In the history of computers, much is said of microprocessors, for good reason, but the relatively unsung RAM chips were equally important. Of particular significance were the 4kbit and 16kbit dynamic RAMs, which had capacity sufficient to enable a single-board personal computer to do nontrivial things; the Apple II, for example, could use either kind of chip, in up to three rows, for a range of 4Kbyte (affordable) to 48Kbyte (capable of running spreadsheets and word processors).

At what process node were 4k and 16k DRAMs first made? For example, the 6um process saw the introduction of the 8080 and 6800 microprocessors, but there is no mention on that page of which RAM chips it covered.

If it makes a difference, by 'first made' I mean commercially significant volume production, not just prototypes.

  • 6
    Interesting question: you'd think the answer would show up within 2 minutes googling.... but nope. – another-dave Dec 25 '20 at 0:35
  • 3
    Ken Shiriff has a blog entry about reversing a decapped 16K DRAM, so I guess you could ask him about the measurements going with those pics, and from there derive the process node. – dirkt Dec 25 '20 at 6:28
  • 6
    @another-dave: difficulty in finding process node information for chips this old stems primarily from one fairly simple fact: the idea of a process node didn't come into most people's thinking until there were a few generations to start showing a trend. In addition, at the time density depended quite a bit on circuit topology (NMOS vs PMOS vs.HMOS, etc.) Process technology only became all-important after everybody settled on CMOS for nearly everything (which was later than the chips being discussed here). – Jerry Coffin Dec 25 '20 at 7:16

Intel manufactured its 1-kilobit 1103 RAM on an 8 μm P-MOS process.

Through most of the 1970s, DRAM was made from NMOS. The first successful CMOS memory was the Hitachi HM6147 SRAM, a 4-kilobit chip which used a 3 μm CMOS process. This was not the first 4-kb chip on the market (it was preceded by, among others, the Intel 2114 SRAM), but it’s more comparable to the CMOS DRAMs that followed it.

The first 16K RAM chip was the Intel 2116, circa 1976, and the market leader was the MOSTEK 4116, introduced circa 1977. Unfortunately, I don’t have a reference for their feature size at hand, but NMOS memories of the sizes you’re asking about would have used a process somewhere in the neighborhood of 6 micron, between the 8 micron process of the 1K RAM and the 5 micron process of the 64K RAM. There were other innovations in process technology beyond shrinking the feature size, particularly in the 4116.

MOSTEK’s 1979 databook said,

At MOSTEK, all new products are being designed to work on either current process technology (5 microns) or new generation SCALED POLY 5. To accomplish this goal all new products are designed to operate on a single 5 volt power supply. All products use advanced state-of-the-art design techniques and perform very respectably on the standard process.

MOSTEK was then selling 64-kibibit DRAM, the MK4164, apparently on an advanced 5-micron process.

I found a claim that NEC introduced 64-kilobit RAM on a 1.5 μm process in 1981, but not a reliable reference for it so far.

From that point onward, when an apples-to-apples comparison becomes possible, this site is an excellent resource. It describes the state of the industry in 1983 as:

256Kb development is at a maturing stage. NMOS 256Kb by Fujitsu with 2.5um NMOS triple poly process, 256Kb by NEC on 1.3um double poly process, 256Kb by Mitsubishi on 2um process, 256Kb by Motorola, 256Kb by Toshiba with 2um double poly process. Intel shows off industry-first CMOS DRAM with 64Kb on 1.2um process.

There was a leap to 1 megabit DRAM in 1984, using a 1 μm CMOS process.

4-megabit DRAM appeared in 1987, using an 800 nm process.

16-megabit DRAM appeared experimentally in 1988, and became mainstream in 1989 on a 500–600 nm process.

64-megabit DRAM became mainstream in 1991 on a 400 nm process.

  • I don't get what you mean by the third paragraph ("Although this refers to a ROM..."); could you clarify? – Alex Hajnal Dec 25 '20 at 10:30
  • @AlexHajnal Ah! I see now that that paragraph was left stranded when I moved the paragraph that was supposed to precede it. Removed now. – Davislor Dec 25 '20 at 10:34

My recollection was that 4 kbit and 16 kbit DRAMS were being manufactured in volume around roughly circa same time frame as when 8 micron NMOS fabrication was common for other stuff. This web site seems to support that hypothesis:


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.