Hardware Reference
In-Depth Information
of 16 kilobytes of memory (imposed by the number of pins on the chip). This de-
sign resulted in the 8080, a small, general-purpose CPU, introduced in 1974.
Much like the PDP-8, this product took the industry by storm and instantly became
a mass-market item. Only instead of selling thousands, as DEC had, Intel sold mil-
lions.
In 1978 came the 8086, a genuine 16-bit CPU on a single chip. The 8086 was
designed to be similar to the 8080, but it was not completely compatible with the
8080. The 8086 was followed by the 8088, which had the same architecture as the
8086 and ran the same programs but had an 8-bit bus instead of a 16-bit bus, mak-
ing it both slower and cheaper than the 8086. When IBM chose the 8088 as the
CPU for the original IBM PC, this chip quickly became the personal computer in-
dustry standard.
Neither the 8088 nor the 8086 could address more than 1 megabyte of memory.
By the early 1980s this had become a serious problem, so Intel designed the
80286, an upward compatible version of the 8086. The basic instruction set was
essentially the same as that of the 8086 and 8088, but the memory organization
was quite different, and rather awkward, due to the requirement of compatibility
with the older chips. The 80286 was used in the IBM PC/AT and in the midrange
PS/2 models. Like the 8088, it was a huge success, mostly because people viewed
it as a faster 8088.
The next logical step was a true 32-bit CPU on a chip, the 80386, brought out
in 1985. Like the 80286, this one was more or less compatible with everything
back to the 8080. Being backward compatible was a boon to people for whom run-
ning old software was important, but a nuisance to people who would have pre-
ferred a simple, clean, modern architecture unencumbered by the mistakes and
technology of the past.
Four years later the 80486 came out. It was essentially a faster version of the
80386 that also had a floating-point unit and 8 kilobytes of cache memory on chip.
Cache memory
is used to hold the most commonly used memory words inside or
close to the CPU, in order to avoid (slow) accesses to main memory. The 80486
also had built-in multiprocessor support, allowing manufacturers to build systems
containing multiple CPUs sharing a common memory.
At this point, Intel found out the hard way (by losing a trademark infringement
lawsuit) that numbers (like 80486) cannot be trademarked, so the next generation
got a name:
Pentium
(from the Greek word for five,
πεντε
). Unlike the 80486,
which had one internal pipeline, the Pentium had two of them, which helped make
it twice as fast (we will discuss pipelines in detail in Chap. 2).
Later in the production run, Intel added special
MMX
(
MultiMedia eXten-
sion
) instructions. These instructions were intended to speed up computations re-
quired to process audio and video, making the addition of special multimedia
coprocessors unnecessary.
When the next generation appeared, people who were hoping for the Sexium
(
sex
is Latin for six) were sorely disappointed. The name Pentium was now so
