Fifty seven years ago today, the US Patent Office granted a patent for a new memory technology to IBM. This approach used a single transistor and a single capacitor as an alternative to SRAM which is built with a six-transistor cell. This Memory Guy blog post was written to commemorate that milestone.
Since the goal of a semiconductor is to squeeze as many functions as possible onto a single silicon chip, the new memory technology was an enormous help to move Moore’s Law along.
The patent, number 3,387,286, carries the name of Robert Dennard, the father of the DRAM chip. Since the name “DRAM” didn’t yet exist, the title of the patent is: Field Effect Transistor Memory.
(At that time the US Patent Office was using a different symbol than we now use for FETs, with boxes symbolizing the channel and a plate alongside it, as we use today, for the gate.)
The industry immediately understood the benefit of storing data on a capacitor to reduce the transistor count. Intel was the first to commercialize IBM’s invention, introducing a 1k-bit DRAM chip called the 1103 in 1970, but the Intel design took a more cautious approach, using a three-transistor—one capacitor (3T1C) cell that did not drain the charge off the storage capacitor as much as Dennard’s original design. Still, Intel’s 3T1C DRAM was far more space efficient than the 6T SRAM it was designed to compete against, and this helped drive down the cost of memory which allowed DRAM to displace SRAM in large memory applications where the additional cost of the refresh mechanism was more than offset by the memory’s cost savings.
Computer designers had already rapidly accepted semiconductor memories, since SRAM chips dramatically cut the cost of a computer’s memory below what was possible with its predecessor technology, the magnetic core memory. Cores had to be hand-wired, one bit at a time, a labor-intensive chore that got in the way of cost reductions. With the introduction of DRAM, the cost savings got even better.
Dennard worked at IBM for his entire career. The company put together a nice page documenting his work that includes a 4-minute video that is well worth watching. Another IBM page about DRAM technology also prominently features Dr. Dennard.
He was a prolific inventor, publishing numerous other patents over the course of his career, but is best known for his 1974 publication of a seminal paper that explained a phenomenon now called Dennard Scaling, in which he accurately predicted that the speed of a transistor would continually increase as its dimensions shrank. This phenomenon held true until around 2005, when increasing leakage currents, also caused by scaling, grew high enough to limit any further speed increases.
Here’s a list of some of the more noteworthy honors bestowed on Dennard for his semiconductor research:
-
-
-
- 1979—Declared an IBM Fellow
- 1988—Received the US National Medal of Technology from President Reagan
- 1990—Was given the Harvey Prize in Science & Technology
- 1997—Inducted into the US National Inventors Hall of Fame
- 2001—Received the IEEE Edison Medal
- 2007—Made a Laureate of the Franklin Institute
- 2009—Received the IEEE Medal of Honor as well as the Charles Stark Draper Prize
- 2013—Given the Kyoto Prize in Advanced Technology for his discovery of Dennard Scaling
- 2019—Received the Robert N. Noyce Award from the Semiconductor Industry Association
-
-
Bob Dennard was known to be quite an optimist. When asked if he expected for semiconductor scaling to end, he replied: “There’s an end to scaling. But there’s no end to creativity.”
Dennard passed away last year, on April 23, 2024, at the age of 91.