Problem with the x86

The exponential part has always 15 bits.
Overflows and underflows, to occur, do not occur.

Ex.)  These double-precision operations result in ...

In case of SPARC:  +∞.     When (2) operates, an overflow occurs.
In case of x86:         2 ** +1023.     No overflow occurs.
Regarding underflow, there is another problem, 
     double rounding.