ASUS Z690 paired with several NVMe storage devices. Performance drops when stress is applied.
ASUS Z690 paired with several NVMe storage devices. Performance drops when stress is applied.
Not long ago, I was working on a discussion trying to solve an odd problem: My ASUS Z690 motherboard didn’t work well with Gen 4 NVMe drives. I shared my experience across several topics—CPU, boards, RAM—and eventually discovered a fix: swap the board to its original BIOS version and avoid updates. This came up after I upgraded from a 12700K to a 14700K, needing BIOS support for the newer chipset. The issue returned when I swapped to a brand new motherboard, same model but different unit.
My setup includes four NVMe drives: two 1TB WD SN850x Gen4, two 1TB WD SN750 Gen3, and a 500GB Samsung 980 (DRAMless). I’ve run the system with these for years—no problems until recently. After upgrading to BIOS, I faced the same glitch: during heavy data moves or game installs, one drive would freeze mid-transfer. It would pause all activity, fail the transfer, and remain visible in Explorer but uneditable. The WD dashboard stopped noticing it. Restarting the system helped only occasionally, sometimes the drive vanished entirely after a reboot.
Interestingly, removing some drives from the system seemed to ease the problem—especially when I kept the 4090FE and SN850x Gen4 together. Taking out the 750 Gen3 or the DRAMless unit resolved the issue for me. The 850 and 750 worked fine, but mixing a 750 with a 980 caused random freezes during large transfers.
I suspect the problem might stem from overloading the PCIe lanes or insufficient bandwidth, possibly due to outdated chipset settings. I’ve tried fresh drivers and updated Intel ME, but nothing stuck. It’s possible the combination of hardware and speed is pushing the limits. I’m still trying to pinpoint exactly what’s causing it, but I’m convinced the motherboard is at the heart of the matter.
For reference, others have reported similar issues with Gigabyte Z690 boards, though no one has replied to my post. My current setup with three drives and careful data handling keeps things stable. If you’re facing this, maybe try a different motherboard or adjust your drive mix. Let’s figure this out together.
Mainstream lacks sufficient PCIe lanes to handle the demand. It seems you're hitting the maximum bandwidth you have available. You're using an x8 DMI 4.0 uplink from the chipset to the CPU, which matches roughly the bandwidth of x8 PCIe 4.0. All the drives in the chipset slots plus your other peripherals also share this connection. Only two full-speed PCIe 4.0 drives can reach that limit without issues, but adding a third drive and loading it fully causes the bandwidth to drop and performance to suffer.
@Zando_ Your response makes sense now. The September 2022 BIOS update likely introduced 13th generation support, which could be linked to the bandwidth issues. It’s possible they adjusted PCIe lane allocation for newer CPUs, possibly reducing bandwidth in M.2 slots. I’m not sure if that explains everything, but it’s a good direction to explore. The USB problems with random disconnections suggest a potential PCIe or driver issue—plugging the hub back helps, so the hardware is present but configuration matters. Regarding the ASUS Hyper M.2 X16 card, using it for Gen3 drives in M.2 slots might help avoid conflicts, depending on your setup.
Uncertainty. The 12th, 13th, and 14th generation models all appear to share the same configuration: 20 PCIe 5.0/4.0 lanes plus 8 DMI 4.0 lanes for data downlink. It seems unlikely they made any BIOS changes that would create problems. Think of it like a highway—more lanes mean higher throughput, but if you try to add too many devices, congestion occurs. Your CPU supports 20 lanes, with 16 dedicated to PCIe slots and 4 going to the M.2 slot. The M.2 ports, PCIe connections, and most I/O share those 8 DMI lanes. When many SSDs each use 4 lanes, especially at 4.0 speeds, they all compete for bandwidth, slowing everything down. Intel and AMD likely assumed most users won’t run many high-speed drives, so this setup works. However, it feels odd considering the rising demand for fast NVMe storage and the fact that mainstream CPUs are marketed to everyday users who will regularly handle such tasks. HEDT is currently limited by expensive components and a focus on workstations rather than consumer needs. For context, my i9 7980XE offers 44 PCIe lanes, allowing full GPU and four M.2 drives at maximum speed—leaving room for additional bandwidth. On this platform, only about four lanes are available, but I suspect the rest are used by the chipset. Modern HEDT models provide even more lanes, which can make the bottleneck seem more pronounced.