The Engineering Book: From the Catapult to the Curiosity Rover: 250 Milestones in the History of Engineering

1966

Dynamic RAM

Robert Dennard (b. 1932)

Every computer needs RAM, or Random Access Memory. The central processing unit of the computer needs a place to store its programs and data so it can access them quickly—at the same pace that the CPU’s clock is operating. For each instruction the CPU (Central Processing Unit) executes, it must fetch the instruction from RAM. The CPU also moves data to or from RAM.

Imagine you are an engineer looking at computer memory options in the late 1960s. There are two possibilities. The first is core memory, which is made by weaving tiny ferrite donuts into a wire mesh. The problems with core memory are many; it is expensive, heavy, and enormous. The second possibility is static RAM made from standard transistor circuits. It takes several transistors for each memory bit, and given the state of integrated circuits at the time, it is not possible to put much memory on a chip.

But in 1966, American electrical engineer Robert Dennard, working for IBM, tried something different in the interest of reducing the number of transistors and fitting more memory cells on a chip. He explored the idea of dynamic RAM using a capacitor to store one bit of data. When the capacitor is charged it represents a 1, discharged it represents a zero. On the surface this seems ridiculous, because capacitors leak. If you store a 1 in memory made of capacitors and do nothing, the capacitor will leak and forget the 1 in less than a tenth of a second.

But the advantage is that this approach greatly reduces the number of transistors, and therefore increases the number of memory cells on a chip. To solve the leakage problem, all of the capacitors are read periodically (for example, every few milliseconds) and rewritten, thus refilling all of the leaking capacitors containing 1s with a full charge. This approach is known as dynamic RAM (DRAM), first manifested in 1970, because it must be dynamically refreshed to keep the capacitors charged.

The dynamic RAM approach yields memory cells that are so much smaller, and therefore less expensive, than static RAM that every desktop, laptop, tablet, and smart phone today uses DRAM. It is a great example of the way engineers can reduce costs by embracing ideas that may seem initially ridiculous.

Pictured: Dynamic SDRAM memory for a computer.