Remembrance of chips past

-April 15, 2014

Sometimes, we forget the implications of Moore's Law, and just how amazing our IC technology is compared to yesteryear's. Pack-rat that I am, it's no trouble for me to peruse what used to pass for high-tech – and now, you can have a look at it too!

IC packages used to be prettier, I think. Lots more gold and white ceramic happening. Packages that look as though they could go to outer space without breaking a sweat.

From top: Labelled "VMMU," from 1989. It looks handmade! What do you think? Part of a mainframe or supermini? A 1970 Teledyne chip – what is it? A CMOS 4049 from 1974; such a fancy package for six measly inverters. And a "100k" series ECL chip from 1979.

FPGAs commonly have multipliers scattered throughout the programmable fabric. Large chips often contain significantly more than 1,000! To me, that is one of the more amazing things about these ICs – that a multitude of multipliers might be left sitting idle in your design, without a second thought (and fast ones, too – maybe 500MHz).

Thirty years ago, multiplier chips (yes, just one multiplier per chip) tended to be big, expensive, hot (a few watts), and, by today's standards, slow (10MHz). Now, we can use them in our designs as easily as if they were a NAND gate.

High-speed ADCs have always been a premium item, but the definition of "high-speed" has changed a bit. Thirty years ago, the TRW TDC1007 pumped out 8-bits at 20MSa/s and burned a few watts. The package just exudes priciness. Now, a part with that performance will run cool and set you back a few bucks. It's not just digital that rides Moore's curve.

From top: TRW 8-bit 20MS ADC, with a flat heatsink top. The famous 68000 processor, as used in the first Macintosh computer. Did it really have 68,000 transistors, or was that just marketing? The archetypal multiplier, TRW's MPY16, from 1977. The nice finned heatsink is part of the package.

The package for Signetics' 8X300 high-speed microprocessor from 1976 looked a lot more striking back in the day than it does to modern eyes. It was an odd chip – fast for the time, at 8MHz RISC – with a very primitive set of 16 instructions. The one application I saw it in was a color video display controller.

A 1979 8X300, and a Cypress 16x16 MAC (multiplier-accumulator). Looks like this puppy ran fast – 15MHz in 1988!

Microcontrollers that used mask-programmed ROM had an issue when it came to developing with them. Here's how Mostek solved it.

Mostek 3870 processors with their piggyback cargo, 1983.

When IBM introduced its "PC-AT" a few years after its original PC, space for all that "advanced technology" was at a premium. So no, those stacked DRAM DIP packages below are not trying to make baby DRAMs. They're actually a pair of 64 kilobit chips soldered together to up the density to 128kb. And you thought 3d IC packaging was a recent innovation. The chips aren't identical, of course – there's some small pinout or signal-polarity difference that allows them to act independently. Who can tell me?

Some PALs were one-time programmable, some were electrically erasable, and at least one manufacturer made UV-erasable chips.

Top: A Fairchild 741 op-amp from 1969, only a year after its introduction. A full-custom analog SSR controller based on my own discrete design, 1988. A DH0008 high-current driver in a rare 10-pin DIP.
Middle: Motorola 2k×8 EPROM. Not an IC, but a cool video delay line. The leads connect to what I assume are input and output piezo transducers on the mitred edge. The screened bits of thick ink on the quartz plate define diagonal paths through it, and the video (modulated onto a carrier) bounces around like a light beam!
Bottom: The Cypress UV PAL, 1989. The stacked DRAMs from the PC/AT.

What interesting chips are in your collection.

A hybrid, high-voltage isolation, DC-DC converter, with the lid popped off. That toroid looks like the Colosseum under a stereo microscope.

Also see:

Follow Michael Dunn at

Loading comments...

Write a Comment

To comment please Log In