GTX 1050 Ti extremely low memory resistance

Joined
Jul 29, 2021
Messages
59
I have I a PNY 1050 Ti I want to repair - no visible damage or shorts on any voltage rails beside VMem. The resistance to GND is 1.2 ohms - probably a short I thought. Then I hooked my power supply up at 0.9V and slowly increasing the current but the GPU got hot and all memory modules as well?! It was only 5W so if there was a short the faulty part should take around 90% of the energy, leaving nothing for the rest. I also don't know why the GPU is getting powered, since VCore didn't experience a rise in voltage up on powering the memory rail and Pascal does not use vMem on the GPU as far as I know.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
Hard to say without the card in my hands, but this is probably a case where the failure occurred inside the memory controller portion of the GPU silicon, which seems to be a pretty common mode of failure.

Edit: And you obviously now have a short to ground as a result. A short is a short - it's just a matter of where it is that determines whether you can repair it or not.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
When you say that the components got "hot," how were you measuring this? Thermal camera? Freeze spray? Isopropanol? Your fingers?
 
Joined
Jul 29, 2021
Messages
59
First isopropanol then fingers (as soon as I'm sure there is nothing on fire xD). The gpu got about as warm, as you would expect from a card with vCore but no vMem. The memory chips where only barely noticeable. I also put it onto my test bench (since if the memory controller is dead it won't get worse and if it isn't the problem - it shouldn't hurt for a few seconds). The GPU was getting about as hot as expected but didn't worked either.
If it still gets detected, do you think it's worth the while running a memory degradation test?
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
If you're supplying power only to the memory rail, and the GPU is getting hot, that's a clear indication that you have a failure on the GPU die. If you look at it with a thermal camera, what you'll probably see is that there's one tiny spot that's getting hot. You may be able to detect this by putting a small amount of isopropanol on it, and looking for exactly where it evaporates from.

Regardless, if what you're describing is accurate, I'd be very surprised if this card is detected as a PCI-E device at all, and I'm curious as to how exactly you think you're going to perform a "memory degradation test" on it in this state.
 
Top