"death trigger" in Nvidia's faulty chips?

Hello, NBR folks!

I'm really happy owner of Thinkpad R61i with Quadro NVS 140M onboard. It's almost 3 years old and I'm it's second owner - both of us never encountered any problems with GPU, but it's well known that this is ticking bomb which sometime will explode, but noone seems to know when.
So I've started to wondering what is causing it to fail. I see these options: 1) high temperature, 2) fast drop/increase in temperature, 3) aging, 4) Murphy's law, 5) random/bad luck.
I'd like to keep my GPU in it's current state as long as it's possible. This is why I've modified fan script and made it to avoid GPU temp go higher than 70 C (in my 2h stress test it was at ~66-68 C) just in case it depend on temp, but "I can't sleep" without 100% conviction that it's safe.

If someone has some thoughts about these chips that could help me - share, please. I would really appreciate any advice (except those about switching mobo - I don't have any cash right now and it seems that I won't be able to work in vacations, so I'm stuck with current hardware).

Thanks,
ThinkLover.

 

Though it widely varies I have read that anything over 60C will start the GPU to be dislodged from the motherboard. But basically do everything you can to keep it as low as possible (repasting maybe in order).

 

It's a combination of random and changes in temperature (fast may be worse, but I've pretty sure any will do). As long as you are using the laptop, you will never be 100% "safe" because the temperature naturally changes as you turn it on and off, but you can reduce the risk by minimizing the maximum temperature.

I would suggest not worrying about it too much. The lifetime of these parts varies widely -- some died after less than a single year, others have no problem even after well over 3 years. If yours has lasted nearly 3 years, then it's probably one of the long-lived ones. It will still die sooner or later, but probably later rather than sooner.

 

Thanks!

It was repasted 2 weeks ago. I will try to run it as cool as it's possible, but keeping it @ 60 C in stress seems to not be doable, as it is idling @ ~55.

I will definitely do some more experiments with controling fan.

 

I think the bigger problem with the solder is cooling down rapidly from over 60c to well under 60c. The problem looks to lie in that once the core is over this it tends to heat the substructures it is conected too as well too these or even over these temps. once stress is reduced from the core we rapidly cool off the die side and through the solder joints try and cool the subsrrtucture etc. this seems to cause the cold jiontts requiring reflow.

If the GPU rapidly cools to say 70c but then slowly cools by either ambient or passive cooling this slow setting of the joints with low speed contraction will prolongate the solder connections. This seems to be the plan with bioses that were modified according to nVidia spec for video card life prolongation.

Now this means at idle you may see some pretty high temps from your card and even higher temps before the GPU fan kicks in. This is what we pay for having a defective video card unless you are one of the lucky few that fell into the class action suit and can get a replacement...................

 

As I understand I should *not* cool down GPU too fast...
Hmm... So, I may set my fan script to fastly switch beetewn fan levels when even there is 1 C change - it should slowly cool down whole system, but it will be noisy as hell ;p
Will do some test though.

 

I think the best thing theoretically is to set the fan to max rpm and do not restart your laptop very often. Plus you should not do anything that will increase the temp of your gpu (i.e. gaming).


--

 

Maxing fan speed (according to TANWare's post) would lead to cooling GPU too fast when quiting GPU extensive application (eg. game).
I don't play games but I enjoy watching fullhd videos from time to time

 

Use it until it lasts then buy a new one
If it lasted in the first 2 years than it's fine.

 

I have an Everex XT5000 and my present P7805. Both use the nVidia recomended cooling policy in the latest BIOS. This is to let the GPU go up to apr. 70c at passive then kick in to apr. 60c then after that it is just passive cooling, read that a slow cool down if any. Since at full load the GPU runs over 70c with the fan at max the GPUwhile being stressed stays on.

now if you can keep the GPU at low temps, say 55c or under all the time, go ahead with a full fan setup. If you can't each and every rapid cool down is a killer according to nVidia's own prolongation policy. Again this isn't saying we still will not eventually suffer a failure, it may just be longer before we do...............

 

Max fan speed in stress keeps GPU @ 66-68 C

I've managed to change fan policy - now it kicks fan to max only at full stress and when stress is taken out, it quickly go down to slowest fan level (and let system to slowly cool down).

Now, I just hope that it won't fail