Does connecting with NVlink help folding?

[MeeLee]

Does it help getting higher folding scores, connecting 2 cards with NVlink?

If it does, does NVlink work when connecting 2 cards (eg 2x RTX 2080) from different brands?

[JimboPalmer]

No.

F@H always considers them as two separate cards.

No technique merges multiple cards to work on a single WU.

[MeeLee]

If one card is connected to a pcie 4x, or 8x port, and the other is 4x or 1x speed, could the slower card get it's data any quicker, using nvlink as passthrough (from the card in the faster pcie slot)?

[JimboPalmer]

Perhaps. You would have to write your own driver stack, and retain compatibility with Nvidias. That is unlikely.

[bruce]

NVlink enhances the speed of video not 3D calculations. The video output is merged on its why to your screen. FAH doesn't use the video production circuitry in your GPU(s), only the calculation hardware.

Some development work has been done by others in attempting to share the computation resources, but not by FAH. See whatever information is available at gromacs.org (for CPUs) and OpenMM.org (for GPUs) The FAHCores use by FAH might incorporate those features SOMEDAY but the capability is not in the present FAH code. It can get VERY complicated (e.g. coordinating and synchronizing work from diverse hardware combinations) so it's not likely to happen any time soon. Coordination of WUs completed separately is performed routinely by FAH's servers.

[MeeLee]

Thanks!
It saves me $80 on hardware.