• Competitor rules

    Please remember that any mention of competitors, hinting at competitors or offering to provide details of competitors will result in an account suspension. The full rules can be found under the 'Terms and Rules' link in the bottom right corner of your screen. Just don't mention competitors in any way, shape or form and you'll be OK.

Gigabyte GTX 1070 ITX mini - Intermittent issues

Associate
Joined
28 Sep 2020
Posts
14
Hi Overclockers!

I managed to get my hands on a faulty second hand GTX 1070, I got it working and all was fine until I noticed a strange pattern occurring when the card was powered down.

When I received it, it was covered in gunk - I suspect thermal pad oil, once I gave it a good clean and replaced the thermals all was good. Ran high load games just fine, however it would not post a display if I turned shutdown my PC (restarting was fine). I noticed some more oil appearing (very little but I understand it is not conductive..) gave it a clean with 99% ISO and it would post a display - but the issue came back as soon as I powered down the card.

So when the card eventually posted, through random acts of removing and inserting, I measured the voltages across the inductors and I get the 5v from the very top coil, then 1.3v at the top coil in the row, with 0.7v from the ones below and just above 1v from the very bottom coil. I also get 1.8v from the coil located at the back of the card. Please see image for more details.

I don't think there is anything wrong with the GPU, RAM as it works when it does eventually power on, however I'm led to believe something strange is occurring with the high side MOFSET (potentially low side FETs too but I'm at a loss at what to measure). With random acts of stabbing, I measured the resistance across the FETs (between the blue dots). The very top FET gave very high Ohms (2 mega and 0.4 mega ohms when switching the probes round) and the ones below much lower Ohms (somewhere in the region of 12k and 25k dependant on which way the probes were used). Also using continuality tests, all the but the very top FET had continuity from the pads to the positive power legs, the same occurred for the very top capacitor as illustrated in the image.

Additionally, when checking for continuity on the PCI slot connector, I get 0 Ohms from the 3v PCI connector, but I get a short beep - indicating a quick short (?) on the 12v PCI slot connector. but this eventually goes to 0 Ohms (is there a dodgy capacitor somewhere or are the resistances so low on the card giving a false positive?).

I'm led to believe that the MOFSET crossed in orange is faulty. Can anyone review this and give us a hand fault finding please? I'm aware that I may have gone about this the totally wrong way so happy to be schooled on how to properly test components!

P.S There are no signs of damaged or missing components after close inspection, but aware that there could be a cracked ceramic capacitor?
P.P.S I note on another forum a user and others reported capacitors C175, C176 and C177 could be at fault and have developed an internal crack due to the mounting screw in the vicinity. Mine seem okay visibility, how do I check them with a multimeter without removing them to measure the capacitance? (Ohms go all over the place, then steadily rise and plateau, but this could mean nothing if they are in parallel which seems the case).

UPDATE:

I plugged the card in and after several successful boots and even removing and re-inserting it worked but only once (!). Whilst seated in the PCI slot with no display I measured the following:

L1 = gave 5v (1.6 k Ohms with no power)
L20 = 1.8v (895 Ohms with no power)

L3 = 0v (normally 1.35v for memory, 22 Ohms with no power [this seems low?!])
L5, L7, L9, L11 and L12 = 0v (normally 0.7v for GPU VCORE, 0 Ohms with no power)
L31 = 0v (1.02v normally for PCI/PEX rail, 51 Ohms with no power))

I measured things randomly and noticed that there were 12v, 5v, and 3.3v present on the card.

Does anyone know the order of which the inductors turn on?

If L3 turns on after L20 then MOFSET Q29 could be the issue?
Or if L31 turns on after L20 then I have no bloody clue!


tHjJ4a1.jpeg
 
Last edited:
I take it you have a solution to repair a component if you find the culprit as all the troubleshooting you do is a bit of a waste if not :)

Also I hope you are using a PCIe slot and not a PCI slot. :)

Yes I plan to repair the board by replacing any faulty components, and yes I have been using PCI-e slots :)
 
It looks like you still have some thermal paste on the green GPU package. Also, the middle VRAM chip to the right doesn't look right. This may be simply because it's not as clean as the others, of course.
 
It looks like you still have some thermal paste on the green GPU package. Also, the middle VRAM chip to the right doesn't look right. This may be simply because it's not as clean as the others, of course.

Thats not a picture of his card its nicked off Toms Hardware, follow my link above and you will see.

@Kiran Parmar post a picture of your actual 1070 Mini PCB so we can inspect for damage.
 
Thanks for the replies so far, and apologies for using a stock photo - my camera is not the greatest. Hopefully the images below show some detail. If not, let me know and I'll take closer ups of any sections.

BTW - the pitting as shown on the fet's is my fault due to aimlessly poking around with probes :)

Gallery view: https://imgur.com/gallery/b5Akq63

Front: https://i.imgur.com/qvn3xOu.jpeg
qvn3xOu.jpeg

MOFSETs (high side?): https://i.imgur.com/9UekrUB.jpeg
9UekrUB.jpeg

5v coil: https://i.imgur.com/e7ja18o.jpeg
e7ja18o.jpeg

Rear: https://i.imgur.com/2WlWiI6.jpeg
2WlWiI6.jpeg

MOSFETs (low side?): https://i.imgur.com/PUZB5JO.jpeg
PUZB5JO.jpeg

1.8v coil: https://i.imgur.com/oY238cO.jpeg
oY238cO.jpeg
 
Last edited:

To right, U3 seems to have some damage. Also there may be discoloration to the left of the washer and above U523.

I'm led to believe that the MOFSET crossed in orange is faulty. Can anyone review this and give us a hand fault finding please? I'm aware that I may have gone about this the totally wrong way so happy to be schooled on how to properly test components!

Take a look at the Adamant IT Youtube channel for how to test and replace at the component level.
 
To right, U3 seems to have some damage. Also there may be discoloration to the left of the washer and above U523.



Take a look at the Adamant IT Youtube channel for how to test and replace at the component level.


Thanks for your reply Quartz,

As for U3, are you referring to the gold colour? Closer image: https://i.imgur.com/Cy1vTec.jpeg I note the white stuff near C1125, which came off easily.

And I am having difficulty locating the discoloration above U523, closer image: https://i.imgur.com/JxtAwRf.jpeg


UPDATE: I left the card in a PCI-e slot overnight, turned on the mobo and it posted with a display. No doubt the issue will re-appear, could there be some sort of residual electrical resistance somewhere?
 
U3 is a vbios chip. It is 1.8 v and if u have 1
8 and card posts , vbios is working.

22ohm for memory sounds reasonable.

It is nierly Impossible To determine intermittent fault.

Your best course of action is make it fail, then start mesuring all voltages on all the coils.
If all voltages are present then it is realy bad.
U can run mats mod to test memory.

If u want more help and learning how GPUs works pm me I'll invite u in a few Discord channels were u can learn tons.
 
L3 = 0v (normally 1.35v for memory, 22 Ohms with no power [this seems low?!])
L5, L7, L9, L11 and L12 = 0v (normally 0.7v for GPU VCORE, 0 Ohms with no power)
L31 = 0v (1.02v normally for PCI/PEX rail, 51 Ohms with no power

Vcore is down, memory wont start pex wont start.
Gpu chip should Stay cold.

U need to investigate Vcore vcc and enable signal.
Hook me up in private message, I'll contact U on Discord after work tonight. We can run some diagnostic.
 
U505 locate data sheet Online and check if this chip have enable signal.

Thanks for your replies @Benass I was literally watching this youtube video with diagnoses/ fault finding very similar to what you are describing when you posted! :eek:

L3 = 0v (normally 1.35v for memory, 22 Ohms with no power [this seems low?!])
L5, L7, L9, L11 and L12 = 0v (normally 0.7v for GPU VCORE, 0 Ohms with no power)
L31 = 0v (1.02v normally for PCI/PEX rail, 51 Ohms with no power

Vcore is down, memory wont start pex wont start.
Gpu chip should Stay cold.

U need to investigate Vcore vcc and enable signal.
Hook me up in private message, I'll contact U on Discord after work tonight. We can run some diagnostic.

Aha, so its 5v, then 1.8v (GPU), then 0.7v (VCORE), 1.35v (MEM) then 1.02v (PEX). I read somewhere high ohms for a mofset is also reasonable, so for now Q29 is not the suspect, and the continuity testing I did at the beginning is incorrect since it's a separate voltage/circuit. Learn something everyday.

I'll get on with checking out U505 schematics later this eve and send you a PM. Appreciate it!! :D
 
Up9511 most likely. Datasheet Should be easy to find.

Tech cemetery, he has a Discord Channel .
I hang out in there.
U can pop in, ppl will give u a hand in any questions U have.
 
@Benass Sorry for the delay, the kids aren't too well!

The card cycled in a working state for about 10 soft and hard resets, it was until I removed it from the PCI-e about three times I managed to get it to fail! :confused:

Anyway, measured pin 36 from the UP9511p IC (traced to R588 for ease of probing) and it gave 2.41v, both when it working and not working. Did some tracing to find pin 36 enable pin was connected to Q519, Q526 and U506 but this was through GND (oddly, to me anyway!). Please see diagram below for measured voltages.

Interestingly, I found U506 pin 1 to give 3.3v when the card worked, but 2.86v when not working. Then I read somewhere Q521 could be providing some sort of signal through voltage difference, where 10.3v is not low enough - I could be talking out of my ass now.

Anyhow, I'll send you a PM.

qqvM9bN.jpeg
 
Back
Top Bottom