Comparing X16 2.0 and X8 3.0 models.
Comparing X16 2.0 and X8 3.0 models.
It's the PCIe version, as shown by the GPU-Z screenshot in the original post. PCIe x16/x8 3.0 You've already tested it yourself and found no performance difference(unless I'm somehow misinterpreting the part of your original post where you stated the result in a Unigine benchmark was the same). GPUs hadn't been able to fully use PCIe 3.0 x16 for gaming by the time PCIe 4.0 came out, including your 1080Ti. This is why you're not seeing any performance difference going from PCIe 3.0 x16 to 3.0 x8. It's fine as it is now.
The second PCI-E X16 port is designed physically for X8, despite its X16 size. In the detailed image of the motherboard, you’ll notice the slot pins extend only halfway. The CPU can handle 16 PCI-E lanes. With two GPUs, the first slot reduces the lanes to 8x, allowing it to pass 8x to the second slot. The second slot also supports up to 8x electrically. In terms of bandwidth, PCI-E X16 2.0 equals PCI-E X8 3.0.
I interpreted it differently, thinking the second PCI-E port would function at x8 only with two GPUs installed. I expected it to work at PCI-E x16 3.0 using a single graphics card on the second slot, but it seems performance remains consistent regardless of GPU count. It might depend more on the specific GPU model than the number installed.