I bought a used Sapphire Pulse 9070 XT and briefly tested it in a friend's PC before accepting the delivery. It displayed an image and reached the desktop, although we did not install the graphics drivers at the time. A few days later, I installed it in my own system with a Ryzen 7 7800X3D, Gigabyte X670 Aorus Elite, and 32 GB of DDR5. The rest of the computer works normally through the CPU's integrated graphics, but the motherboard does not detect the card and stays on the VGA diagnostic LED.
We updated the BIOS and tried different settings, including PCIe Gen 3 and Gen 4, ReBAR, 4G decoding, CSM, and fast boot. Testing other graphics cards, including a GTX 1650 and RTX 3090, showed that the motherboard, CPU, and power supply were working correctly. The 9070 XT also failed to POST in my friend's system.
Oddly, after moving it back into my PC, the card briefly posted again and displayed an image through HDMI. However, the system crashed while installing the graphics drivers. Afterward, the card stopped posting again. We used the integrated graphics and another GPU to remove the partially installed drivers with DDU, but the 9070 XT still failed. It only briefly worked again after being removed and reinstalled, then crashed during another driver installation and has not worked since.
The card has now been tested in three computers with four different power supplies. Since it usually fails before the operating system even loads, I suspect a hardware problem such as corrupted firmware, damaged VRAM, or a cracked solder joint rather than a normal driver issue. The seller does not have the original receipt, so warranty service may not be practical. Has anyone dealt with similar symptoms or found a reliable way to diagnose or repair a card like this?
3 Answers
The pattern strongly suggests the card itself is failing. A driver crash can expose an existing fault, but it should not normally prevent the system from completing POST in several different computers afterward. Possible causes include damaged VRAM, a failing power stage, corrupted VBIOS, or physical damage. Before attempting a risky reflash, document the failures and contact the seller or the marketplace’s dispute process. A repair shop with board-level diagnostic equipment would be the safest option if no return or warranty route remains.
Check Windows Reliability Monitor and look for hardware-error entries around the crashes. You can also try every available PCIe slot if you haven’t already, but a card that fails in multiple known-good systems and with several power supplies is very likely defective. At that point, returning it or pursuing a buyer-protection claim is safer than repeatedly installing drivers or flashing firmware.
It may be worth testing from a Linux live USB. If the card manages to boot, run `lspci` to see whether the system detects it, then investigate the firmware version and hardware diagnostics. You could compare the card’s firmware checksum with the correct version and consider reflashing it, but only if the card is consistently detected and you can confirm the exact firmware. VRAM and power-delivery tests may also help identify a hardware fault.
I can try that, although the card usually fails to POST at all, so I’m not very confident this is only a driver problem.
Reliability Monitor does report a hardware error. Unfortunately, we accepted the delivery after the first brief test, so getting a refund may be difficult.