The Intel 80386 SX

by Lloyd Borrett
Technical Cornucopia, January–February 1989

Much has been said recently about the introduction of the Intel 80386SX or P9, but as usual little has been said about the real benefits of the 80386SX, and where it fits into the new generation of personal computers and PC based solutions that are coming our way.

What is it?

When IBM released the IBM PC back in August 1981 it was the first mass produced 16 bit personal computer. It was a big break away from the existing 8 bit Z 80, 6800 and 6502 based systems that then dominated the market. And yet at the same time the IBM PC was also an 8 bit machine.

Instead of choosing the Intel 8086 chip, IBM chose the Intel 8088 chip. The difference was that although both internally were full 16 bit processors with the same instruction set, the 8088 talked to the outside world via an 8 bit data bus, while the 8086 used a 16 bit data bus. The result was a 50% reduction in performance. But more significantly, it meant a relatively low cost new machine that could run a new generation of software.

The Intel 80386SX is to the 80386 as the original 8088 is to the 8086. Internally the 80386SX is a full 32 bit 80386 processor with the same instruction set as the real 80386. But as far as the outside world is concerned, it's a 16 bit 80286 like chip.

Why was it created?

Two reasons.

Firstly, Intel have licensed other chip makers to manufacture the 80286 chip. These other manufacturers have been gradually increasing the clock speed of the 80286 and lowering prices. Intel's share of the 80286 market has been gradually declining. By introducing a baby 386 chip Intel can increase sales and profitability.

Secondly, the 80286 is actually a flawed design. It's okay for use as a faster 8088/8086 but it doesn't support multi user multi tasking software as well as it could. The 80386 is much, much better at this.

Operating System/2 (OS/2) has a lot of code built into it to handle the multi user multi tasking and memory management instructions that are built into the 80386 chip. That code means increased overhead. Increased overhead means reduced speed.

Many software developers are already recognising this and producing 386 versions of their software to make use of the full potential and performance of the 386. It's quite likely we'll see a 386 version of OS/2 by the end of 1989.

The benefits and problems

The significance of producing a 386 with a 16 bit data bus is that it simplifies the chip. This means increased production yields. It also means simpler system design to produce a working motherboard.

With a 16 bit data bus there is no need for manufacturers to come up with yet another 32 bit bus design. They can use the cheaper 16 bit peripheral chips and circuitry. The lower number of pins on the 80386SX chip makes board design and manufacture a lot easier.

But there are some disadvantages. The 16 bit data bus means that full 32 bit memory accesses take at least twice as long. The high 16 MHz clock speed of the 80386SX means that many 80386 like speed up tricks have to be incorporated into the design. Static column RAM, interleaving and RAM caching are just some of the speed up tricks needed to allow the 16 bit peripheral chips to cope.

However, the performance difference between the 80386 and the 80386SX is not as great as that between the 8086 and the 8088. An 80386SX should only run about 10 percent slower than an 80386 running at the same clock speed. (This is because of the improved memory access design inherent in the 386 design, along with fewer wait states.)

The promise of the 80386SX

What the 80386SX promises is 32 bit processing at a speed not much below that of the 80386, but at a much lower price.

All of the special features of the 386 over the 286 remain. Manufacturers, software houses, and users can all upgrade more cheaply to 386 32 bit processing.