What's wrong with my 780m?

Hello

I have encountered this issue lately. Every time I run something that uses the GPU, when the heat reaches 75c,
the GPU goes haywire, and causes the CPU to do the same as well, until laptop is restarted.

On my last test (one of several with same result) with Fumark, this is what I get:
The following test has been done like this:

Because the system becomes extremely non responsive, I have scheduled a 60 second delayed restart command with
- shutdown -r -f -t 60
- I made GPU-Z to output the data to a file
- and I ran Fumark (GPU stressing tool)

At fumark when the temperature reached 75-76c things went crazy (i.e. Sound lags, Mouse lags, everything lags), I let it stay until the shutdown timer expired and restarted the computer.
http://pastebin.com/raw.php?i=hWBrQFQj

Is this throttling issue?

The reason why I'm sceptical about it being a heat issue is that the heat is not even reaching critical levels... My GPU could handle ~90c without problems... but this issue is new and it doesnt even pass 76c and it goes crazy...
As you can see on the link above, the heat drops down when it reaches 76c and the volts increases and stays stuck... even when the GPU accelerated program stops... The PC will be unusable until restarted, if I manage to restart normally...
Also, I usually clean my laptop from dust on monthly basis...
I wont be able to re-paste the GPU until next weekend, until then I can only do "software" related solutions.

I tried:
- Fresh reinstall with DriverSweeper the GPU driver
- Tried multiple versions of the GPU driver... and even installed a version I used to use without problems...

I've also thought about reflashing the VBIOS but;
1. I didnt have a copy of the original from another laptop.
2. I've only done VBIOS flashing once and it was following an old undervolting guide. So, I didnt want to try this one yet. (Yeah, I'm aware of how to blind flash the VBIOS )

p.s. This is not the first G series for me. I've had G53SX and several other laptops in my life and I've seen throttling before, but this one is unique.

p.s. G750JH

 

I think it's best to backup the data, and then reload the OS clean, or restore to factory defaults. So if you can eliminate this possibility and if problem persists then it can be a hardware failure.

 

Have you logged the cpu temperatures and behaviour?

 

I have done a new test with MSI Afterburner. Told afterburner to limit the GPU to 75c and it was fine but as soon as I increased the limit by one or two (76-77c) the GPU went crazy. Afterburner reported that the core clock is 0 when the gpu started hanging.
I have done a separate test (Prime95) to stress the CPU and it went fine, so it's not the CPU.
The CPU only goes crazy when the GPU does.

 

You should probably link the pictures so we can have a look.

 

Just completed the disassembly's and the assembly of my laptop.
After a small stress test for the GPU, it was fine...

The old thermal paste was still not so "frozen", but I reapplied it with a non-metal thermal paste again and assembled the laptop.
Not sure if the new thermal paste or removing the bios battery fixed the issue, but either way the max temp is now 86c instead of 92-94c

 

A disturbed paste application could have caused hot spots on the core giving you issues.

 

.. this is a kepler card at some .. 90w? I think it's set up so that if you trip the bidirectional throttle temp, the graphics card will drop to the lowest "power state", and the cpu will cut out, over and over again to drop the temp below the limit. So at that point, even if the cooling paste is caked, the temp will be stuck at whatever that limit is set to. While the frequencies tend to go up and down too much to get a stable frame-production - or, you'd get better results by just setting the card to the lowest frequency.

And you typically don't get into that situation unless the cooling doesn't work. So get a screwdriver and some cooling goop, and so on. It's possible the power control isn't working as well, that you don't have the bus and chipset drivers installed properly.. are a few quirks with those on the asus machines. Frequencies dropping, and the fans dropping along with it, that sort of thing, so it's always stuck in panic-mode. But you're only going to see that behavior if you're hitting the bidirectional throttle. Best guess: the cpu has a burnt spot of goop on the side that looks like old bubble-gum, and the goop has been put on unevenly and too thick.