A level of floating point accuracy that usually requires half the space for each value than does double precision, but provides considerably less precision. For most systems running the SPEC CPU tests from the OSG (e.g., CPU95), single precision implies a 32-bit value