I have a Z790-BTF-WiFi motherboard with built-in 2.5GbE, and my ISP recently increased my internet speed. I added several 10GbE adapters, but the PC randomly loses its physical network link for about 4–7 seconds before reconnecting at the same 10Gbps auto-negotiated speed. The problem disappears when I force the connection to 1Gbps.
I reproduced the issue with the motherboard's Intel I226-V controller, two Intel X540-T2 cards, and a TP-Link TX401 using a Marvell AQtion controller. One of the X540-T2 cards worked normally in another computer. I tested both PCIe slots, multiple ports, two unmanaged 10GbE switches, a direct PC-to-switch connection, several Cat6A cables, and a 100-foot cable while bypassing all in-wall wiring and keystones.
I also tried auto-negotiation, forced 5Gbps and 1Gbps modes, changing Energy-Efficient Ethernet, a high-performance Windows power plan, disabling PCIe link-state power management, reseating the cards, and improving cooling on both the adapters and switches. There is no NIC teaming, bridge, Hyper-V virtual switch, third-party network filter, or matching WHEA, PCIe, Kernel-PnP, Kernel-Power, or NDIS reset event.
With the TX401 installed, Windows logs aqnic650 Event ID 14 when the link drops and Event ID 15 when it returns. What motherboard, PCIe, driver, firmware, or compatibility issue could cause this behavior across multiple network adapters?
3 Answers
The fact that several unrelated network adapters show the same physical-link drop points more toward the motherboard, firmware, PCIe slot behavior, or a shared power-management issue than toward the switches or cables. I’d update the motherboard BIOS, chipset package, Intel ME firmware, and each adapter’s driver, then check Device Manager to make sure Windows isn’t using a generic or mismatched driver. It may also be worth testing with onboard devices and aggressive PCIe power-saving options disabled.
Since the same behavior follows the PC despite different adapters, a spare compatible motherboard or a test system with the same cards would be the cleanest way to isolate it. A GPU dependency makes swapping the motherboard difficult, but even temporarily testing without the usual GPU setup—or using another system—could distinguish a board-level PCIe or firmware problem from an Ethernet compatibility problem. The lack of reset events doesn’t rule out a negotiation or physical-layer issue; the adapter is simply reporting that the link disappeared and later returned.
Be careful interpreting the forced-speed tests: the X540-T2 supports 1Gbps and 10Gbps, but not an intermediate 5Gbps mode. The TX401 does support 5Gbps, so that test only applies to that card. The TX401 can also be sensitive to PCIe slot placement; try the first or second usable slot rather than a lower-numbered slot near the bottom of the board, and compare the results after a cold boot.
The 5Gbps test was only done with the TX401. The X540-T2 cards were tested at 1Gbps and 10Gbps using their proper drivers. The problem has occurred in both available PCIe slots, but I’ll verify the slot preference and check for newer TX401 drivers and motherboard firmware.
Testing another motherboard is difficult because my GPU requires the BTF board’s rear-mounted power arrangement. I’ve already actively cooled the adapters and switches, so overheating seems less likely, but a separate compatible system may still be the best way to confirm whether the motherboard is responsible.