64KB

3-1/4 silicon wafer with an astounding 109 64k memory chips etched on its surface. Circa 1978. Prices of these leading edge chips dropped from $25 apiece in 1981 to $5 each.

 

December 4, 1985, Los Angeles Times:

Six Japanese semiconductor firms “dumped” memory chips in the U.S. market in violation of tariff laws, the Commerce Department ruled Tuesday in a preliminary decision.

The U.S. semiconductor industry has been battered by plummeting prices for more than a year, a phenomenon that it blames on Japanese pricing tactics as well as a worldwide slump in demand. A trade group claims that the slump has cost 54,000 jobs this year.

The case involves a type of semiconductor called a 64K DRAM, (for dynamic random access memory). It can store about 64,000 pieces of information and is the most common type of chip in computer memories.

Prices for the 64K product plummeted to as low as 35 cents apiece from $3.50 within 18 months, with disastrous financial consequences for some U.S. firms.

To provide a frame-reference, an inexpensive 256GB memory stick currently retails for about $10 at Walmart.

There are one million "KB" in one "GB" so the 256GB device is 4 million times larger than the 64KB chip.

Put another way, the 64KB capacity is analogous to a two-seater outhouse while the 256GB device is a city of 8 million people...like New York City. I will avoid the temptation to compare the quality of the contents in either case. My readers are better than that.

What does four-million times more capacity and faster access "buy" the consumer?

In all fairness, it buys you less power consumption and faster access times because the discretes on the device are smaller.

It buys you more space to put "code".

Increased "code" means that the systems the chips are built into have the capability of being coded to recognize and adapt to rapidly changing events or novel-but-not-unique conditions.

Example: There were two early contenders for gasoline engine control algorithms.

One example looked at a gasoline engine as an air pump. It measured the absolute pressure in the intake manifold between the throttle-plate and the intake valves. It measured the engine RPM. The algorithm used those two bits of information to "look-up" the amount of oxygen passing through the engine and adjusted the injector's pulse-width to modulate the amount of gasoline to match the available oxygen. The system used an interpolation scheme to guess values between the ones in the look-up table. The algorithm used data from the O2 sensor that monitored the exhaust gas chemistry to slowly update the look-up table(s) to comprehend clogging of the catalytic converter, wallowing out of cylinder walls and other aging-related phenomena.

It was a simple, robust approach that was not computationally intensive. In my humble, out-dated opinion, this system worked spectacularly well in 99.9% of the situations drivers encountered.

The other system used an insulated post covered with a heated foil film and thermometer in the intake stream to measure the instantaneous velocity/temp/density of the incoming air-flow. The O2 sensor data was used to rapidly update the look-up table.

The advantage of this system is that you could run out of standard gasoline fuel in Poontang, Minnesota and after you refilled with E-85 the system would rapidly recognize the lower quality fuel and make the necessary adjustments in terms of the amount of fuel to inject (assuming you could get the vehicle to start). It should be obvious that the second approach is more computationally intensive.

And once you, the person responsible for programming the engine control system, have virtually limitless computational capability, you can do things like monitor the spark/ignition system on a cylinder-by-cylinder basis. If the system recognizes that cylinder Number 3 is behaving abnormally (say a squirrel chewed the insulation off of the spark-plug wire) then it can shut off fuel to that cylinder to save the catalytic converter and reduce the risk of emitting excessive pollutants.

"Limitless" refers to the fact that there is more room for "code" than you have competent coders to fill that space and more speed and memory to run said code than that code can demand.