
Originally Posted by
churchy
ElGreco: imho 0.25-0.28gpm should be minimum for gpu full cover blocks. If parallel, that means you nead at least 0.5 gpm in overall loop with two cards in parallel and 1 gpm with four in parallel. If you get less with four, i'd sugest to connect in parallel by pairs of cards this way both lowering a bit flow resistance yet not dropping partial flow through each card too much. (imho it would benefit for most vendors that make SLI/CF bridges to design this way their longest ones for biggest number of cards). (and maybe middle length for three cards as well, as in first cards in parallel, last one in serial. but that's my wild guess, in my mind compensating with higher flow in last one a little bit preheated water by first cards)
Bookmarks