MLAB or M9K?

Hi there

It's getting quite crowded on my Stratix III device now (101 of 108 M9K used) and I intend to build in further data processing functions with need for intermediate storage.

As I learn from earlier threads, quartus does a quite good job in assigning memories that were selected to be "AUTO".

I notice that my FIFO_1W1R_8x18 with its 144 bits is assigned to an M9K (where it would perfectly fit into the 320 bits of an MLAB to my understanding).

Is there any advantage in using M9K in this case (from a resource, power consumption, wiring point of view) that makes quartus doing so?

Differently asked: If I require 9 more M9K blocks, which will eventually quartus push to use MLAB's for the samll memories as I expect, what drawback will I have to face?

Anyone got experience in that direction?

Regards,

Peter