@BB
The answer I got from SM was that synchronisation between the cores calculations was the main bottleneck.

A render can be cut in pieces and distributed as buckets among cores, but a cloth calculation that has to be one fluent movement calculated among several cores is not so easy.

sorry Enlish is not my native language, but You know what I mean.

Poser 1, 2, 3, 4, 5, 7, P8 and PPro2010, P9 and PP2012, P10 and PP2014 Game Dev
"Do not drive faster then your angel can fly"!