IBM SG24-4576-00 manual External Interfaces, Processor Types

performance. (Please see 1.3, ªMemoryº on page 3for a discussion of memory speeds and system performance.) The point is that you cannot compare system performance by simply looking at the speed at which the processor is running.

A 90 MHz machine with a set of matched components can out perform a 100 MHz machine which is running with slow memory. IBM PC Servers are always optimized to incorporate these factors and they always deliver a balanced design.

1.1.2 External Interfaces

The processor data interface, or data bus, is the data connection between the processor and external logic components. The Pentium family of processors utilizes a 64-bit data bus, which means that they are capable of reading in 8 bytes of data in one memory cycle from processor main memory. The Intel 486 has a data bus of only 32-bits, which limits its memory cycles to 4 bytes of data per cycle.

The width of the processor address interface, or address bus, determines the amount of physical memory the processor can address. A processor with a 24-bit address bus, such as the i286 class of processors, can address a maximum of 16 megabytes (MB) of physical memory. Starting with the i386 class of processors, the address bus was increased to 32 bits, which correlates to 4 gigabyte (GB) of addressability.

1.1.3 Processor Types

IBM currently uses two processors in the PC Server line:

∙80486DX2

The 80486DX2 has a 32-bit address bus and 32-bit data bus. It utilizes clock doubling to run at 50/25 MHz or 66/33 MHz. It is software compatible with all previous Intel processors. The 80486DX2 has an internal two-way set associative 8KB cache.

∙Pentium

The Pentium has a 32-bit address bus and 64-bit data bus. It has internal split data and instruction caches of 8KB each. The instruction cache is a write-through cache and the data cache is a write-back design. The Pentium microprocessor is a two-issue superscalar machine. This means that there are two integer execution units (ALUs) in addition to the on-board floating point unit. The superscalar architecture is one of the key techniques used to improve performance over that of the previous generation i486 class processors. Intel was able to achieve this design while maintaining compatibility with applications written for the Intel i368/i486 family of processors.

Note: A superscalar architecture is one where the microprocessor has multiple execution units, which allow it to perform multiple operations during the same clock cycle.

2NetWare Integration Guide