Folding Forum

For instance, if I have a board that will only do x16/x4 on two slots, would moving to a board that does x8/x8 make any difference?

It is complicated.

Windows is much more limited by PCIE bandwidth than is Linux, if you can, use Linux for multi-GPU setups.

You will notice reduced Points Per Day with less than x8; older, slower cards will be effected less.

I feel sure real experts will wake up and give better answers eventually, I just used one GPU per box.

With Windows, you might notice a small benefit but not one that's worth replacing a motherboard for.

With Linux, there's no noticeable difference between a machine with x16/x4 and one with x16/x16. My x4 slot has a TitanX in it and it runs fine, in fact it's faster than the 1070 in the x16 slot by a significant margin.

Thanks for the replies. I'm using Windows for the ease of RDP in managing things. Not sure if that's doable in Linux, but I might give it a look.

If the GPUs are identical, you can compare results yourself. If they're NOT identical, be sure the slower one is in the 4X slot.

One issue with RDP, it can shutdown GPU folding while you are connected using RDP.

If your looking for ease of managing the folding machine its easy to use Advanced control on a windows machine to control your Linix machine.

Welcome to the F@H Forum CaptainHalon,

Previously, I have used TeamViewer on a headless Linux system (headless was after I fully configured it) to monitor it from my Windows system. It didn't kill the GPU folding but that was back in the era of FahCore_15. I haven't tested it with FahCore_22 as I no longer have that setup. However, you can give that a go instead of RDP.

Thanks for the suggestion on Teamviewer. Does the particular distro of Linux matter much?

Shouldn't matter at all. I have one Linux Mint 18 and one on 19. Both work really well.

At work, we have always used TightVNC for Linux remote desktop. It's pretty much the industry default for Linux from what I have seen. They even have a Windows client so you can remote from Windows into your Linux boxes.

You might play around with TightVNC or TigerVNC and see how it goes!

Just remember that each generation of PCIe has double the throughput of the previous generation; e.g., a PCIe Gen4 x4 slot has the same throughput as a PCIe Gen3 x8 slot.

Yes, but only if both the slot and the device support Gen 4.

jrweiss wrote:Just remember that each generation of PCIe has double the throughput of the previous generation; e.g., a PCIe Gen4 x4 slot has the same throughput as a PCIe Gen3 x8 slot.

So actually I should be able to use let's say a 1080ti under Linux on a Pci-e gen2 16x without performance loss. Right?

jrweiss wrote:Just remember that each generation of PCIe has double the throughput of the previous generation; e.g., a PCIe Gen4 x4 slot has the same throughput as a PCIe Gen3 x8 slot.

That's true, and x4 is probably fine, but also what concerns me is when two slots are running x16 and x4, then the latter lanes are going through the PCH rather than directly to the CPU. With x8/x8, the lanes are strictly CPU lanes. Now whether or not the PCH introduces significant latency, I'm not sure.