The second gen focusrite scarlett interfaces have decent round trip latency, but not quite as good as the new motu interfaces.
Get 2nd gen if you are going with focusrite, or the MK4 if you go motu.
Neither are actually that exceptional on round trip latency compared to RME.
See here for focusrite figures at various buffer sizes and sample rates, on windows and OSX:
https://us.focusrite.com/scarlett-in-depth
Motu claims 1.6ms round trip, at 32 samples buffer, 96K, but they don't specify if it can do that in windows as well as OSX. I imagine the results would scale from there to be slightly lower than focusrite at higher buffer settings.
Focusrite claim 2.74ms at 32 samples 96K in OSX, and 3.22ms in windows.
As for how much system load they can take at these settings compared to each other, I have no idea. I know that MOTU interfaces I've used are stable at high loads.