Short circuit on GPU

[klazaridis1]

My pc shuts off itself, when I was playing League of Legends, I power it on again and when the game started it shuts off again. I check the drivers and download the most recent. I started a new game and it was working fine. I was curious and I test it with the furMark ( stress test on the GPU). Everything was fine and when I closed the furMark my PC went off again... It didn't power ON again, I found out that the GPU was shorting the power supply of my PC. I thought that the MOSFETs of the card was fried because everything was shorted (diodes, MOSFETs, etc) I removed all the MOSFETs that were shorted on Drain and Source but the short (7 ohm) remains. I can't find any diagram of my graphic card, so can anyone help me? My graphics card is an AMD R9 200 SERIES (VGA ASUS RADEON R9 290 R9290-DC2OC-4GD5 4GB GDDR5 PCI-E RETAIL). I don't know what to check next, nothing seems damaged.

When I plug the 8 pin Molex connector my PSU shut off itself.

[nctnico]

First check the MOSFETs you removed for shorts or open between source and drain in order to verify these are actually broken. The square chips labelled PQ are probably also switching PSU chips. Anyone of them can be broken as well.

What you can do is use a lab PSU set to 10V and connect it to the 12V supply of the graphics card and set the current limit to 1A. Then use a DMM (at least 4.5 digits) to measure the voltage between ground and 12V supply points (capacitors) on the board. The short is near where you measure the lowest voltage.

[amyk]

You can also use a thermal camera to locate the shorted component if you have one, or spray IPA on the board and see where it evaporates the fastest. (If this was a few decades ago, I would've suggested using something nonflammable like R11.)

[Someone]

1.2V power rail, 100-200W card.  What resistance would be expected on that?

[kripton2035]

7Ω on a GPU is not a short indeed...
you should build a tool like this one : http://kripton2035.free.fr/Projects/shorty-display.html

[AndyC_772]

1.2V power rail, 100-200W card.  What resistance would be expected on that?

I don't know. What is the expected resistance across a complex CMOS device when it's not actually being clocked?

I can't say I've ever measured it on a device as complex as a GPU. Presumably it's about a billion times the leakage current through any individual gate, which is anybody's guess.

[Someone]

1.2V power rail, 100-200W card.  What resistance would be expected on that?

I don't know. What is the expected resistance across a complex CMOS device when it's not actually being clocked?

Well it has been measured if you could be bothered to do some basic searching.

... thats about as useful as your post.

Easy to raise issues and pretend like you are clever, but you don't put in the tiny effort to answer your own question (barely more effort than posting)

[klazaridis1]

Shouldn't the drain and the source have infinite resistance? The MOSFET acts like a switch so If it's already connected what's the point?

[Buriedcode]

As the OP has discovered, there's a big difference between measuring components in-circuit and out of circuit.  Measuring parts in-circuit requires you to know what exactly that circuit is, or at the very least for repair purposes, a known working unit of the one you're working on to compare it to.

AndyC's point is that GFX cards are remarkably complex, and measuring resistance between two random nodes, on a non-passive circuit designed to switch three figure watts without know exactly what that circuit is is somewhat futile.

As you have already removed the devices in question, you can at least measure these.

It's going to be hard to clean up those pads and solder on parts (either the originals if they are fine, or new ones) because those boards are made for heat transfer, and so may require preheating.

I've only repaired two graphic cards in my time, and both times it was a painful experience.