Hi everyone,
just registered a few days ago because I'm currently working on a Zotac GTX 1080 Amp. Got it very cheap because, according to the previous owner, it won't give an output on the monitor.
Indeed, no display on none of the outputs. Even after a few minutes there's not even a black screen signal, like you sometimes see on GPUs with faulty VRAM. Nevertheless, the GPU is kind of detected by the OS. The correct drivers are installed, GPU-Z shows 0 MB VRAM, revision FF and can't figure out the BIOS version and various other details about the GPU. But it's alive.
After making sure that all voltages are present, I went ahead and replaced the 27 MHz crystal (which was giving a flat line on the scope) and the VBIOS SPI chip (which got CLK and CS from the GPU, but no activity on SI and SO).
None of that helped.
No, sometimes problems arise from bad solder connections and so during POST I pressed down the corner of the GPU package in which the SPI lines are connected: "We have a picture". And it is even reproducible.
- No pressure during POST: No display
- Pressure during POST: Display
- Let go after POST: Picture disappears again
- Pressure again: Picture reappears
So it seems to affect more than just the SPI lines, because otherwise the output should be stable even without pressure. Maybe the 27 MHz lines to the crystal which btw are located in the same corner?
Whatever, this GPU has at least 2, probably 4 and perhaps even a few more disconnected pads.
What to do next? I have the equipment to rework SMDs but I never attempted to work on anything bigger than VRAM chips. Not sure if it's feasible to only reflow the critical corner of the GPU or if doing so would introduce additional stress.
But would it be enough to apply flux, heat the GPU to the point where it moves, wiggle it a bit and hope for the best? Otherwise it would need a complete removal and reballing procedure, which I don't feel comfortable doing.
Any suggestions?
just registered a few days ago because I'm currently working on a Zotac GTX 1080 Amp. Got it very cheap because, according to the previous owner, it won't give an output on the monitor.
Indeed, no display on none of the outputs. Even after a few minutes there's not even a black screen signal, like you sometimes see on GPUs with faulty VRAM. Nevertheless, the GPU is kind of detected by the OS. The correct drivers are installed, GPU-Z shows 0 MB VRAM, revision FF and can't figure out the BIOS version and various other details about the GPU. But it's alive.
After making sure that all voltages are present, I went ahead and replaced the 27 MHz crystal (which was giving a flat line on the scope) and the VBIOS SPI chip (which got CLK and CS from the GPU, but no activity on SI and SO).
None of that helped.
No, sometimes problems arise from bad solder connections and so during POST I pressed down the corner of the GPU package in which the SPI lines are connected: "We have a picture". And it is even reproducible.
- No pressure during POST: No display
- Pressure during POST: Display
- Let go after POST: Picture disappears again
- Pressure again: Picture reappears
So it seems to affect more than just the SPI lines, because otherwise the output should be stable even without pressure. Maybe the 27 MHz lines to the crystal which btw are located in the same corner?
Whatever, this GPU has at least 2, probably 4 and perhaps even a few more disconnected pads.
What to do next? I have the equipment to rework SMDs but I never attempted to work on anything bigger than VRAM chips. Not sure if it's feasible to only reflow the critical corner of the GPU or if doing so would introduce additional stress.
But would it be enough to apply flux, heat the GPU to the point where it moves, wiggle it a bit and hope for the best? Otherwise it would need a complete removal and reballing procedure, which I don't feel comfortable doing.
Any suggestions?
Attached Files
I placed a small piece of 1 mm think pad on the corer of the interposer. The heatsink mounted as usual and the little extra pressure is enough to "fix" the card. Temps look normal in idle (32 °C, 43 °C hot spot). I only dared to run a few seconds of Unigine Heaven but temps also seem normal here. That's a huge success and I'm almost certain now, that it's worth reballing the chip.
.
Comment