Dead 6800XT Troubleshooting

[Mercury Monkey]

I built my first PC back in the summer of 2021 with an Asus Tuff 6800xt I barely managed to get my hands on due to the GPU mining craze at the time. It worked really well other than a rare crash I would have while gaming that I was not able to replicate through the benchmarks I had access to at the time (heaven, and cinebench) and being my first PC i was unfamiliar with characteristics of what component may be failing. Fast forward to fall of 2024 when I ended up finding 3 games that would consistently crash my PC, Armored Core 6, Shadow of the Tomb Raider, and Horizon Zero Dawn. The crashes would cause the PC to fully restart, but display would not return until I manually restarted the computer. In hindsight, this should have been my indicator that it was a gpu problem but given how tough it was to get it I was relucted to accept that it was the issue. This lead me to start troubleshooting again but without knowing/accepting what the cause was i was shooing in the dark. This did result in me finding some faulty RAM that did not fix the issue, so I buckled in to try and wait it out until the next gen would come out. Sadly it died fully on me on a Friday when I came back from work, and my PC that should have been in sleep would not give me any display out.

 

Since the timing of its death put it outside of its warranty, i figured it would be more fun to take it apart and see if I can slowly learn how to fix it. I have a mechanical engineering background so sadly my hardware and circuit knowledge is limited, so any help in this endeavor is much appreciated. So far I have just taken it apart to give it a visual inspection which has yielded nothing as expected, so I would love to know what to start with next.

[emosun]

ok well step one would have been to actually know what the issue was with the computer before randomly deciding to take the gpu apart. A "no post" issue can be a problem with several different components that aren't the gpu so taking it apart probably now made the situation worse.

At any point in the last 3 year have you checked the system temps. 

also do we get to know what the system specs are or is that part of the challenge for us

 

[KidKid]

1 hour ago, Mercury Monkey said:

I built my first PC back in the summer of 2021 with an Asus Tuff 6800xt I barely managed to get my hands on due to the GPU mining craze at the time. It worked really well other than a rare crash I would have while gaming that I was not able to replicate through the benchmarks I had access to at the time (heaven, and cinebench) and being my first PC i was unfamiliar with characteristics of what component may be failing. Fast forward to fall of 2024 when I ended up finding 3 games that would consistently crash my PC, Armored Core 6, Shadow of the Tomb Raider, and Horizon Zero Dawn. The crashes would cause the PC to fully restart, but display would not return until I manually restarted the computer. In hindsight, this should have been my indicator that it was a gpu problem but given how tough it was to get it I was relucted to accept that it was the issue. This lead me to start troubleshooting again but without knowing/accepting what the cause was i was shooing in the dark. This did result in me finding some faulty RAM that did not fix the issue, so I buckled in to try and wait it out until the next gen would come out. Sadly it died fully on me on a Friday when I came back from work, and my PC that should have been in sleep would not give me any display out.

 

Since the timing of its death put it outside of its warranty, i figured it would be more fun to take it apart and see if I can slowly learn how to fix it. I have a mechanical engineering background so sadly my hardware and circuit knowledge is limited, so any help in this endeavor is much appreciated. So far I have just taken it apart to give it a visual inspection which has yielded nothing as expected, so I would love to know what to start with next.

Do you have integrated CPU graphics? If that works, it's the GPU that's dead, and not something else.

[emosun]

4 minutes ago, KidKid said:

Do you have integrated CPU graphics? If that works, it's the GPU that's dead, and not something else.

no that would confirm that the machines integrated graphics work.... not that the gpu is dead. You have to actually TRY the gpu in other machines to confirm it's dead.

this is why morons sell "dead" gpus on ebay all the time having no clue that the gpu works fine and their methodology is flawed. 

 

[Mercury Monkey]

I had managed to borrow another GPU. With my old GPU, the motherboard would get stuck on the vga LED and not post, but worked fine with the other gpu. Had no issues after switching GPU's. The Asus Tuff also has a dual bios so I tried flipping the switch in case the vbios i was on somehow got corrupted in some way but that did not work either, hence my conclusion that the GPU is dead.

 

system specs were: amd 5600x, MSI B550 Tomahawk, Corsair vengeance 3200 RAM, Corsair RM850x, and the mentioned Asus Tuff 6800xt OC.

[YoungBlade]

Flipping the BIOS switch was a good troubleshooting step. The disassembly is potentially premature if you have access to another machine to test the card in, but it sounds like you don't.

 

Before attempting to do anything significant to the card, since you've already taken it apart, you should simply repaste it, reassemble it, and try again.

 

If the thermal paste was dried out, it could have been overheating - dry paste effectively becomes an insulator, and could make the card shut down almost immediately to save itself.

 

Additionally, it could be that a fan connector had come loose or that a fan has died. Some cards will not start up properly if they detect that a fan is disconnected or dead. When you try booting the system, do you see the fans spin up? If all three don't spin, then it could be that one of them has died. If none of them spin, then it could be that the card isn't getting power properly.

 

The next thing I would check on before trying to do anything to the card is the power supply. The fact that you have a modular one means it could be that the PSU end of the cable has come loose - that's an easy thing to check.

 

What GPU are you using to test against? Does it also require two 6+2 pin connections? Or just one? Or none?

[Mercury Monkey]

I do now have a new rig set up and i have kept all my old components so i could quickly design and 3d print myself a test bench and see if it was just a thermal issue. Admittedly, I am reluctant to take apart the new rig to test the old one unless is absolutely necessary.

 

I tested it with two cards, first a 3060 from my dad's dell prebuilt he was kind enough to let me take out, and second was the one used temporarily until i made my new rig which was an evga 3070 i borrowed from work. Both those cards worked fine with no issues. I had checked the power supply and even after confirming post and boot with the other cards, tried putting the 6800xt back in to see if maybe the motherboard just got stuck for some reason and getting it to work with another card may have helped alleviate the problem but to no luck.

[freeagent]

Ded. You could try Northridge fix, but I don't think he likes AMD GPU's that much haha.

[Mercury Monkey]

Rather then sending it off to someone else to try to repair it, I figured it would be a fun challenge to try and repair it myself and learn a bit along the way.