How to fix the "GPU not detected" error after adding a new card to my rig?

Power Supply and Physical Connection Verification

1. Confirm that the GPU’s auxiliary PCIe power connectors are fully seated and matched to the PSU’s available 6-pin or 8-pin PCIe cables — mismatched or under-rated cables often cause silent initialization failure.

2. Measure rail stability using a multimeter on the 12V line during POST; voltage sag below 11.4V indicates insufficient wattage or aging capacitors in the PSU.

3. Inspect PCIe slot retention clips for deformation — physical misalignment prevents full electrical contact even when the card appears flush.

4. Test the GPU in another known-working motherboard slot to rule out lane enumeration issues specific to the primary x16 slot.

BIOS/UEFI Configuration Adjustments

1. Enter BIOS and disable CSM (Compatibility Support Module) to enforce UEFI-only boot path, which ensures proper PCIe device enumeration before OS handoff.

2. Set PCIe Link Speed to “Gen3” manually instead of “Auto”, as some motherboards fail negotiation with newer GPUs at Gen4/Gen5 speeds during cold boot.

3. Enable Above 4G Decoding and Resizable BAR support — absence of these settings blocks memory-mapped I/O space allocation required for modern GPU firmware handshake.

4. Reset BIOS to optimized defaults after updating firmware, then reapply only necessary changes — corrupted NVRAM entries from prior overclocking can stall PCIe device discovery.

Kernel-Level Driver Initialization Failures

1. Boot into recovery mode and run dmesg | grep -i 'pcie|nvidia|amdgpu|gpu' to capture early-boot hardware detection logs — missing “bridge window” or “BAR 0: assigned” messages indicate PCIe topology failure.

2. Check if the GPU appears in lspci -vv -s $(lspci | grep VGA | head -n1 | cut -d' ' -f1) output — absence confirms no enumeration, while presence without subsystem IDs points to incomplete configuration space read.

3. Verify kernel parameter pci=noacpi is not set in GRUB_CMDLINE_LINUX — ACPI-based resource allocation conflicts with PCIe hotplug-capable motherboards during cold boot.

4. Load kernel modules explicitly via sudo modprobe nvidia-drm nvidia-modeset nvidia and monitor dmesg for “failed to get fb info” or “timeout waiting for GPU to become ready” — these indicate firmware hang during VBIOS execution.

Firmware and Hardware Compatibility Constraints

1. Cross-reference GPU model against motherboard QVL (Qualified Vendor List) — many AM5 and LGA1700 boards list only specific AIB partner cards as validated, excluding reference designs with nonstandard power sequencing.

2. Update GPU VBIOS using manufacturer-provided flash utilities before installation — outdated VBIOS fails PCIe ASPM negotiation on newer chipsets.

3. Disable Fast Startup in Windows or disable systemd’s “hibernate” target in Linux — residual PCIe link state from hibernation prevents clean re-initialization on next boot.

4. Test with minimal hardware: remove all non-essential PCIe devices, disconnect M.2 drives sharing CPU lanes, and boot with only one RAM stick installed — latent timing conflicts often manifest solely as GPU invisibility.

Common Questions and Direct Answers

Q: Why does lspci show my GPU but nvidia-smi says “NVIDIA-SMI has failed because it couldn’t communicate with the NVIDIA driver”?A: The GPU is electrically present but the kernel module failed to bind — check lsmod | grep nvidia and inspect dmesg for “probing for device” followed by “ignoring BAR” errors indicating memory region conflict.

Q: Can a GPU be physically damaged during installation without visible signs?A: Yes — excessive insertion force can shear internal micro-solder joints on PCIe edge connectors, resulting in intermittent or absent enumeration detectable only via thermal imaging or boundary scan testing.

Q: Does using a PCIe riser cable affect GPU detection?A: Passive risers introduce signal integrity degradation; active risers require separate power and correct PCIe generation matching — mismatched generations (e.g., Gen4 GPU on Gen3 riser) cause enumeration timeout before OS load.

Q: Why does the GPU appear in BIOS but vanish after GRUB loads?A: Kernel command line contains nomodeset or rd.driver.blacklist=nouveau,nvidia, forcing driver exclusion before probe — verify current parameters via cat /proc/cmdline.