If you have a nuclear reactor, the power vs mass loss could end up working in favor of the emdrive, especially for long range interstellar flights. None of this is, of course, practical anytime soon...
1. Details on submarine nuclear reactors are classified, but five minutes' Googling shows that a reasonable guess for their total mass (including shielding) is 1,000 tons. Meanwhile, the entire payload to LEO of the largest rocket ever successfully launched, the Saturn V, is only 155 tons -- and that's for the entire top stage.
2. Submarine nuclear reactors use water for cooling. Water that is not available in space. Cooling would be a huge problem.
So, existing submarine designs are not practical. You'd need to design something from the ground up that is much lighter, and the cooling system would be entirely different and likely more massive, since you don't have all of that free water available to dump heat into. Instead you're talking massive radiators.
But any mission involving humans is likely to carry a large amount of water beyond the crew's personal needs, because it makes such a good radiation shield. So presumably the same water would be used for cooling the reactor.
A 3 MW reactor puts out a hell of a lot of heat, and without the benefit of air-cooling in space, I'm not sure what exactly you would do with all of that waste heat. Consider how massive the space shuttle orbiter's radiators were (they are on the inside of the cargo bay here: http://i.stack.imgur.com/Flgzb.jpg ), and all of that is only capable of shedding waste heat in the amount of ~6 KW! We can put a much more capable reactor into space than we can possibly cool, so we haven't bothered. Cooling is the real problem. The total mass of the radiators and the structure required to support them ends up being way more than the reactor itself.
So for good long distance transportation in space, not only do we need a working, efficient emdrive, but we also need better power generation that is much more efficient from a waste heat perspective. These are really hard problems.
There are several other reactor designs that have been researched over the last half century that can operate in a closed loop with much smaller amounts of nuclear fuel and more manageable radiation emissions. Designs like the nuclear lightbulb, which was researched by United Aircraft Corporation and NASA for almost a decade before the plug on the Mars mission was pulled in the 70s, are much better suited and are what proponents of active nuclear propulsion most often have in mind. Once there is some political will, we have decades old research to start from to build a flight capable nuclear reactor.
You can seed expanding liquid hydrogen in the outer cavity with tungsten nanoparticles which absorb the UV radiation to heat the hydrogen for use as a propellant. You can pump the outer cavity with a UV transparent coolant and line the walls with parabolic photovoltaics (that convert UV instead of visible spectrum) for direct conversion of the black body radiation to electricity. You can theoretically even create a magnetohydrodynamic "turbine" in the outer cavity that is coupled to the spinning nuclear fuel (which can be charged plasma).
You may be thinking of an earlier that was mistakenly called nuclear lightbulb or another design that was lumped into the concept. The variation I have in mind is just as general purpose as any other nuclear reactor, it just uses a lot less fuel and has to actively maintain temperature and pressure to keep the neutron cross section energy high enough to sustain power positive fision.