"This is a format that came into standard usage with the advent of the Intel '287 and '387 math coprocessors and then the 80486 wherein the math coprocessor was built onboard the CPU."

I believe the x87 coprocessors actually use an 80 bit floating point representation - they don't work natively with IEEE754 doubles. At least one compiler I use has a switch to force conversion between the coprocessor format and IEEE double at intermediate points during evaluation of a complex expression so that the loss of precision is increased to make the results consistent with those from a 'true' IEEE implementation.