Two GPUs with different PCIe generations? will performance suffer?

What will happen if I try to use a PCIe 2.0 card and a PCIe 1.0 card on a single motherboard (for example nForce 750i based)? Will the faster card scale down to the lowest common speed (PCIe 1.0) or will I have both running at their best bandwidth?

I don’t plan on using SLI, I’m thinking of using a G92 card for main processing/graphics and a 8600GT for other stuff (display for CUDA or PhysX in games).

It shouldn’t–PCIe is a point-to-point interface, so there’s no reason why a Gen1 card in one slot should impact another slot’s performance. But that’s theoretical; I’ve never tried it. Let us know how it turns out…

I used to have a GT200 prototype card and an 8800 GTX inside the same motherboard with the AMD 790FX chipset. The PCI-e 1.0 and 2.0 devices transfered data at the different speeds you would expect.

As long as the motherboard is configured to autodetect the speed in all slots (true for the Gigabyte motherboard I used, but not true for the MSI motherboard until I flipped some BIOS settings) it should be fine.


My 2 cents:

Although PCI-E is one-to-one or whatever — I still think the RAM would still be a bottleneck.

Unless the RAM controller can service different memory requests simultaneously (+ system bus support for multiple talkers…) – there would still be a drop in performance – as far as memory copies are concerned!