Type and size of operands
Type : Integer, floating point: single precision float, double precision float, character.
Floating point : IEEE 754 format.
Support for decimal data : Direct decimal operations? packed decimal?
byte, word, double word, quad word, etc.
With SPEC92 suite of benchmarks, on 32-bit address machine, word is the size commonly used for integer computation and 64bit (double) is common for floating point computation.