GeForce GTX 1060 short on PCIe 12V

[Traceless]

Hi everyone,

I've been trying to find the cause for a short on 12v (~20 Ohms to
ground) on a GTX 1060 for a while now, The card has no branding but
as far as I can tell it is a Palit model. The short only shows
up on the PCIe-connector but not the PSU connector.

Initially I suspected a faulty MOSFET in the memory power supply.
However after removing most of the MOSFETS/integrated power stages
from the board: Q16 (SM4377 CS70X), Q17, Q18 (4503NH CS740), U10,
U12, U14 (SIC632) and the PWM controller (NCP81174) the
short is still there. As a last resort I decided to inject .7 V into
the 12V rail on SC87, hoping that a faulty component would heat up.
Unfortunately the card only draws ~40mA, not enough to produce any
significant heat.

I already took a look around here on the forum to find some
schematics, unfortunately the Palit GTX 1060 cards seem to differ
a lot from many other GTX 1060 models and use a different naming
scheme for the components on the PCB.

Since I've run out of ideas on how to tackle the problem
systematically I hope that someone here has some advice how I could
proceed cleverly to find the faulty component(s), without resorting
to randomly removing/replacing components until I get lucky.

Disclaimer: I realize that probably more info is needed to help me
with the problem. If you need any additional data/info/measurements
to be taken let me know which data is missing and I'll gladly try to
provide it.

Cheers
Traceless

[ahmad9794]

12v is usually caused by mosfets make sure again if you properly checked the mosfets. Check datasheet and connect one probe of multimeter to gate of mosfet and other to ground on gpu.

[Traceless]

Hello ahmad9794,

thanks for the response. That was my first thought as well. As you can see on the first two pictures I have already removed all MOSFETs responsible for supplying GPU and memory from the board, unfortunately the short is still present.

Since the board only draws ~40mA I suspect that some smaller component maybe a capacitor, inductor or sth. like that went short. However so far I'm unable to locate the source of the problem.

[MasterCap]

I would not think a 1060 would use 12v from the pci-e slot, on any large phase, as in the large ones around the main chip.

I think 12v-mobo is used a lot for 5V to the video outputs, maybe the fan, maybe for some aux rails

I'd look at the similar schematics if I knew where they were

Start at the 6pin connector, and see what big caps are connected to it, and the upstream side of the big power mosfets. Do the same for the mobo connectors.

That's what I would do anyways. I just fixed another power issue I caused on my GPU, but now I'm back to no video, but all the main voltages are running.

[Traceless]

Today I finally found some time to take a look at the board again.

Originally Posted by MasterCap

I would not think a 1060 would use 12v from the pci-e slot, on any large phase, as in the large ones around the main chip.

Thanks for the hints MasterCap. I assume the card might use PCIe 12V for the memory. Without schematics there is no way to be sure though. I think Q16, 17, 18 may belong to the memory power supply whereas U10 ... U14 belong to the GPU supply (U11, 13 were unpopulated though, probably the same board is used for 1070s as well, which require the additional phases).

Originally Posted by MasterCap

I think 12v-mobo is used a lot for 5V to the video outputs, maybe the fan, maybe for some aux rails

I checked the HDMI output there seems to be a short on TDMS CLK+ (see this video), note sure though if that has to do with the cause of the problems or if it is just a symptom.

Originally Posted by MasterCap

I'd look at the similar schematics if I knew where they were

Start at the 6pin connector, and see what big caps are connected to it, and the upstream side of the big power mosfets. Do the same for the mobo connectors.

That's what I would do anyways. I just fixed another power issue I caused on my GPU, but now I'm back to no video, but all the main voltages are running.

Unfortunately I don't have schematics and there is funky stuff happening all over the board. I made a video showing some examples of these problems. Sorry for the bad image quality my camera is not the best, so here is some additional explanation of what I'm doing:

First I test two caps and an IC. Both caps have low resistance to ground on both ends (~.25 Ohms on the actual ground and ~.65 Ohms on the other end).

Next I measure a capacitor on the other end of the board (SC87 c.f. pictures from my first post to be able to see the component names). SC87 seems to be connected to PCIe 12V (I soldered a wire to it for voltage injection). Both the cap and the PCIe connector have < 20 Ohms resistance to ground.

Next I measure C67/C60/C59 both measure also < 20 Ohms to ground as does the cap below U14 (above Q17). In contrast the according caps below U10 and U12 measure ~500 k Ohms so I guess no problem there.

I do see a lot of problems but still have no clue what is going on exactly.

[Traceless]

I think I might have finally found two suspicious parts, SC532 and SU5 (see picture below). Using Diode mode I measure a drop of 0.000V between pins 2 and 4 of SU5 and both sides of SC532. Can anyone identify part SU5 on the picture below? SU5 is labeled "08=R3M".

[Traceless]

Originally Posted by Traceless

Can anyone identify part SU5 on the picture below? SU5 is labeled "08=R3M".

I think I might have identified the part, I'm not 100% sure but I think it might be an RT9403EGV8

[ktmmotocross]

Originally Posted by Traceless

I think I might have identified the part, I'm not 100% sure but I think it might be an RT9403EGV8

yes

[harp]

20 ohms, 40mA?
you may try this: put pcb on expanded polistyrene surface, connect 0.8v on "short" (btw what is voltage drop at 40mA?) and leave this about half hour on peace. Then use a NTC 10k flat sensor (from some liion batt) connected to multimeter, and scan surface that have accumulated heat spot.

[Traceless]

Originally Posted by harp

20 ohms, 40mA?
you may try this: put pcb on expanded polistyrene surface, connect 0.8v on "short" (btw what is voltage drop at 40mA?) and leave this about half hour on peace. Then use a NTC 10k flat sensor (from some liion batt) connected to multimeter, and scan surface that have accumulated heat spot.

Thanks for the tipp with the NTC, harp. Unfortunately the current draw was just to low to get any conclusive hints based on temperature and the card ended up in my drawer for quite some time. Yesterday I found some time to look at it again and finally found the fault. A ceramic capacitor (SC532) caused the short. I removed it and now 12V to ground resistance is back to 4.9kOhms, which seems healthy to me.

I also took of another ceramic capacitor next to C113, which has a thin black line on top, could be a surface scratch or a fissure. Either way I'd like to replace it just to be sure. It doesn't read short but I'm not sure if its capacitance (10 nF) is correct.

Also during voltage injection I accidentally damaged SC87. I soldered a wire to it and put some pressure on the board while it was lying flat on that wire - so one end broke of and I have to replace that one as well.

Since there seem to be no schematics for the board and the component numbering is very different from that used by other 1060 designs I was hoping someone could tell me the correct values for:

SC87
SC532
the missing SMD ceramic capacitor next to C113

I included pictures for reference.

[harp]

The manufacturer wish to lower cost as posible cost of his products, so it use only few common components on pcb. You may assume that other capacitor with same shape, size, and color be the same value, and if is at same arrangement or function it is even much possible that are the same.

The c532 is near a u5 chip... so find same size cap near other similar chip, desolder it and do measurement of them,
and c86 is in arrangement with near small cap, so find on pcb something simmilar... I think it is not critical value, if is on rail

Or you can trace capacitor function and value in datasheet of RT9403EGV8 if is contain schematic.

[Traceless]

Thanks for the reply harp. According to the RT9403 datasheet could be 100 nF. Eyeballing the cap sizes is definitely plan B. If anyone knows the exact sizes or has schematics it would be great if they could post them here. If not I'll do it exactly the way you suggested.