It occurred to me that Havard Eidnes wrote in gmane.os.netbsd.ports.x86-64:
> On my local amd64 test machine, which does little else than run ntp
> using the GENERIC kernel, all the kernel processes have 8028K in top's
> "RES" coloumn, so if he is indeed seeing 233M for each and every
> kernel process in the "RES" coloumn, at least that's not what everyone
> else is seeing.

fwiw, on my amd64 machine (Currently running a bulk build) it's 8700-ish K.