This is the internal INTEL and MOTOROLA floating point "extended precision" that uses 64 bits mantissa, 15 bits exponent and 1 bit sign in their math coprocessors.
This data type has been standardized by IEEE as ten bytes floating point formats.
You can access to this option for X86 processors in assembler using the types "TBYTES" or "REAL10".
Specific assembler instructions handle conversion from extended precision to standard precision.
For more info see
http://en.wikipedia.org/wiki/Extended_precision and INTEL assembler manuals.