Originally Posted by gurusan
hmm ok that makes sense....but then what allows you to use cards with different amounts of texture memory?
each GPU/card renders its part (or parts) of the screen independently; then one of the cards joins each cards part (or parts) together
Even if they have different amounts of texture memory, each card would render their parts the same way as they would without crossfire; If a card doesn't have enough local memory to store all its textures, it would have to acccess textures from main system memory.
If one card has less texture memory, and it has to use main system memory, it's probably going to run slower than the one with a higher amount of texture memory, and therefore it will slow down the other card, if it has to wait longer for all the rendered data.