RTX 2080 short or something else?

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
So I've got an ASUS ROG Strix 2080 that doesn't work. The LEDs on the housing come on but there are no other signs of life and I thought this GPU would make a good project to learn electronic repair - I've been trying to get into this for some time.

I could really use some help as I think I'm hitting the ceiling of what I know (very little) about electronics.

As far as I can see, there are no physical signs of component damage, and whilst the card LEDs do turn on, the fans don't spin and I'm not able to POST. The GPU does not get warm so that's somewhat reassuring.

So far I've been able to determine that none of the 12V rails are shorted to ground nor to Vcore or Vmem. There are no boardview type images/documents I can rely on for identifying rails, but an initial voltage test revealed 12V coming through to three inductors on the right side of the board (labelled) as well as the bottom left, as well as 3.3V in the bottom left. 5V were identified at the top (labelled), and one of the inductors appears to short to ground (approx. 0.2-0.4 ohms).

The part I'm not so proud of... after a long exam today I decided I would come home and jup right into diagnostics... and that's where things went a bit sideways. I tried measuring across a small capacitor (don't ask why I thought this was a good idea) and ended up touching another component (a resistor I think) at the same time. This caused the PCI-e extender ribbon to smoke (I ended up burning the first two pins (12V) on the back and front of this ribbon. I used a riser cable as my current tower isn't large enough to house the bare GPU and try to test voltages etc very well simultaneously.

There's no identifiable damage on the GPU pcb. I decided to hook the GPU up to the motherboard directly, and the PC would no longer start. Knowing I probably shouldn't, I tried to restart the computer which caused a spark across what appears to be a sense resistor adjacent to one of the Vcore phases - nothing else appears to have changed with the GPU in terms of shorts to ground (the original unidentified power rail is still shorted to ground) and there doesn't appear to be any 12V shorts to the Vcore either. I've got a new Corsair RM850 so i'm hoping it kicked in quick enough to prevent any serious damage to the board.

I haven't connected power since in case it burns the card outright, and I'm having difficulty narrowing the potential original short to the potentially culprit component (I'm hoping it's not a short within the PCB) as well as trying to figure out if I've now introduced a new problem (I suspect it is related to the 12V power supply from the PCI-e as that was what was blown on the riser cable I was using).
I can't seem to find any datasheet about the powerstations/MOSFETs to know where I should be testing and my current attempt at trying to hot air the one off the board closest to the sense resistor that blew hasn't been successful (I will try again tomorrow).

I feel like there's more testing to be done before I completely throw in the towel, and with any luck my little mishap hasn't caused any irreversible damage. Any ideas where to go from here?
 

Attachments

  • back_full.jpg
    back_full.jpg
    473.3 KB · Views: 0
  • s-l1600.jpg
    s-l1600.jpg
    183.4 KB · Views: 0
  • front_full - id.jpg
    front_full - id.jpg
    522.3 KB · Views: 0
Last edited:
Joined
Nov 18, 2011
Messages
601
“Any ideas where to go from here?”

To the store to get another card, fire extinguisher, and an electrical engineering book.

This is well above my pay grade but my guess is that when you shorted the resistor it may have damaged a PCB trace which is now causing it to spark. I had that happen once. If the power got dumped into the VRM, then you may have fried the chip itself. But I’m guessing here. We need one of [H] engineers to give you a good answer.
 

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
Thanks, to be honest this is more of a learning thing for me at this stage, and we're still probably several months away from reasonably priced and available GPUs.

I am a bit worried about power getting dumped into the VRM as you said, I thought I'd find a short circuit to verify this, but all 12V rails show decent resistance to the VRMs so I have to assume there's no short directly to the VRMs.

I'm tempted to run power to the minor rails and check for voltage dropoff...not so keen on running power to ay 12V rails just yet.

The resistor that blew appears to be a shunt resistor...I can't quite figure why it would blow as if there's a short when there's no identifiable short from the PCI-e 12 V rail that appeared to be the source as evidenced by the damaged 12 V PCB traces on the riser cable when my mishap occurred.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,808
I have some experience with repairing graphics cards.

If the system won't power on at all with this card plugged in, you have a short to ground. You said you don't have a short on the 8 pin connectors, so that leaves the 12V input from the PCI-E slot. Check for a short to ground on pin 2 of the slot connector.

The shunt at the bottom of the card that your picture says is "smoked" is most likely wired up to the 12V input from the slot. You will need to figure out where the short is in that circuit. It's hard to say without the card in my hands to test, but if it was shorted for long enough to visibly damage a PCI-E riser, chances are that the damage is actually in the 12V power plane itself, which means the card is beyond saving. Nevertheless, it would be worth removing the suspect shunt and checking to see if that clears the short. If it does, then there's a chance the damage is a component situated between the shunt and ground, such as a power stage or a capacitor.

Also, do not use short detection ("beep") mode when troubleshooting. There are circuits on this card (such as the core and main memory power) that measure as only a few milliohms to ground by design. The core should measure about 300 mohms, as I recall, and the memory should be maybe 10-15 ohms. Both of these are well below the threshold that most multimeters consider to be a short. The proper way to check for short circuits on a graphics card is to use regular ohms mode, and compare your readings to the known-good values.

The thing I would do next is remove that shunt, and see if that clears the short on pin 2 of the PCI-E connector. Do you have access to hot air? Do not attempt to use a soldering iron.
 
Last edited:

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
I have some experience with repairing graphics cards.

If the system won't power on at all with this card plugged in, you have a short to ground. You said you don't have a short on the 8 pin connectors, so that leaves the 12V input from the PCI-E slot. Check for a short to ground on pin 2 of the slot connector.

The shunt at the bottom of the card that your picture says is "smoked" is most likely wired up to the 12V input from the slot. You will need to figure out where the short is in that circuit. It's hard to say without the card in my hands to test, but if it was shorted for long enough to visibly damage a PCI-E riser, chances are that the damage is actually in the 12V power plane itself, which means the card is beyond saving. Nevertheless, it would be worth removing the suspect shunt and checking to see if that clears the short. If it does, then there's a chance the damage is a component situated between the shunt and ground, such as a power stage or a capacitor.

Also, do not use short detection ("beep") mode when troubleshooting. There are circuits on this card (such as the core and main memory power) that measure as only a few milliohms to ground by design. The core should measure about 300 mohms, as I recall, and the memory should be maybe 10-15 ohms. Both of these are well below the threshold that most multimeters consider to be a short. The proper way to check for short circuits on a graphics card is to use regular ohms mode, and compare your readings to the known-good values.

The thing I would do next is remove that shunt, and see if that clears the short on pin 2 of the PCI-E connector. Do you have access to hot air? Do not attempt to use a soldering iron.
You're being too modest...I've read your posts and seen one of your YouTube vids.

Thanks for the advice. I'll have to check properly this weekend as I'm on nights this week (I'm a doctor by trade). I did manage to remove the resistor with hot air and there appears to be several kilo ohms of resistance between pin 2 and just one of the resistor pads.

I will upload proper details and pics later this week when I can finally sit down and figure this out. From what I can see, the resistor is located near some vias, which have two small capacitors before them. This makes me think the resistor comes before the power stages. Is it reasonable to assume this or would a sense resistor generally come after a power stage rather than directly from the 12v plane?

I'm not able to detect a short in the two capacitors I removed when checking resistance nor when using short detection mode. I'm probably going about it wrong.
 

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,808
You're being too modest...I've read your posts and seen one of your YouTube vids.

Thanks for the advice. I'll have to check properly this weekend as I'm on nights this week (I'm a doctor by trade). I did manage to remove the resistor with hot air and there appears to be several kilo ohms of resistance between pin 2 and just one of the resistor pads.

I will upload proper details and pics later this week when I can finally sit down and figure this out. From what I can see, the resistor is located near some vias, which have two small capacitors before them. This makes me think the resistor comes before the power stages. Is it reasonable to assume this or would a sense resistor generally come after a power stage rather than directly from the 12v plane?

I'm not able to detect a short in the two capacitors I removed when checking resistance nor when using short detection mode. I'm probably going about it wrong.
When you say you removed the resistor, do you mean the shunt?

Assuming yes, you should see the following on a healthy card:
Right side of shunt -> PCI-E Pin 2 : 0 ohms
Left side of shunt -> PCI-E Pin 2: Thousands or more
Right side of shunt -> Ground: Thousands or more
Left side of shutnt -> Ground: Thousands

If you mean the tiny 0402 resistor you said you were probing when you bridged something you shouldn't have, don't worry about that area for now. You need to figure out what general part of the board your dead short to ground is in first.
 

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
When you say you removed the resistor, do you mean the shunt?

Assuming yes, you should see the following on a healthy card:
Right side of shunt -> PCI-E Pin 2 : 0 ohms
Left side of shunt -> PCI-E Pin 2: Thousands or more
Right side of shunt -> Ground: Thousands or more
Left side of shutnt -> Ground: Thousands

If you mean the tiny 0402 resistor you said you were probing when you bridged something you shouldn't have, don't worry about that area for now. You need to figure out what general part of the board your dead short to ground is in first.
Yes I meant the R005 shunt not the resistor I accidentally probed.

Right side of shunt to pin 2 - 0.6 ohms
Left side of shunt to pin 2 > 2 MOhm

Right side of shunt to ground > 2MOhm
Left side of shunt to ground - 13.3 kOhm

Regarding the 2MOhm readings, I note they occasionally start rising to 2 MOhm then give me a "1 .", so I'm assuming the resistance is greater than this upper limit of my multimeter.
 
Last edited:

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,808
Yes I meant the R005 shunt not the resistor I accidentally probed.

Right side of shunt to pin 2 - 0.6 ohms
Left side of shunt to pin 2 > 2 MOhm

Right side of shunt to ground > 2MOhm
Left side of shunt to ground - 13.3 kOhm

Regarding the 2MOhm readings, I note they occasionally start rising to 2 MOhm then give me a "1 .", so I'm assuming the resistance is greater than this upper limit of my multimeter.
That all looks healthy. You may have gotten lucky, and not done any actual damage. Maybe.

What resistance do you have on the left side of each of the other shunts? Edit: That is, resistance to ground.
 

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
That all looks healthy. You may have gotten lucky, and not done any actual damage. Maybe.

What resistance do you have on the left side of each of the other shunts? Edit: That is, resistance to ground.
Shunt at:

J8 adjacent to 12v from PSU
Right = 44 kOhm
Left = 36.7 kOhm

J11 adjacent to 12v from PSU
Right = 9.04 kOhm
Left = 9.24 kOhm

Shunts were still on the board when I took these readings if that changes anything.
 
Last edited:

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,808
Shunt at:

J8 adjacent to 12v from PSU
Right = 44 kOhm
Left = 36.7 kOhm

J11 adjacent to 12v from PSU
Right = 9.04 kOhm
Left = 9.24 kOhm

Shunts were still on the board when I took these readings if that changes anything.
Ideally, you want to take these measurements with the shunts still on the board. The shunts have a very low resistance (typically about 0.5 ohms), and thus don't contribute much to the resistance value.

Your measurements there are all over the place, but still within the realm of normal, so that's a good sign. It does not appear that you have a dead short to ground. If you did, you'd have a really low resistance to ground on one or both of them.

Next, check the resistance through the shunt you removed. You should have about 0.5 ohms. If you have significantly greater than that, you'll need a new one. If not, reinstall it on the board. You can probably get away with using a soldering iron on that one, if it's easier than hot air. Just be quick, and use a good quality flux.

After that, with the card unplugged from the system completely, take resistance measurements between the locations shown in these pictures. (Click to embiggen)
strix_2080_low_voltage_resistance_checks.jpg

Here's what I suspect each one is:
1. Memory Power - expected resistance is something like 15-30 ohms
2. GPU Core Power - expected resistance is ~300 milliohms
3. 5V? - I'd guess about 2K ohms expected
4. ~1V? - I'd guess about 3500 ohms
5. 5V for usb-c? - Resistance should probably be a few K
6. 3.3V? - You usually see 200-400 ohms on this

Each one of those locations is the output of a power supply circuit. By checking the resistance on each one, we're checking to see if a short circuit has formed either through the load or some part of the power supply. If either of those things had happened, that would explain why our card doesn't work. I may be wrong about which one of the ones with question marks is which, which is part of the reason we're doing these checks. We need to at least kind of know what each one does before we do the next step.

Edit: When taking your resistance measurements, don't be afraid to scrape at the solder connections a little, in order to get a solid connection with your probes. You'll sometimes get a layer of oxidation on top of the solder, which throws your resistance measurements off, and you have to scrape through it to get accurate readings. Obviously, just be careful not to knock any of the tiny components off the board. You won't damage any of the big ones doing this.
 

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
Ideally, you want to take these measurements with the shunts still on the board. The shunts have a very low resistance (typically about 0.5 ohms), and thus don't contribute much to the resistance value.

Your measurements there are all over the place, but still within the realm of normal, so that's a good sign. It does not appear that you have a dead short to ground. If you did, you'd have a really low resistance to ground on one or both of them.

Next, check the resistance through the shunt you removed. You should have about 0.5 ohms. If you have significantly greater than that, you'll need a new one. If not, reinstall it on the board. You can probably get away with using a soldering iron on that one, if it's easier than hot air. Just be quick, and use a good quality flux.

After that, with the card unplugged from the system completely, take resistance measurements between the locations shown in these pictures. (Click to embiggen)
View attachment 328712

Here's what I suspect each one is:
1. Memory Power - expected resistance is something like 15-30 ohms
2. GPU Core Power - expected resistance is ~300 milliohms
3. 5V? - I'd guess about 2K ohms expected
4. ~1V? - I'd guess about 3500 ohms
5. 5V for usb-c? - Resistance should probably be a few K
6. 3.3V? - You usually see 200-400 ohms on this

Each one of those locations is the output of a power supply circuit. By checking the resistance on each one, we're checking to see if a short circuit has formed either through the load or some part of the power supply. If either of those things had happened, that would explain why our card doesn't work. I may be wrong about which one of the ones with question marks is which, which is part of the reason we're doing these checks. We need to at least kind of know what each one does before we do the next step.

Edit: When taking your resistance measurements, don't be afraid to scrape at the solder connections a little, in order to get a solid connection with your probes. You'll sometimes get a layer of oxidation on top of the solder, which throws your resistance measurements off, and you have to scrape through it to get accurate readings. Obviously, just be careful not to knock any of the tiny components off the board. You won't damage any of the big ones doing this.
The shunt measured around 0.4 ohms, and it's now back on the board.

1 - 1.0 ohm
2 - 0.9 ohms
3 - 14.8 kOhms
4 - 3.12 kOhms
5 - 81.8 kOhms
6 - 10.2 ohms

Edit: before my probing incident, the ohms from Vmem to ground were in the 20's, so the 1 ohm appears to be new.

EDIT #2: ignore the above comment as I cannot confirm since I didn't write it down.

Also point 3 had measured at 5V before.
 
Last edited:

RazorWind

2[H]4U
Joined
Feb 11, 2001
Messages
3,808
The shunt measured around 0.4 ohms, and it's now back on the board.

1 - 1.0 ohm
2 - 0.9 ohms
3 - 14.8 kOhms
4 - 3.12 kOhms
5 - 81.8 kOhms
6 - 10.2 ohms

Edit: before my probing incident, the ohms from Vmem to ground were in the 20's, so the 1 ohm appears to be new.

Also point 3 had measured at 5V before.
#1 is your problem. 1.0 ohms is basically a dead short to ground on that rail. I went and looked at some footage I took when I repaired an EVGA reference 2080 (which I'm using to type this post :cool: ), and I got 55 ohms on the memory rail on that card.

Your next step is to figure out what component is in short, and replace it. Unfortunately, this is likely to be pretty tough, unless you have some elaborate equipment. Probably 40% of the components on the board are related to the memory rail, and your short could be in almost any of them, including the memory ICs or the GPU die itself.

There are a few ways of going about this:

* You may get lucky, and figure out which components are damaged with a closer visual inspection. You mentioned you're a doctor - if you have access to a stereo microscope at work, you might try using it to examine the board.
* Plug the card into power and use a thermal camera, isopropanol, or your fingers to look for components that get hot.
* Attach a bench power supply to the memory rail, and inject voltage (it's designed for 1.35 volts), and look for components that get hot
* Speculatively remove components until you find the short.

If you strike out with the visual inspection and voltage injection, it might be worth at least removing the memory rail power stages. Just be aware that if you can clear the short by removing them, you'll likely need a board preheater and a great deal of luck to get new ones soldered on properly.
 

MSIGamer72

n00b
Joined
Feb 6, 2021
Messages
7
#1 is your problem. 1.0 ohms is basically a dead short to ground on that rail. I went and looked at some footage I took when I repaired an EVGA reference 2080 (which I'm using to type this post :cool: ), and I got 55 ohms on the memory rail on that card.

Your next step is to figure out what component is in short, and replace it. Unfortunately, this is likely to be pretty tough, unless you have some elaborate equipment. Probably 40% of the components on the board are related to the memory rail, and your short could be in almost any of them, including the memory ICs or the GPU die itself.

There are a few ways of going about this:

* You may get lucky, and figure out which components are damaged with a closer visual inspection. You mentioned you're a doctor - if you have access to a stereo microscope at work, you might try using it to examine the board.
* Plug the card into power and use a thermal camera, isopropanol, or your fingers to look for components that get hot.
* Attach a bench power supply to the memory rail, and inject voltage (it's designed for 1.35 volts), and look for components that get hot
* Speculatively remove components until you find the short.

If you strike out with the visual inspection and voltage injection, it might be worth at least removing the memory rail power stages. Just be aware that if you can clear the short by removing them, you'll likely need a board preheater and a great deal of luck to get new ones soldered on properly.
Ahh shucks. A microscope will be out of the question, the labs are swamped right now with this pandemic. I did use a pair of magnifying glasses I have lying around to do an inspection a couple weeks back, and there was no obvious damage.

I've got a power bench supply at home, I could try and inject voltage to see if this heats up any components. Where exactly would I apply voltage to? And would you say 2 amps is the max current I should allow that rail to draw during testing? (I see a lot of people setting 2 amps as their go to current draw limit).

And when the pc wouldn't turn on after the probing mishap, the shunt sparked on a subsequent attempt. If I now plug my GPU into the motherboard, how can I be sure that this won't cause it to happen again and potentially cause permanent damage?

If you wouldn't mind posting the link to that repair video it might be useful for reference.

I haven't connected one of the small capacitors below the shunt back to the board yet (the smaller one closest to the shunt) as I seemed to have misplaced it 😅. if I don't find it in my work area, do you reckon it would have a significant impact on function?

Thanks for all your input so far, looks like I've got some testing to do. If I find anything, I'll post back in here.

Edit: don't mind about the video, I thought you meant a video online not one you did yourself... probably not as useful for me without commentary.
 
Last edited:
Top