How do you tell a hardware problem from a driver problem?

I have a Compal JFL92 laptop with an 8600M GT that is 2 years and 9 months old. A few days ago, I was playing some silly flash game when the machine crashed -- hard. The screen became corrupted and there was nothing I could do except reboot, but even when I rebooted the display on the POST screen was likewise corrupt. Windows would not boot anymore and Linux would only boot in command-line mode (I have a dual boot system) and even then the display was so bad I couldn't see what I was typing. Since the 8600M GT's are known for being defective and this one has been used practically every single day for almost 3 years, I assumed that the GPU is dying.

I am currently in Europe and I'd need to mail the machine back to the US for warranty service so I tried to make it work again (at least until Christmas when I go home). I unplugged the power and took out the battery and after that Windows was able to boot in Safe Mode. But here is the strange part: I tried to take some load off the GPU by uninstalling all of Nvidia's drivers until it was down to bare VGA. After I did this, the problem practically went away. I am able to boot both Windows and Linux normally and not only that, but I can even use video (with, say Skype) and I played the same flash game as before without any problems for the past hour or so.

So my question is: how can I be sure that this is really the GPU dying rather than the drivers corrupting the memory of the card?

 

try another driver. what was the one you removed? was it the latest.
for the older card like yours i recommend the 186.81 get it here NVIDIA Driver Downloads - Advanced Search
if the problem still comes back they your card has died or on its last legs.

edit: have you cleaned out the fans lately as this will help with temps.
also download HW Monitor to check on your temperatures.

 

How fast was your fan running while doing all of this? Do you have any temp readings? I had a similar card (8400m gs) a while back and I went through a similar experience, but I noticed my fan would go from quiet to full blast in less than a few seconds upon rebooting and my temps would be close to 100 C just by idling.

If you're still under warranty then there is no harm in sending it back to the US for a replacement GPU, motherboard, w/e they feel like doing. In the end

 

If your POST screen is corrupted it is most likely a hardware failure. Your GPU is about to die.

I had an old laptop with nvidia GPU die on me by getting a similar experience when gaming, but it was usable when I uninstalled all VGA drivers for obvious reasons, as the GPU was no longer being used per se. Reinstalling drivers later on killed it permanently and I had to send it for repairs.

 

Yes, it was the latest. I've used DriverSweeper and went back to a couple of older ones that worked perfectly fine for months, but anything other than bare VGA kills it.

Temps are normal (around 72C idle, the same as it was when I first got this laptop). The fan never goes to full blast when it's idle (or at all unless I'm gaming or stress testing which I don't want to do right now).

Yes, that was my thought as well. The POST screen has no clue of what the drivers are, but the corruption goes away if I don't aggravate the condition. I was wondering if it is possible for the firmware to corrupt the memory of the card to the point where it wouldn't work until the memory was cleared (of course, the more likely scenario is that part of the memory is physically damaged and the whole thing will die shortly).

 

I can almost guarantee this is exactly what is going on.