Not a hugely scientific test, I rode an hour tempo with a bit of monkeying around with cadence and resistance in erg mode. Calibrated CT and SRM after 20 minutes warm up at 150-170w.
Outcome? The SRM was consistently less than the Computrainer, but only by about 3%. Average watts for the workout from CT was 206 vs 200 from SRM.
The complete dataset is available here (edge 705 download into wko vs CT CDF import into wko then both exported as PT CSV and spliced together using the zero watt markers I placed in the workout when stretching etc).
And here is a chart.... (hi resolution version here)
I'm planning on a more scientific comparison tomorrow since the forecast is for rain and I may end up indoors :-( The geek in me wants to look at the peak power and high/low cadence variances ...