So the Nord used a DSP - case in point, it used a DSP which has hardware registers / accumulator, etc...designed for double floating point precision operations. That DSP had "Double precision 48 x 48-bit multiply with 96-bit result in 6 instruction cycles " - among other hardware support that even things like the M0, M3, and M4 just can't compete with. Don't just judge performance by MIPs.
Still excited to hear it though - any videos?