ROG 1080Ti failing to post (got it to post twice)

[Harmsway1283]

I've been troubleshooting this 1080Ti (rev 2.01) and I've also been reading through the posts of other people having similar problems. I was hoping I could ask some questions and get some advice while moving forward.
The card isn't recognized on start up, I get the 1 long 3 short beep code. Using IGPU to check through the MB, hardware monitor gives me code 43, and NVFlash gives me the eeprom error. GPU-Z recognizes it as a 1080Ti, but no other info (bios version unknown, blank spaces for clock speed, memory, etc) and I couldn't access NVidia control panel. So far I've checked:

- All voltage rails, all check good
- no detected shorts
- on visual inspection I did find a missing cap near a memory module, replaced with no change. Also, there is a missing 6-pin IC that after checking the boardview it's related to the #2 DP socket output signal only (not too worried about it at the moment).
- Checked voltages on the BIOS chip itself (Windbond 25Q40EWNIG) and what I found was:
- during start up nothing on pins 2, 5, and 6.
- Pin 3, 7, and 8 all get 1.83v.
- Pin 1 pulls low and eventually climbs to 1.83v

Now then, after tracking current paths in boardview, I found unoccupied pads for a resistor in a row of other resistors (see picture) that looked like it would feed 1.8v to pin 5. The boardview file I have doesn't list any part numbers or specs, so I metered the resistors to the left (.5 kΩ resistor ties to ground) and right (.09 kΩ resistor tied to pin 5) of the unoccupied pads. I put a .5 kΩ resistor in the unoccupied space and decided to try it. The card posted on the first attempt and after checking hardware monitor code 43 was gone (this device is working properly), GPU-Z had all information available, NVidia control panel allowed settings changes and driver installs.

I thought I had it fixed, but after putting everything back together, no post with1 long 3 short beeps. I took it back apart and I guess my solder job wasn't the best because the resistor moved after I touched it with tweezers and it was lost to Nth dimension. I did not check any voltages after it posted that time (I had assumed that it was fixed). Since that point, I have soldered on other resistors (.09 kΩ, .12 kΩ, & 0 Ω) and I've been getting worse results. Now I can't get it to post again. I get no output from pin 2 and pin 6 is not receiving anything (not sure if it ever was but I assume it must have to post).

My questions are:

- Do any of you have information on the input/output/clock voltages (pins 5, 2, and 6)?
- Is pin 5 strictly input and pin 2 strictly output? (checked the datasheet and not sure if the bios runs in standard SPI or dual SPI)
- Pin 6 (clock) should get it's input through the core and the core gets it from the oscillator right?

There are some scuff marks on the GPU making think that it has been dropped. According to the Pascal GPU Diagnostic guide on the repair wiki, that could damage the oscillator and in turn effect the operation, but then again, I did get it to post. All this was done over a 2-3 hour period yesterday and I'll keep digging later today. Any and all suggestions or information is much appreciated.

***Update***

Found out that I have a short in the VMem and VCore rails if not more. I re-checked all the rails while I was getting the voltages from the oscillator. Found VMem half a volt low, VCore slightly high and two 12v coils near the SLI bridge were reading 1.4v. Checked the resistance across VMem and VCore rails; both were approx 5.73 kΩ's.

 

[qubit]

Is this a used card off a private seller on eBay?

 

[r9]

Is this a used card off a private seller on eBay?

I've been troubleshooting this 1080Ti (rev 2.01) and I've also been reading through the posts of other people having similar problems. I was hoping I could ask some questions and get some advice while moving forward.
The card isn't recognized on start up, I get the 1 long 3 short beep code. Using IGPU to check through the MB, hardware monitor gives me code 43, and NVFlash gives me the eeprom error. GPU-Z recognizes it as a 1080Ti, but no other info (bios version unknown, blank spaces for clock speed, memory, etc) and I couldn't access NVidia control panel. So far I've checked:

- All voltage rails, all check good
- no detected shorts
- on visual inspection I did find a missing cap near a memory module, replaced with no change. Also, there is a missing 6-pin IC that after checking the boardview it's related to the #2 DP socket output signal only (not too worried about it at the moment).
- Checked voltages on the BIOS chip itself (Windbond 25Q40EWNIG) and what I found was:
- during start up nothing on pins 2, 5, and 6.
- Pin 3, 7, and 8 all get 1.83v.
- Pin 1 pulls low and eventually climbs to 1.83v

Now then, after tracking current paths in boardview, I found unoccupied pads for a resistor in a row of other resistors (see picture) that looked like it would feed 1.8v to pin 5. The boardview file I have doesn't list any part numbers or specs, so I metered the resistors to the left (.5 kΩ resistor ties to ground) and right (.09 kΩ resistor tied to pin 5) of the unoccupied pads. I put a .5 kΩ resistor in the unoccupied space and decided to try it. The card posted on the first attempt and after checking hardware monitor code 43 was gone (this device is working properly), GPU-Z had all information available, NVidia control panel allowed settings changes and driver installs.

I thought I had it fixed, but after putting everything back together, no post with1 long 3 short beeps. I took it back apart and I guess my solder job wasn't the best because the resistor moved after I touched it with tweezers and it was lost to Nth dimension. I did not check any voltages after it posted that time (I had assumed that it was fixed). Since that point, I have soldered on other resistors (.09 kΩ, .12 kΩ, & 0 Ω) and I've been getting worse results. Now I can't get it to post again. I get no output from pin 2 and pin 6 is not receiving anything (not sure if it ever was but I assume it must have to post).

My questions are:

- Do any of you have information on the input/output/clock voltages (pins 5, 2, and 6)?
- Is pin 5 strictly input and pin 2 strictly output? (checked the datasheet and not sure if the bios runs in standard SPI or dual SPI)
- Pin 6 (clock) should get it's input through the core and the core gets it from the oscillator right?

There are some scuff marks on the GPU making think that it has been dropped. According to the Pascal GPU Diagnostic guide on the repair wiki, that could damage the oscillator and in turn effect the operation, but then again, I did get it to post. All this was done over a 2-3 hour period yesterday and I'll keep digging later today. Any and all suggestions or information is much appreciated.

Test bench specs:
Asus Prime390-A
i3-8100
16Gb GSkill DDR4
EVGA BQ 650W

If it was shorted you would have had better chance of fixing it under assumption the the short is not coming from the GPU.
Being able to post couple times could mean that easily could be a resistor that went out of spec in any of the dozen current sensing circuits and triggering a shutdown.
Without a diagram or same exact working card to compare it will be impossible to fix.
If anything else fails you can try to reflow the gpu.

 

[Harmsway1283]

Is this a used card off a private seller on eBay?

That is correct.

 

[qubit]

I don't see what's so funny, it's not like the OP stated it, is it? He's confirmed it in his reply to me, too.

That is correct.

ok, was it sold as working and then you found out it didn't work? If so, then I strongly suggest returning it for a refund is possible.

Any background info you can tell us about it will help members to advise you better.

 

[Harmsway1283]

ok, was it sold as working and then you found out it didn't work? If so, then I strongly suggest returning it for a refund is possible.

Any background info you can tell us about it will help members to advise you better.

Understood. I won this in an eBay auction of a bunch of GPU's that aren't working. I knew this GPU would need work when I bought it. There was zero information provided on what kind of or if any troubleshooting had been done by the seller. The card has definitely been worked on before; the sticker over the heatsink screws was broken, a cap and that 6-pin IC were missing, and honestly the solder on the bios IC doesn't look factory. No way of knowing whether the bios IC was replaced or removed and reinstalled. I received the GPU yesterday, hooked it up to my test bench, it didn't post. Checked programs (GPU-Z, device manager, etc) tried to reboot with a MB speaker this time; 1 long 2 short beeps later I took it apart to begin a visual inspection. From that point is where I did everything I talked about in my first post.

If it was shorted you would have had better chance of fixing it under assumption the the short is not coming from the GPU.
Being able to post couple times could mean that easily could be a resistor that went out of spec in any of the dozen current sensing circuits and triggering a shutdown.
Without a diagram or same exact working card to compare it will be impossible to fix.
If anything else fails you can try to reflow the gpu.

Having the boardview has been a huge help, and there is a ton of information on diagnosing and repairing these 1080's. I'm reaching out to ask these questions because the information I'm looking for about pins 2, 5, and 6 I haven't been able to find anywhere. A shot in dark I know, but I'm hoping someone has those numbers.

Reflowing is something I have considered if I reach a dead end. It does kinda make sense that maybe there's a bad solder joint to the GPU core pad that sends the clock signal to the bios IC. Maybe after warming the board while I was working on it made that connection just a little better and that's why it posted. Maybe it cooled off in the time that I was putting it back together and the connection got worse again and that's why it didn't post then.

When it did post, it didn't crash on me. I shut it down to reinstall the heatsink (assuming that I had fixed it) and it would fail to post again.

 

[qubit]

Understood. I won this in an eBay auction of a bunch of GPU's that aren't working. I knew this GPU would need work when I bought it. There was zero information provided on what kind of or if any troubleshooting had been done by the seller. The card has definitely been worked on before; the sticker over the heatsink screws was broken, a cap and that 6-pin IC were missing, and honestly the solder on the bios IC doesn't look factory. No way of knowing whether the bios IC was replaced or removed and reinstalled. I received the GPU yesterday, hooked it up to my test bench, it didn't post. Checked programs (GPU-Z, device manager, etc) tried to reboot with a MB speaker this time; 1 long 2 short beeps later I took it apart to begin a visual inspection. From that point is where I did everything I talked about in my first post.

Ok, so you intentionally bought yourself a techie challenge and knew the state of the card near enough - how very enthusiast of you, the qubit approvez.

I think the fact you got it to post at all is impressive and hope you're able to fix it. At least you know that the actual GPU works. I'm sorry I can't help with that board level repair unfortunately.

 

[eidairaman1]

See if they have websites/emails you can ask about pcb repair

 

[Count von Schwalbe]

@Mussels has a method of reflowing you could ask him about...

 

[Harmsway1283]

***Update***

I'll be sure to add this to my original post as well. After trying to find any information on how to determine if an oscillator is bad without an oscilloscope, I decided to gather what information I could from the one on my card (.89-.91v but I haven't determined where it's getting voltage from). I re-checked all the voltage rails because it seemed like an odd voltage and found the VMem (should be 1.4v) low and VCore (should be 0.7v) slightly high (both at .82v) and the two 12v coils below the SLI bridge are showing 1.4v. I checked the resistance on VMem and VCore (Mem should be 30-60 Ω) and both are roughly 5.73 kΩ. So I've got a short. I'll find it tomorrow. Coincidentaly, thanks for that third video recommendation eidairaman1. I'll watch that one, I've seen the other two. Northridgefix puts out some good videos too.

 

[eidairaman1]

***Update***

I'll be sure to add this to my original post as well. After trying to find any information on how to determine if an oscillator is bad without an oscilloscope, I decided to gather what information I could from the one on my card (.89-.91v but I haven't determined where it's getting voltage from). I re-checked all the voltage rails because it seemed like an odd voltage and found the VMem (should be 1.4v) low and VCore (should be 0.7v) slightly high (both at .82v) and the two 12v coils below the SLI bridge are showing 1.4v. I checked the resistance on VMem and VCore (Mem should be 30-60 Ω) and both are roughly 5.73 kΩ. So I've got a short. I'll find it tomorrow. Coincidentaly, thanks for that third video recommendation eidairaman1. I'll watch that one, I've seen the other two. Northridgefix puts out some good videos too.

You should email those people for guidance on the repairs

 

[qubit]

@Mussels has a method of reflowing you could ask him about...View attachment 249692

Oh, hell no...

 

[Count von Schwalbe]

Oh, hell no...

Why the f*ck not? - a bad ideas only build, by Mussels!

TL;DR: Second hand parts bonanza, forgot i needed a PSU... Long version: Since i fixed my 2700x with a snapped pin by throwing a SATA cable in the socket, i've had a ryzen 1400 sitting around being sad pals with a GTX 980 in the closet. They needed a home. Enter facebook, with some guy selling...

www.techpowerup.com

Link provided for the full story... Much amusement. Some really good tips on reflowing there by lexluthermiester though.

 

[qubit]

Why the f*ck not? - a bad ideas only build, by Mussels!

TL;DR: Second hand parts bonanza, forgot i needed a PSU... Long version: Since i fixed my 2700x with a snapped pin by throwing a SATA cable in the socket, i've had a ryzen 1400 sitting around being sad pals with a GTX 980 in the closet. They needed a home. Enter facebook, with some guy selling...

www.techpowerup.com

Link provided for the full story... Much amusement. Some really good tips on reflowing there by lexluthermiester though.

Thanks, I'll take a look. I see it's a big thread...

 

[Deleted member 24505]

Mine is a ROG 1080ti i am sure but it is a 1.01X and is flawless