Math Co-Processor

Hardware entity

Won't run without math co-processor hardware. Some early IBM compatibles required a separate chip. Not always include in some later reduced cost CPUs.


The first video game about Math Co-Processor was released in 1995.

This article is way to IBM-PC compatible specific. Some Macintosh, Amiga, MSX, Atari, Commodore, RISC, Amstrad, Android (the kernel looks for one, but this may be GPU related), Apple ][gs, Apple ][ (yowzers!), Coin-ops, HP Calculators (some of these have MCP sockets!), and NeXTstep information is greatly needed.

These are mainly for performing floating-point operations that would otherwise be done in software by an application. Many applications (and games) use the co-processor to operate at maximum speed but happily run slowly if there is no co-processor is found. However, if speed and/or timing is critical, calculating in software may not be an option.

Math co-pocessors often include exponential, logarithmic, trigonometric functions. Useful for such things as calculating with a very precise Pi (say, 50 digits or more for example). A 2D game of the 386 era or previous, shouldn't need anything more precision than 3.14.

Early IBM compatibles had a socket for an 8087 math co-processor. This was a popular add-on for customers using CAD or working with massive spreadsheets. Generally, games did not run any better with the 8087. In the 486 era, math co-processors were integrated. A "DX" model CPU included an MCP; "SX" models did not. Games of this era generally did run slower without an MCP. Some game required it. Pentium CPUs and later all have MCP functions in the main CPU. However, modern games may still be made to utilizes a co-processor for math. They use the GPU on the 3D graphics accelerator card for this. Do not apply mathcoprocessor to these games as there are specific groups for that.