It's hard to say which is a more efficient use of fissiles. It's easier to burn up all your fuels in the context of an explosion as in Orion than in a saltwater rocket. On the other hand by bulking out the fuel with water to form a larger mass of propellant the saltwater uses it's energy more efficiently at the cost of lower ISP.

But really, if we're willing to leak radioactive byproducts I think an open cycle gas core might be the best bet in terms of not-yet-developed rockets.

http://www.projectrho.com/public_html/rocket/enginelist.php#...

The big advantage of Orion is that we're reasonably confident it would work, and could probably build one within the decade with a serious effort. Zubrin's NSW rocket, on the other hand, falls into the "hasn't yet been conclusively proved impossible" category. We're not sure if we can control the reaction, and we're not sure how much heat would be transferred to the rocket if we can.

Orion would have exhaust velocity ~1000-10000km/s. Compare that to NSWR's 66km/s. NSWR seems fast compared to today's chemical rockets but it's nothing compared to Orion.