I think your mistake is trying to manage all your data in one chunk – rarely does anyone have any need to work with all assets at once. Although I agree that none of the code repository software handles large binary data well (at least SVN and Perforce)

I actually recently blogged a bit about this here:
http://systematicgaming.wordpress.com/2009/06/24/asset-management/

(although on a quick re-read it’s a pretty meandering article)

At my studio, it has turned into quite the battle to get everything synced up properly. Often, several people try to add new files with the same names which causes a lot of problems in addition to the usual bad check-ins. Locking/corruption issues are also much more frequent than I’d like them to be, and none of these problems tend to surface in the source tree.

It’s also fairly obvious that SVN in particular really wasn’t made to handle big repositories with gigabytes of data, as it tends to grind to a halt when dealing with them. Even when located on the same intranet as the server, a fresh check out of a 5-6GB repository commonly takes several hours (at least here). And 5-6GB really isn’t much these days.

We’re always on a look out for a better solution, but have yet to find any. Alienbrain looks interesting, but for us at least it’s a major downside that the server doesn’t run on *NX platforms since that’s what our dev servers are running. We might try the evaluation at some point though to see if it’s worth it.