Why not multiple precision in Pentium?

I am trying to write a little piece on the virtues of multiple precision arithmetic software. Unfortunately, I cant seem to come up with much solid information.

Does anyone have any historical info regarding Intel's decision not to include multiple precision arithmetic in the CPU? In other words, how did they come to decide on 32 bits for integers and 64 bits for doubles and so forth? Or am I looking at it backwards? Perhaps this is an inherent feature of the programming language, C and C++. While it is obviously possible to write multiple precision arithmetic routines using C and C++, are there any languages that inherently contain native arbitrary or multiple precision data types?

How has multiple precision arithmetic proved superior to the customary PC processor arithmetic, other than for encryption and pedogogical purposes? In other words, when would anybody have the need for more than 64 bit data types?

Any and all ideas greatly appreciated. Thanks.