I'm troubleshooting an ASUS TUF Gaming Z790-Plus WiFi system with an Intel Core i9-14900KF, ASUS ROG Astral RTX 5080 OC, 32 GB of DDR5-6000 with XMP enabled, and a Samsung 990 Pro 2TB PCIe 4.0 SSD. BIOS 1836 and the NVIDIA driver are up to date. GPU-Z consistently reports the graphics card as PCIe 5.0 x8, even during the GPU-Z render test, Steel Nomad, and several games. The link never changes to x16 under load. I have already cleared CMOS, confirmed the card is installed in the primary PCIe x16 slot, had it reseated, and verified that no riser cable is being used. Temperatures, power usage, SSD performance, and system stability all appear normal, with no WHEA errors. The SSD is installed in the primary M.2 slot, which the motherboard manual says should not reduce the graphics slot to x8. Could this be caused by a BIOS or PCIe negotiation issue, bent LGA1700 socket pins, poor CPU contact, a defective motherboard, or a problem with the GPU? Has anyone run this motherboard and an RTX 5080 at PCIe 5.0 x16, or found a BIOS setting or slot configuration that fixes it?
4 Answers
The performance impact is probably small. PCIe 5.0 x8 provides roughly the bandwidth of PCIe 4.0 x16, and an RTX 5080 generally won’t lose much performance in ordinary gaming at that connection speed. It’s still worth fixing if the system is supposed to run at x16, but the result alone doesn’t necessarily indicate an imminent hardware failure.
First, make sure the reading is being checked while the GPU is actually under sustained load. PCIe links can downshift when idle, but since you’ve already tested with GPU-Z, benchmarks, and games and it stays at x8, this sounds like a genuine negotiated-link limitation rather than normal idle behavior.
Try moving the Samsung 990 Pro to a lower M.2 slot temporarily and then check the GPU link again. Even though the manual suggests the primary M.2 slot should not share lanes with the main graphics slot, testing another slot is an easy way to rule out lane routing or firmware behavior.
If changing the M.2 slot does nothing, I’d inspect the LGA1700 socket and CPU seating next. The main PCIe lanes come directly from the CPU, so bent socket pins, uneven cooler pressure, or poor contact can sometimes cause reduced lane width. After that, testing with another motherboard or GPU would help separate a CPU/socket problem from a board or graphics-card fault.
That’s what I suspected too. The link remains at x8 during the render test and benchmarks, so it doesn’t appear to be an idle power-saving state.