Can a driver kill a video card?

[Carfax83]

Some AMD fans tend to throw this around fairly often on the forum, especially in retaliation for saying that NVidia makes superior drivers to AMD..

But is it true? And if so, how does it happen?

The 196.75 drivers were rumored to cause the GPU fan to stop spinning or working properly thus causing heat death, but I've always thought that ultimately it was the BIOS that controlled the GPU fan, and not the driver. You can use third party software to override the BIOS settings to a certain extent sure, but the BIOS has fail safe's I believe which cannot be overriden unless you edit the BIOS.

Am I wrong in this?

As for the 320.18 drivers, that to me is more hear say than anything. There was never any solid evidence that those drivers bricked any card, other than rumors on various forums.. Personally speaking, I've used NVidia as long as I've been a computer gamer (about 13 years now), and I've never had a driver brick my card..

 

[SolMiester]

Lots of rumors, never saw any links to users that had proven the drivers bricked their card though....

 

[toyota]

it was not a rumor. its a fact that the driver did not let fans operate properly which let some cards overheat. thats why the driver was pulled as soon as the issue was reported.

 

[SolMiester]

Which is interesting in that the cards will throttle when they get too hot anyway, in which case, did any actually die through use of this driver...it was only posted for a day or two at the most!

 

[BallaTheFeared]

Fermi had thermal protection, it would shut the PC down.

Fans stopping isn't enough to kill cards, same as removing your heatsink from your cpu won't kill it either.



Edit: The only other case I saw was the 590 popping at start up, which wouldn't have anything to do with the driver since the driver isn't loaded until windows whereas this pop took place at post.

 

[toyota]

wrong. I guess you guys dont understand that the driver can prevent the protection from kicking in. the gtx590 had an issue with a set of drivers and the drivers had to be updated to keep the card from killing itself in certain scenarios.

 

[BallaTheFeared]

What protection would a stock card need at post?

 

[TerryMathews]

Some AMD fans tend to throw this around fairly often on the forum, especially in retaliation for saying that NVidia makes superior drivers to AMD..

But is it true? And if so, how does it happen?

The 196.75 drivers were rumored to cause the GPU fan to stop spinning or working properly thus causing heat death, but I've always thought that ultimately it was the BIOS that controlled the GPU fan, and not the driver. You can use third party software to override the BIOS settings to a certain extent sure, but the BIOS has fail safe's I believe which cannot be overriden unless you edit the BIOS.

Am I wrong in this?

As for the 320.18 drivers, that to me is more hear say than anything. There was never any solid evidence that those drivers bricked any card, other than rumors on various forums.. Personally speaking, I've used NVidia as long as I've been a computer gamer (about 13 years now), and I've never had a driver brick my card..

Obviously you are wrong. How else can you reconcile that NVidia officially acknowledged the problem via their forum, pulled the driver, and never re-released it?

If there was an "A ha! It wasn't us, it was xxx." don't you think Nvidia would have updated their forum post to reflect it?

I mean it's not great publicity right?

Both sides have screwed up at various points. Why are you so defensive about Nvidia? Why do they have to be perfect to you?

 

[jaqie]

To answer the original question: Yes, a driver can kill a video card, but it is rather rare and requires an extreme derp move by the driver maker, such as forcing the fan to power down and allowing the gpu to build up heat to the point of failure.

 

[Gloomy]

Was it the driver or the vbios? I find it hard to believe that a program that lives in the OS can be made to cause hardware to fail by mistake.

 

[BallaTheFeared]

It seems the 590 was a case of the wrong driver on an overclocked 590. The 590 was pushed well past 580 SLI performance and it killed it due to not having the vrm circuity to support such clocks.

It seems either the driver wasn't designed for the 590 or it didn't work correctly, but the end result was overclocking well outside of specifications wouldn't trip the internal circuits overvolt protection and ended up frying some cards on test benches.

And now we have kepler

 

[3DVagabond]

Was it the driver or the vbios? I find it hard to believe that a program that lives in the OS can be made to cause hardware to fail by mistake.

nVidia "fixed" the issue with the 590 by releasing a new driver, not a new bios. All we can take from that is it was the driver.

 

[BrightCandle]

There was definitely a duff driver that killed cards. They pulled the driver pretty quickly and its not happened since. As far as I know they also replaced the cards of the people who suffered from it. GPUs and CPUs have gone through the process of adding more and more hardware for managing their temperature so nowadays I think its probably not possible to kill a 600 series card in this way, but the 500 were partially software managed.

The difference between the two manufacturers approach to dealing with driver problems is really quite different. You can see that difference today with the number of bugs that impact a large chunk of games. NVidia's bugs are limited to individual games, and they tend to get fixed within a release or two. AMD has multiple priority 1 bugs that impact a number of games and are often completely game breaking without any form of workaround. The bug list I maintain of the two even with the frame pacing driver now out still shows a significant difference in the relative severity. That comes from the attitude of the driver teams to bug reports, not from infallible teams. Nvidia makes bad drivers as well, the latest beta has a bug with 120Hz screen performance in Windows. I reported it and they will take it seriously. The only bug I raised with AMD that has been fixed was the frame pacing one and I am pretty certain it wasn't my bug report that got that one fixed!

 

[OatisCampbell]

nVidia "fixed" the issue with the 590 by releasing a new driver, not a new bios. All we can take from that is it was the driver.

IIRC, what they did was not allow high over volting (which is what burned the cards)

Pretty much what they've done with all their cards since, and angered the "must run cards as far out of spec as possible, and want my warranty too" crowd.

No GTX590 ever burned at stock settings AFAIK.

I think some laptop cards did though, and the question was "Was the assembler at fault for hot cases, or was NVIDIA at fault for not accounting for hot cases?" NVIDIA ended up paying in any case.

 

[Zanovar]

Early cards possibly.these days hmm dont thik so

 

[Gikaseixas]

Well reported and admitted by Nvidia that it was the driver. Every manufacturer has their own low point in history. It could have been AMD/ATI so this should not be an opportunity for an attack from the red side.

 

[BallaTheFeared]

The driver didn't actually kill the card in this case, it was excessive voltage/clock speed that popped the vrms.

All the driver did was prevent smart people from overclocking it properly, or in other words, make it so stupid people couldn't destroy cards and claim warranty.

 

[3DVagabond]

The driver didn't actually kill the card in this case, it was excessive voltage/clock speed that popped the vrms.

All the driver did was prevent smart people from overclocking it properly, or in other words, make it so stupid people couldn't destroy cards and claim warranty.

That was the 590. There was also a driver after that caused a malfunction with the fans and some cards died. I think we are combining the 2 into one which is creating the confusion.

 

[BallaTheFeared]

I still find it highly unlikely a broken fan killed a Fermi or better card.

As was stated they have thermal throttling and will actually shut off the PC if they get too hot.

 

[TerryMathews]

I still find it highly unlikely a broken fan killed a Fermi or better card.

As was stated they have thermal throttling and will actually shut off the PC if they get too hot.

True. Unless the driver flipped off whatever register bit controls the thermal throttling, whether intentional or by accident.

 

[3DVagabond]

I still find it highly unlikely a broken fan killed a Fermi or better card.

As was stated they have thermal throttling and will actually shut off the PC if they get too hot.

Maybe you weren't around then? It's not just rumor or made up. The drivers were pulled by nVidia and people were told not to use them. They were killing cards.

 

[BallaTheFeared]

I'm not saying a driver didn't mess up the fan control, what I'm saying is I've boiled water with my 470s several times and all that happened was system shutdown.

 

[ruhtraeel]

It is 100% possible for a driver to kill a card.

In fact, in my Computer Science Intel x86 class, we were required to write a driver for a computer's gpu; granted, it was an old Pentium 3 machine. The TA warned us not to use certain interrupts at certain times, such as turning up the refresh rate of your monitor so high that it would light on fire.

 

[Silverforce11]

Starcraft II MENU + a bad NV driver release killed a lot of older geforces, the blizzard forum was full of rage. Apparently it was a vsync disabled issue that cause the menu to hit excessive fps (many hundreds), causing older poorly vented (think DUSTY) cards to just die/artifact from overheating.

Too old now to dig up the threads.

 

[zlatan]

It is possible to kill a card without a proper hardware protection system. But a bad driver is needed for it.
The real problem is that the only fully implemented protection system is Powertune. All Geforce and most of the past gen Radeons are use a software based solution.