GeForce RTX 5060 Ti PCIe Lanes Problem: Do You Have This Issue?

While testing the GeForce RTX 5060 Ti, I encountered a recurring issue with active PCIe lanes. On PCIe 5.0 motherboards, the card often fails to run in x8 mode even in a x16 slot, starting instead in slower x1, x2, or x4 modes. The same behavior occurred across three RTX 5060 Ti models from MSI, Gainward, and Asus. The issue likely stems from how these motherboards handle compatibility with PCIe 5.0 graphics cards using eight lanes.

How to Tell if Your GeForce RTX 50 Is Affected by a PCIe Speed Issue?

The most reliable method currently is the PCI Express Feature Test from 3DMark. Unfortunately, it’s only available in the paid 3DMark Advanced Edition or higher. This test measures PCIe bandwidth in a dedicated benchmark.

The following table shows the theoretical maximum bandwidth for each PCIe version and corresponding lane count:

Version Line code Transfer rate
(per lane)
x1 x2 x4 x8 x16
1.0 8b/10b 2.5 GT/s 0.250 GB/s 0.500 GB/s 1.000 GB/s 2.000 GB/s 4.000 GB/s
2.0 8b/10b 5.0 GT/s 0.500 GB/s 1.000 GB/s 2.000 GB/s 4.000 GB/s 8.000 GB/s
3.0 NRZ 8.0 GT/s 0.985 GB/s 1.969 GB/s 3.938 GB/s 7.877 GB/s 15.754 GB/s
4.0 128b/130b 16.0 GT/s 1.969 GB/s 3.938 GB/s 7.877 GB/s 15.754 GB/s 31.508 GB/s
5.0 128b/130b 32.0 GT/s 3.938 GB/s 7.877 GB/s 15.754 GB/s 31.508 GB/s 63.015 GB/s

3DMark results aren’t 100% accurate. For example, during RTX 5090 testing, the test allegedly showed around 121 GB/s due to a driver bug – twice the theoretical maximum of PCIe 5.0 x16.

However, with the RTX 5060 Ti and recent drivers, measured slot speeds approximately matched the number of active lanes as detected by diagnostic utilities. This held true across all four systems tested. Measured speeds varied slightly but were within a few GB/s of theoretical values.

In the 3DMark DirectStorage feature test, the card actually appears to perform better in PCIe 5.0 x1 mode than in PCIe 5.0 x8. However, the final score reflects how much the GPU accelerates data decompression during transfers from storage to the card’s memory. What we’re focusing on here is the speed at which data moves from storage and system memory to video memory—specifically, Storage to VRAM and RAM to VRAM—and in these tests, the bandwidth over PCIe 5.0 x1 is several times lower.

DirectStorage Feature Test, PCIe 5.0 x1 vs PCIe 5.0×8

One simple, though not entirely reliable, way to check the bus speed is through GPU-Z diagnostics. It’s less certain mainly because, until the recent update to version 2.65.1, GPU-Z 2.64.0 incorrectly reported the RTX 5060 Ti as running at PCIe x16 5.0—even when only eight lanes were active. That didn’t exactly boost confidence in the accuracy of the lane count.

Version 2.65.1, which fixes this and adds RTX 5060 Ti support, was released just before the card’s official launch.

Bandwidth testing is also possible in AIDA64, but the current beta version (7.65.7404) doesn’t yet support the RTX 5060 Ti, so tests won’t run.

You could also measure bandwidth using developer tools for OpenCL and CUDA, though I didn’t install those on my test bench.

Alienware vendor locks Ryzen X3D CPUs, effectively bricking chips

AMD processors have an unfortunate feature: the PSB security technology that makes it possible to lock a CPU so that it cannot be used in any motherboard other than the one that locked it with its key. This typically makes a CPU unable to boot anywhere else and this lock is irreversible. It used to affect just server processors, Threadripper and Ryzen Pro chips, but beware: Dell is apparently locking CPUs in gaming PCs as well. Read more “Alienware vendor locks Ryzen X3D CPUs, effectively bricking chips” »

New Radeon GPUs Likely Won’t Arrive Until Second Half of 2027

We recently reported the new generations of GPUs from Nvidia (GeForce RTX 6000 / Rubin) and AMD (RDNA 5) will not arrive until 2028. There was supposed to be one exception, however—AMD could reportedly release one of the new-generation cards in 2027. The information at the time did not say when exactly in that year this might happen. Now we have clarification, and it seems it will not be anytime soon either—as might have been expected. Read more “New Radeon GPUs Likely Won’t Arrive Until Second Half of 2027” »

AMD Reportedly Raising GPU Prices. Apart From One Radeon card?

Radeon GPUs have been significantly better value than similarly fast Nvidia cards lately. These lower prices (by today’s dismal standards) may not last much longer, however. According to news from China, AMD and AIB makers are now preparing a substantial price increase, despite similar price move happening recently. The reason is the AI mania again, which has massively driven up the cost of memory, as well as manufacturing itself. Read more “AMD Reportedly Raising GPU Prices. Apart From One Radeon card?” »

Comments (6) Add comment

  1. The 8GB version is the one you should be testing for PCIe bottleneck, not the 16GB.

    1. I don’t think it matters much in this case. Interestingly, the biggest performance drops happen with settings that are actually less demanding on memory.

  2. Mobos Asus strix z370-h
    CPU QTJ1 (i9-10980HK)
    Ram 2x16gb ddr4
    Rtx5060Ti 8gb…

    Damn! Always downgrade to pcie 2.0 instead pcie 3.0 like my old rtx3060ti!
    Windows 10 (last driver) and debian Linux bookworm kernel 6.12.32 (driver 575)

    Considering plus last firmware official

    1. It’s difficult to say whether Gigabyte has already resolved the issue. Adam, who was dealing with this topic, no longer works for HWC. However, if you are still observing similar behaviour, then it probably hasn’t been completely fixed yet. 🙂

  3. Is this issues fixed? i seems to have similiar issues with my z490 motherboard and rtx5060ti 16gb, do i need to reflash my bios using different gpu , and then plug the 5060 back?

Leave a Reply

Your email address will not be published. Required fields are marked *