Maybe this is the reason why the 8400M and 8600M keep failing?

I just discovered an interesting issue after undervolting my DV9700's CPU and overclocking its graphics chip but forgot to post it earlier. When playing GTAIV (which is extremely GPU intensive as you know), the whole notebook will get warm. I already know from FurMark that my 8400M GS will reach 80C in its current configuration so i can understand why the thing gets warm. However the fan always runs at minimum speed when playing GTAIV...

This leads me to believe that the cooling fan is exclusively controlled by the CPU temperature. It wouldn't matter if the graphics chip would be on fire - the fan will still not ramp up unless the CPU reaches the temperature threshold. I'll check exactly what these threshold values are later, but i remember that on my DV5 it's 75C for high and 85C for maximum. The fans on these things have 4 speeds IIRC.

Now, my DV9700 is a large and rather solidly built machine. It easily copes with the heat even with the fan turned down, i have no issues with GTAIV or any other game, and if you recall it had a faulty fan when i got it yet it was still able to work for 20 minutes before reaching shutdown temperature.

But if this is true for all laptops, i can see why these GPUs fail. Notebook manufacturers should consider implementing a fan controller that reads from both CPU and GPU sensors.

 

Yup. That's exactly how it works. It's certainly contributed to the defective NVIDIA GPU burnout problem. The HP BIOS should definitely monitor both the CPU and GPU thermal sensors as you suggest.

 

FWIW I used to think that my fan was also only controlled by the CPU but later on discovered that I couldn't really hear the fans change in speed until it hit 90% drive. A combination of bad hearing, high ambient noise and quiet fan perhaps.

After testing by artificially injecting temperatures for GPU and then CPU that the ACPI uses for thermal control I got these values.

Code:

 CPU   FAN
<67C   30%
 67C   50%
 75C   70%
 81C   90%
 85C  100%

=========

 GPU   FAN
<66C   30%
 66C   50%
 74C   70%
 80C   90%
 87C  100%

Note that the BIOS is set for fan always on mode hence the 30% drive. Of course different BIOSes even BIOS versions may do something completely different but I thought it worth mentioning because I used to have the same belief for my machine.

 

nooooooooo the reason the gpus fail is that NV cheaped out on the fab process for the chips. No new discovery needed.

 

Mine does NOT speed up when my GPU gets to 80C. In fact it never gets past 50% in any case because my CPU's full load temp is 72C. Undervolting ftw.

Yes i know the whole thing about nvidia's fabrication process, but it can't be that alone. Solder doesn't melt on its own.

 

What I heard was that due to the solder used.
Constant power cycling causes the solder to crack.

 

Yes that's what i heard too. And they are using "eco-friendly" solder which is rubbish and breaks really easily. I use a soldering iron frequently and i know the strength of a soldering joint depends a lot on temperature - the higher the temperature of the iron and the lower the working temperature of the device soldered, the less likely for the solder to crack.

That eco-friendly stuff on the other hand, has a lower melting point than the regular (leaded) solder, and over 120C it already starts turning into paste. It may not be noticeable, however, a mechanical shock at that temperature will definitely crack the solder. Anything over 90C is a threat for this type of solder.

And here's the irony - is eco-friendly solder so eco-friendly anymore if tons of devices are thrown away or come back to the factory because it cracks? I'll keep using the "poisonous" stuff, thank you.