Without knowing what applications you will run on the computer, the better performance bet would be on a single higher-performance video card (whether it has more memory, faster Graphics Processing Unit chip, or what-have-you), rather than going with two lower-performance cards with Crossfire, simply because not all applications can make use of the Crossfire dual-card setup, and will just use one of your two cards while the other one sits and does nothing. Since these applications will be using one lower-performance card in that scenario, you'll get better performance just going with a single higher performance card.

That's the "in general" answer, but if the primary use of the computer is to play one or more games that *do* make use of the Crossfire dual-card setup, then the opposite may be the case, and you would likely get better performance with two lesser Crossfired cards than a single higher performance card.

A previous answer is correct that just the memory of a video card is not a good spec to use for comparing two cards. Fortunately, other people have gone to the trouble to test lots of video cards, so you don't have to! One of the best sites to view and compare video card statistics is Tom's Hardware guide here: http://www.tomshardware.com/charts/gaming-graphics-charts-q1-2008/3DMark06-v1.0.2-HDR-SM3.0-Score,538.html

It additionally categorizes cards by chipset manufacturer and price range to make it easy to quickly find and compare different cards.

Another thing to consider is that high-performance video cards tend to be power hogs, so using two Crossfire cards might require a higher output power supply than going with a single high-performance card, which would be another expense to consider when deciding on cards. Tom's Hardware also explores the power requirements of video cards here: http://www.tomshardware.com/reviews/geforce-radeon-power,2122.html

It really depends on which card.

Memory is probably the least important factor when comparing two cards.