Intel 80287

The intel 80287 was a floating point co-processor to intel's 80286 models of single chip computer. The 80287 was introduced in 1982.

Integer arithmetic requires much less silicon real estate to implement than calculations on values that had a fractional component. Intel off-loaded the hardware to perform those instructions to a co-processor -- a support chip devoted solely to that task. Computers could be equipped with an 80286, and without an 80287. If the 80287 were present floating point instructions would be off-loaded and executed relatively quickly, by the 80287. Unlike intel's earlier floating point co-processors, the intel 80186 and the intel 80386, the 20286 generated a specific hardware interrupt when it encountered a floating point instruction. This meant programs did not have to be specifically compiled to use the co-processor. If the co-processor was present, the hardware interrupt would trigger it to perform the floating point operation. It would pause the CPU until it was finished. If the co-processor was not present the interrupt would be processed by an interrupt service routine, which would perform the floating point operation in software. The program would not be able to tell how the floating point operation ws performed.

Some sources report the intel 80287 used the same execution unit as the intel 8087. Other sources report the two chips had very different number of transistors.