Hey there everyone, I had a question for some of those experienced with fixing video cards.
So my aging GTX 460 1GB (Axle-brand) suddenly died while playing a game the other day. I've tried it on two different computers that have worked with the video card in the past and the same result happens:
When starting the computer from a cold boot, the computer spins up for a fraction of a second, and then shuts down. When I try to start it up again; nothing happens. The computer is getting the 5V fine, since the motherboard LED is lit up. The computers both work fine with another GTX 460 I have and both systems work fine with an AMD card too.
Since I have tried the video card on more than one system, including my one system with an 850-watt seasonic PSU, I'm fairly certain it is the video card. The video card was not overclocked by me when it died and heat/noise had never been a problem before. It is a model of card that is designed to be overclockable though. I've owned it for about 6-months now without issue up till now; I bought it used, but it seemed in great condition. Unfortunately RMAing the card now would cost me as much as the video card is worth... no receipt and Axle requires me to pay the shipping and parts to and from Hong Kong or something. eek!
In my attempts to figure out what might be wrong with it, I've cleaned the graphics card of dust and I have visually inspected the card for signs of component damage; which there seems to be none. I did notice though that the graphics card and system would start (no display of course) if I plugged in just one PCIe power cable, but only in one of the particular PCIe power ports on the video card. If I try plugging the power cable into the other PCIe port, the computer does not boot again.
Obviously, if I could run the system with just one PCIe power cable attached, I'd be content to run the card undervolted and underclocked, but it really does seem as if you need both plugs attached to get the card to actually do more than just power on. I'm not sure of anyway to bypass this requirement?
Anyways, I'm assuming, because of the PCIe plug issue, that the problem is related to either a voltage regulator or maybe a capacitor, and not the GPU chip itself.
Here are some high-res photos of the card, front and back:
While I do have a basic soldering gun and analog oscilloscope, I'm not confident enough in my ability to just start burning holes randomly to see if I get lucky.
I am hoping for some advice or for an idea of what may be wrong with this video card, and how to fix it before I try and just make things worse...
So I've been looking online some more and I found one other person with the same issue as me so far, but no solution. We both have come to the same conclusion though -- that it is a voltage regulator issue. A mosfet ic shorted I guess, but due to the PSU's short circuit protection, the chip hasn't been able to explode and leave any visual clues for me.
Anyways, I've pulled out my multimeter and tested each of the two dozen 3-legged mosfets/ICs . There are 4 mosfets named MDD2605 and 8 MDD2601's, all by the two PCIe power inputs.
One of the MOSFETs, the one right next to the PCIe power plug that has been giving me issues, is short circuited. Now, without taking the IC off the board, I can't be sure if it is dead or if another chip near it is dead that is creating the short, but all the other MOSFETs are not shorted.
Testing the PCIe power plug, I can confirm that 5 of the 6 pins are shorted; that should not be the case obviously, so I think I found my problem
I assume the MDD2605 is a 30V 9mΩ DPAK low voltage MOSFET of some sort, with high switching, but I have no idea for sure if that is the case. I also can't seem to find any place that is in english online that sells one.
Dare I just stick on something close to my guess and hope for the best? Any suggestions? Would just cutting one of the legs of the MOSFET, and breaking the short, be enough to get the card working again? Would 3 of the 4 mosfets working be enough to power the card do you think?
More I play with this card, the more likely I will break it. heh
So I was able to resolve this problem, although I am left a bit wary of my solution.
I hunted around for a spare MOSFET of equal or better value to replace the one on the video card. I did find one close on an old dead LGA775 motherboard. 25V (vs 30V), n-channel enhanced MOSFET with a 4mΩ rating. Overall, it seemed a better MOSFET and about the same size.
I tried to get it off with a 30-watt soldering gun -- no luck at all there. Practiced on some other MOSFETs first. Then pulled out a heat gun that I have for stripping paint off walls and went to work with that. 30% chance I could get the MOSFET off in one piece. The other times I either got it off with a leg missing, sheered it in two, or took off the PCB with it.
Anyways, I got lucky and got it off. Tried the same thing on the video card though (I masked the video card with aluminium tape/foil btw), but I took off the PCB copper pads with it. WHich is bad, cause now I can't replace it.
Anyways, I figure though that the video card can run on 3 out of 4 of its MOSFETS if I don't push it. I already indirectly knew it wouldn't run on just 2 though... I tested first for the shorting issue however using my multimeter to be safe; everything tested out fine. I put the video card back into the system and booted it up.
System booted!!! Video card works.
Just to be safe, I am undervolting this card down from its 1.00V stock to 0.825V .. and then knocking down the clockrate to something stable. ~500mhz. It is a far cry from the 900mhz it can go up to, but at least I still have a video card.
I guess I could always try to solder on new PCB pads or replace some of the other MOSFETS with higher quality ones, or maybe even try stacking one of them, but that seems like a challenge that I'm not sure is worth it.
Anyways. So, IF your computer is not starting at all, and you know it's the video card and not your PSU, use a multimeter to see if the 6-pin power input port is shorted anywhere on the card. ie: if 4 or 5 pins are shorted, you have a problem -- only 3 should be shorted at most. If this is the case, test the MOSFETs (3 legged black things near the power plug) on the board for a short. Sometimes the MOSFETS are under a heatsink. They should not be shorted. If you find one that is, cut the legs off of it so it becomes redundant or better yet, replace it with a good MOSFET -- you might want to take it to a repair shop to get the job done right though. Test the 6-pin power input again then for any shorts. If no more shorting, try it out! If your board isn't too power hungry, it should load up!
as always, remember to wear protective gear and use common sense when applicable, like remembering to apply thermal paste to the GPU chip if needed.