It isn't so much 'urban legend' as lack of clarity. Any application that uses the integrated Desktop Windows memory controls -are- capped at 2gig max address; those apps that have integrated control code that directly accesses the DMA controls can chew as much memory as you have. The 4 gig limit is due to that being the max a 32 bit address bus can access. But a custom coded memory routine can swipe an old trick, and use a set byte from the data bus to break the memory access into banks. At the penalty of adding a lot of non-processing time(comparatively) to set a buffer from the data bus (you've heard of server boards that address 32 gigs? That's how they do it. A simple buffer that reads an output on the lower byte of the data bus. Each line enables the readwrite for one discreet bank The OS's that use more than 4 gig are either some variant of Nix (either U of Li), or MS Server...both of which does the bank address trick.