Dead RTX 2080 Super

[gamaliux]

Hello everyone!

I'm looking for support on how to fix Gainward RTX 2080 Super Phantom GHL. These are the symptoms:
1. Fans spins at 100%.
2. Black screen. Even screen monitor does not detect anything connected to it.
3. I get five beeps from a buzzer.
4. Windows shows Error 43;
5. nVflash, Mats/Mods shows that there is no device detected.


I have done many measurements:
1. I have measured all rails resistances and voltages. Every voltage is present.

2. I have measured each pair of PCIe diff lanes with diode mode, each pair has around 0.6-0.7V voltage drop, which is correct.

3. I have measured XTAL clock, it is present.

4. I have tried to boot up this GPU with iGPU and check everything. This is what I found:
4.1. Windows detects card with error 43. Driver install did not change anything, also no blue screen.
4.2. I have tried to use nVflash to rewrite bios, however it says GPU device not found. After that I have desoldered flash and reprogram it with Beeprog2. It is dual bios device, I have rewrite both of them, nothing changed.
However I believe I have used wrong Bios. I have saved old dump from flash also. I have used these bios: https://www.techpowerup.com/vgabios/205193/gainward-rtx2080-8192-180918
It says it is Pheonix bios, but photo is for Phantom.
4.3. I have tried to run Mods/Mats. It also says PCIe device not detected.

5. I have measured PGOOD, CLKREQ, WAKE, PEX_RST lines.
5.1. PEX_RST has AND gate, output is correct.
5.2. PGOOD also has AND gate, output is correct.
5.3. WAKE has 10k pull up resistor. There is always high level.
5.4. CLKREQ has 10k pull up resistor. There is always low level (issue?). No short to GND.

6. I have checked both flashes CS signal. I can see that it makes some communication, signal goes high and low for a few times. Which flash works is regulated by one N-channel mosfet and one two N-channel mosfet in one case. It seems everything works correctly here.


One more thing to mention, that GPU heats up very fast if there is no heatsink present. Within 30secs it can reach up to 100degC.
With heatsink, bottom of the board does not seem to heat a lot. I believe that is normal, to heat up fast without heatsink.


If any one would have some tips what else to check, I would be very gratefull!

 

[pavle]

GPU heating up is a good sign from what I know, other than that I can't help you. Perhaps @madness777 would be so gracious to chime in?

 

[gamaliux]

Hello pavle,

Thank you for your reply.

As far as I know heating means that voltage is preset which is of course a good sign. However I do not know how hot it is normal for GPU to get in such scenario.
I have seen some graphs of PCIe interface on the internet, where it shows that power cosumption is higher until the device is initialized. Also not sure if it is true.
Maybe that would be the reason why it could heat up more or maybe it has to heat up like that when no heatsink present.

 

[rafik2019]

When fans go at 100% its symptom of dead gpu chip i think.
I might be wrong with those, im not an expert but i did watch many videos about gpu diagnose and repair and many says that.

 

[gamaliux]

Hello rafik2019,

I always keep in mind that GPU may be dead, however I do not know how to prove it. If the GPU would be dead, I believe I should see some unusual measurements on PCIe pins.
If the GPU is dead, so the card goes to trash bin.
Other option that some BGA ball may be broken. Which means it would help to reball the GPU.

I believe 100% fans work releated to that the gpu is not initialized, which in order stops program to work correctly. Without working correctly program, fans will spin as they could.
But this is my opinion. Always glad to get more knowledge.

 

[rafik2019]

Well u can cook your gpu for about 10-15 mins(guide on yt). I did that with my old 7870 worked fine for some time(Do it before u gonna trash it).
Four months after i had to do it again.
Or use hot-air station to make better contact.
Another question to ask is what u did exactly with card before it stopped working.

 

[Bomby569]

the board can be fine and the VRAM die for example. Hard to say what it is.

Anyway do not trash it under any circunstance, there is always someone interested in buying even if only for parts.

 

[gamaliux]

rafik2019,

Yeah I know about that stuff.
This is the last optiom to go for hot air heating GPU with some flux, to clean oxides and to see if this helps.

Bomby569,

I do not think that it is releated to VRAM.
I had GTX1070, where two of the VRAM Vref voltage were zero. However it did not spin fans as crazy, also monitor found that card connected but it did not shown anything on screen. And Mats/Mods also nVflash worked just fine. I believe this could be more related to some kind of logic or bios. Also high probabilty of damaged GPU.

 

[pavle]

Just please do not put the card in the oven, gently heat it with the heat gun (adding flux)...

 

[gamaliux]

pavle,

Yes, I would never put it in the oven.
Hot air station with flux is the only way. But this is the last option, currently I'm looking for what I could test more.

 

[eidairaman1]

Rtx 2000 series were known for bad ram/imc.

And from your description the gpu should be replaced or a professional service repair it, try krisfix.de

 

[gamaliux]

Hello eidairaman1,

Thank you for your response.
I also thought about IMC, well which is basically the GPU. So yeah, I think the only option left is to try to heat GPU, if it will not help, new GPU required, not sure if it is possible to get it anywhere.

 

[eidairaman1]

Hello eidairaman1,

Thank you for your response.
I also thought about IMC, well which is basically the GPU. So yeah, I think the only option left is to try to heat GPU, if it will not help, new GPU required, not sure if it is possible to get it anywhere.

They are but by the time you have the die in hand you would have bought a brand new gpu from Zotact/Inno3D/Galaxy

 

[gamaliux]

They are but by the time you have the die in hand you would have bought a brand new gpu from Zotact/Inno3D/Galaxy

I agree with you. But electronics is also as a hobby to me, so I like to try fixing it.
If you know where it is possible to buy GPU, can you share? Or is it confidential?

 

[hallmark1730]

Hello, I have an RTX3090 that is in exactly the same situation. You have not made any updates to the issue. Have you been able to find out what the problem is during this time?

I agree with you. But electronics is also as a hobby to me, so I like to try fixing it.
If you know where it is possible to buy GPU, can you share? Or is it confidential?

?