Performance and microarchitecture of the i486 processor

The i486 microprocessor includes a carefully tuned, five-stage pipeline with an integrated 8-kB cache. A variety of techniques previously associated only with RISC (reduced-instruction-set computer) processors are used to execute the average instruction in 1.8 clocks. This represents a 2.5× reduction from its predecessor, the 386 microprocessor. The pipeline and clock count comparisons are described in detail. In addition, an onchip floating-point unit is included which yields a 4× clock count reduction from the 387 numeric coprocessor. The microarchitecture enhancements and optimizations used to achieve this goal, most of which are non-silicon-intensive, are discussed. All instructions of the 386 microprocessor and the 387 numeric coprocessor are implemented in a completely compatible fashion