In MySQL 5 Certification Study Guide, Numeric data types are described as below:
The floating-point data types include FLOAT and DOUBLE. Each of these types may be used to represent approximate-value numbers that have an integer part, a fractional part, or both. FLOAT and DOUBLE data types represent values in the native binary floating-point format used by the server host's CPU. This is a very efficient type for storage and computation, but values are subject to rounding error.
FLOAT represents single-precision floating-point values that require four bytes each for storage. DOUBLE represents double-precision floating-point values that require eight bytes each for storage.
You can specify explicit precision and scale values in the column definition to indicate the number of significant digits and the number of decimal places to the right of the decimal point. The following definitions specify a single-precision column with a precision of 10 digits and scale of 3 decimals, and a double-precision column with a precision of 20 digits and scale of 7 decimals:
If you specify no precision or scale, MySQL represents values stored in FLOAT and DOUBLE columns to the maximum accuracy allowed by the hardware of the MySQL server host. The following definitions include no explicit precision or scale:
Floating-point values are stored using mantissa/exponent representation, which means that the precision is defined by the width of the mantissa and the scale varies depending on the exponent value. The result of these factors is that stored values are approximate.