I think it is limited to 6xx and below, from reading around the net.