Finding a short on a graphics card

Joined
Jul 29, 2021
Messages
59
I have a Zotac 960 mini and a R9 280X from Asus with a short to ground - the 960 on vMem and the 280X on the upper two vCore phases. Since there is no visible damage on both cards i proceeded to hooking up my power supply and setting it to 0,9V/1,5V. After slowly increasing the current up to 10A and nothing even got slightly warm beside the cables?
I hooked it up just like for example Eli Tech - ground to ground and positive to for example the coil on the shorted phase. Am I doing something wrong or is the resistance of the faulty part simply to low to get hot and how would I locate it in this case?
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
Ok, first, you can't just say "I've got a short, how do I find it?"

The 960 is probably a lost cause, but let's talk about the 280X. Post a photo of the card marked up to show exactly where you're measuring from, and describe to us what exact resistance measurement are you getting at that spot. Also, mark the exact spot you're attaching the positive lead from the power supply.
 
Joined
Jan 16, 2013
Messages
2,855
Ok, first, you can't just say "I've got a short, how do I find it?"

The 960 is probably a lost cause, but let's talk about the 280X. Post a photo of the card marked up to show exactly where you're measuring from, and describe to us what exact resistance measurement are you getting at that spot. Also, mark the exact spot you're attaching the positive lead from the power supply.
And be sure to PP Razor $10 for his help. :)
 
Joined
Jul 29, 2021
Messages
59
Ok, first, you can't just say "I've got a short, how do I find it?"

The 960 is probably a lost cause, but let's talk about the 280X. Post a photo of the card marked up to show exactly where you're measuring from, and describe to us what exact resistance measurement are you getting at that spot. Also, mark the exact spot you're attaching the positive lead from the power supply.
Firstly - sorry for the late reply and also for the unspecific title. I just wanted to be as general as possible.
I tested the 280x once more and found out that the two shorted phases are the only ones not connected to the others. This makes me believe that those are actually not vCore but VDDCI. That means it's probably not a dead cap and much more a fault in the memory controller with very low resistance - am I right?
Sadly I wasn't able to find any information regarding the position of VDDCI on this particular card but there are no other phases except the one in the bottom left. But I think that's the video rail.

(I probed on the green marked spots.)

*I can't add images for some reason right now but I placed the positive electrode on the inductor of the shorted phase and the ground on the casing of a display output.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
What variety of 280X is it? I'm not sure a reference card exists, and it appears that every brand had its own design.

Edit: You said Asus. Is it this one?
https://www.techpowerup.com/review/asus-r9-280x-direct-cu-ii-top/3.html

If so, the two phases up in the top left are almost certainly VDDCI. The memory, AKA "MVDD," is most likely the top two phases in the big row of them. You can tell because the bottom eight have phase doubler ICs, but there's no doubler for the top two.

Next, again, what is the EXACT resistance measurement that youre getting on the "shorted" phases? You need to post the EXACT measurement you're getting. "Shorted" is not enough information to help you.
 
Last edited:
Joined
Jul 29, 2021
Messages
59
Yes, it's the direct cu II.
The resistance from the phase to ground is smaller than 100m ohm and in between is probably non (as to be expected). I can't tell the exact numbers though since my multimeter is very imprecise. They are also not shorted to ground.
 

RazorWind

Supreme [H]ardness
Joined
Feb 11, 2001
Messages
4,152
Yes, it's the direct cu II.
The resistance from the phase to ground is smaller than 100m ohm and in between is probably non (as to be expected). I can't tell the exact numbers though since my multimeter is very imprecise. They are also not shorted to ground.
Ok, so 0.1 ohms to ground on those two phases? I don't understand what you mean by "in between is probably non." They should be connected together if they're part of the same regulator - you'd have zero ohms when measuring with a DC ohm meter from the positive side of one to the positive side of the one next to it.

If current injection didn't help finding the source of the short, about the best thing you can do at this point is to try removing the low side mosfets from the board and see if that clears the short. Other possibilities include any of the various capacitors on that rail and the GPU itself. Current injection is your friend if you're looking for a shorted cap, and the GPU is generally diagnosed by elimination, unless you have a thermal camera.
 
Top