Is it mainly a matter of too many layer?
Or algorithms don't sufficiently flexible to be able to use at best the hardware?
Or both?
Note:
I speak for situations when the documentation is present as happen on Intel (at least on most part) and partially on Amd.


Reply With Quote
