Driver killed GPU/VRAM

[Mad]

I was wondering, could you actually kill a GFX card by installing a new driver (which was not really meant for the card)?

My story:

I installed the 169.28 driver on my GeForce Fx Go5200 (which resides in my marvellous (at least when I bought it) Toshiba M30-344 :) ).

Of course it does not work "just like this", so I took Larry's INFs for Dell 5150/5160 and modified it to accept my 5200.

Everything woked fine for a week or so. (Actually more than fine, this driver boosted my FPS by 100% in comparison to my 44.46 driver)

This morning, I switched on my Laptop, and I got these nice picture disturbances (even on BIOS level) you get when there is something terribly wrong with the VRAM. Only thing I can think of, which might have killed the cat is the "sideband addressing" which is switched on in Larry's INF. But could that actually fry my GFX card?

Any comments on this would be appreciated - I ordered a new mainboard for my M30, and I don't want to fry the next GPU...

Thanks! :P

Ps.: Maybe I should take the time to say hello. So, hello! :)

Thanks for this great spot on the net. Took me quite some time to find it, but I'm pretty impressed. :P

Edited February 7, 2008 by Mad

[Guest Jesus duarte]

I dont really know... but i too have that theory.

It happend with my ATI Radeon card on my Gateway laptop when i installed the OMEGA drivers.

It worked awesome for about two weeks.... BUT

After I installed Safary for windows... about two hours of use made my GPU to crash and turn to software rendering because of a hardware deadlock. But... safary ain't video intensive... so... i discarded that option. Anyways.. i uninstalled safary. SIDENOTE: accelerated graphics programs (games) ran... but had this weird missing segments on the image generated and crashed frequently.

Problems seemd to be gone... till two weeks later that my laptop had another hardware deadlock. After that, hardware deadlocks were more frequent and windows turned to software rendering. Sometimes it couldnt recover and windows had to force a restart.

The funy thing is that when my screen was at 32bit coloring, normal copmuter use (word, internet, windows desktop) seemd to look fine (except that navigation and display of programs seemed kind of slow). But at 24bits or 16 bits, there were awful vertical segmented lines running through the whole screen.

The removal of the OMEGA drivers did not fixed these issues, nor the installation of original Gateway drivers.

I took the laptop with a technician, and the diagnostic said that it had nothing wrong. (i wont ever go to that store again).

The issues persist, but laptop doesnt crassh as long as it doesn't get extended periods of use. I currenlty use it for quick email checks, or quick internet access.

[Mad]

Hm. Interesting. That raises the question if drivers do set a clock frequency for the GPU.

I know that some Laptop manufacturers clock their GPUs down to allow their cooling hardware to cope with the heat.

Could a driver which was not originally meant for that special model change the VRAM or GPU clock to a "normal" level by default, thus sending the GFX card to its termal death?

edit: I guess I'll instal the old drivers as soon as my laptop is up and running again and read out the clock settings with GPU-Z - Riva Tuner provided some strange readings in the past (now that I come to think of it, especially after installing the 169.28 drivers...) and the nTune Sensors page always crashed on me (BOD)...

Edited February 7, 2008 by Mad

[mobilenvidia]

The modded INF here do NOT include and tweaks to cause a GPU to burn out.

No performance enhancing settings, no voltage or Clock speed modding.

Nothing but the ability to install a driver and then have as many settings available.

Larry has been using his INF for what must be centuries now :)

With no ill effect, he probably wouldn't mind if something happened so he can upgrade with an excuse.

1000's are using his modded INF for the go5200.

Hope this explains it.

[Mad]

Thanks. So this is just a strange coincidence then?

[mobilenvidia]

I would think so, or there would be many others with the same ailment.

[Mad]

That of course is true... :)

But a strange feeling remains. Mainly because I enabled sideband adressing. Oh well, will toy around with this again when I've got the new mainborad...

[Darmdorf]

:mad

As our Forum host already stated, a driver on its own cannot willfully misbehave and "destroy" your hardware. The clocking values are fixed within the cards video BIOS, but CAN be made alterable if the driver was installed with certain features enabled in the INF at installation. Thus the door is open to misuse by the user, but not by the driver's own default settings. As you know there are also tools (like the Nvidia control panel) which will also overclock your graphic hardware. The issue you mention of sideband-addressing is indeed harmless, and is merely a performance feature. It usually is NOT supported by most chipsets, and is basically akin to allowing the farmer's cows to try and enter the barn 2 abreast when there is really only room for one at a time. If enabled it merely tends to make the system unstable, (i.e. may crash), but it certainly won't fry anything.

Having problems of a strange nature as you've mentioned, can however result when the "patient" is unhealthy to begin with, and the new drivers & settings entice the owner to raise the RPM limit now that things seem to run so nicely. The result can be thermal instability, and in the case of certain Toshiba models I have read that the manufacturing process chosen can lead to degradation of the graphic chip contacts over time. This deterioration would naturally be negatively influenced by any higher ambient heat levels.

Hope that helps.

[Mad]

Thanks for the clarification. Indeed, I never heard or would've thought that a driver could actually damage the hardware, regardless of the environmental settings. I do have some experience in hardware, but well, the presented case just puzzled me.

But the explanation involving the degrading contacts between GPUand MB does sound interesting. Thanks for the insight :)

[mobilenvidia]

As the new giant "news' post now can vouch for.

My GPU nearly died I managed to bring it back from death.

My predicament was not driver but prossibly fan related, it won't spin at full speed, and then the temp goes up.

At some point it mus have gotten too hot and the memory chips must have been close to melt down,.

This is not drver related a eensor on the video card might be faulty.

The above could happen at anytime and just plain cook the GPU or Memory.

[Guest Kenneth]

Thanks for the clarification. Indeed, I never heard or would've thought that a driver could actually damage the hardware, regardless of the environmental settings. I do have some experience in hardware, but well, the presented case just puzzled me.

But the explanation involving the degrading contacts between GPUand MB does sound interesting. Thanks for the insight :)

After installing a new nvidia driver (156. somthing) for my XPS M170 my GPU would overheat to the extend that it went

past the threshold and BIOS would turn off my laptop. I installed the old driver (84.64) and everything was fine again.

I played with this for 6 months atleast. Temperature was max 102 degrees C with the old driver.

Some time later Dell released a beta driver made for the XPS series

machines and I installed this (163.75). I had the excact same problem. Overheat - BIOS shutdown.

This time I started getting artifacts on my screen - in Windows strange horizontal lines.

In post letters all over the screen. Temperature accordingly to i8GUI - well I saw it at 110 degrees C.

You can see pics of it here:

http://www.kenneth-iversen.dk/gallery_test...p?g2_itemId=820

It turned out my cards RAM had died.

The fans were checked. Totally clean so it wasnt cause because of lack of cooling.

With i8GUI the fans were even forced to full speed all the time with no difference.

Well... to me it seems like the driver changed something that pushed the card over its

limits and it got hotter.

Since the first driver I installed was downloaded directly from the Nvidia site it could

not have been altered in any way. I didnt alter it either not even with a new inf file.

I ofcourse had contact with Dell support but I never really heard anything of them

after I contacted them. At the time Nvidia would not give support since it was

a driver made for Dell so I have no idea what could have caused it.

I have a refurbished card now from Dell and I still use the old 84.64 driver.

The temperature still goes up to 102 degrees C and I am guessing this

is normal for the card I have (7800GTX)

So... my experience is that a driver can change the temperature and

destroy the card. I cant explain why.

Best Regards

Kenneth Iversen