SLI/Crossfire and X-Plane
From X-Plane Wiki
X-Plane, typically, does not benefit.
This is because performance boosts using these technologies depend on whether or not the computer is limited by the fill rate.
For X-Plane, at a resolution of 1024 x 768, the simulator’s speed is limited by the geometry going across the bus, not the fill rate. So, in that case, SLI and Crossfire are useless-—one card can fill 1024 x 768 at 60 fps.
Furthermore, in SLI, the two cards must communicate with one another, sending textures back and forth between them. This can be slower than one card doing all the work! This is because sending these textures from one card to another can be slower than doing a whole frame, if the whole frame can be done on one card.
So, where the computer is not limited by the graphics card’s fill rate, SLI can actually slow the simulator down, not speed it up.
There are performance testing programs (benchmarks) that can test SLI/Crossfire setups and show a higher performance with them enabled. This is because the test program does not have to draw the world from the sky when it tests (as X-Plane does), so when it tests fill rate, it does nothing else. Thus, you see only fill rate performance. This has nothing to do with X-Plane because X-Plane is typically not fill rate-limited. As well, the test program might not copy textures back and forth between the cards. Again, the performance you see in the program has nothing to do with X-Plane, which must copy cloud shadows, reflections, and the like around between video cards.
So, the bottom line is that these technologies can make the simulator slower, not faster, in many cases!
To learn more, read Benjamin Supnik’s blog here.