10

Early pioneers building memory cards for personal computers, tended to use static RAM, because it's quite a bit easier to get to work. Later, dynamic RAM became de rigueur, for the simple and sufficient reason that it costs a lot less per kilobyte; as far as I know, the Vic-20 was the last computer to use static RAM for its main memory.

As far as I can tell, the difference in price per kilobyte became larger over time, as density increased. The ads in the back of Byte magazine are a good way to check memory prices... at least they become so from the late seventies.

But looking at 1975-76, I'm surprised I can't find any ads for DRAM chips at all. I wonder is that more because the cost advantage was not yet that great, or because it was that great but the hardware hacking community was not yet ready to deal with them, no matter what the advantage?

So: in, say, 1975, when the personal computer industry was just getting off the ground with the Altair and the Homebrew Computer Club: what was the cost differential between DRAM and SRAM?

Improve this question
5
  • 4
    the Vic-20 was the last computer to use static RAM for its main memory - NOT AT ALL. Perhaps "the last mass-market computer". Certainly plenty of military and scientific (e.g., space travel) computers continued to use SRAM, and I suspect plenty of supercomputers in order to not waste any cycles on refresh.
    – manassehkatz-Moving 2 Codidact
    Jan 7, 2020 at 23:53
  • 4
    @manassehkatz-ReinstateMonica The Cray Y-MP series didn't replace SRAM with DRAM until the M90 model in 1992. The Cray 1 used 16x4bit 6ns SRAM chips for register storage, and 1024x1bit 48ns chips for main memory.
    – alephzero
    Jan 8, 2020 at 0:15
  • 4
    ... the last computer ... Most embedded systems, which are also computers, use static RAM as main (writable) memory. Their number is bigger than the number of personal computers, since long and to these days.
    – the busybee
    Jan 8, 2020 at 7:06
  • 5
    @manassehkatz-ReinstateMonica you should flesh this out into a full answer because it addresses the technological reasons for not making the switch initially, not just financial ones. Even though the question was 'answered' quickly. It is often that a more nuanced answer will get voted up over time. For instance along with your comment, did implementation of sram required less system board component resources than dynamic ram, so while sram was per bit higher cost, did sram have a lower BOM price for the whole solution?
    – Rowan Hawkins
    Jan 8, 2020 at 7:07
  • 3
    @thebusybee - it also depends on the application. Fast SRAM is the preferred solution (including now) for many embedded video products and other applications where refresh (and for SDRAM the transaction based operation) gets in the way of determinism (although they can get really toasty when in use).
    – Peter Smith
    Jan 8, 2020 at 11:11

3 Answers 3

Reset to default
11

I did find some prices in BYTE:

November 1975, page 91

2107 4Kx1 Dynamic: $19.95 (0.49 cents/byte)
2111 256x4 Static: -- not listed
1101 256x1 Static: $2.25 (0.89 cents/byte)

April 1976, page 89

2107 4Kx1 Dynamic: $19.95 (0.49 cents/byte)
2111 256x4 Static: $7.95 (0.77 cents/byte)
1101 256x1 Static: $2.25 (0.89 cents/byte)

Byte May 1977 - Page 170

2107 4Kx1 Dynamic: $4.95 (0.12 cents/byte)
2111 256x4 Static: $6.95 (0.68 cents/byte)
1101 256x1 Static: $1.49 (0.58 cents/byte)

So it looks like even in 1975 dynamic ram was cheaper per bit.

Improve this answer
2
  • 1
    Cheaper and bigger. Sometimes size does matter :)
    – PTwr
    Jan 8, 2020 at 9:06
  • 1
    According to MeasuringWorth.com $20 in 1977 is equivalent to (2018 $): - real price of that commodity is $82.70; real value in consumption of that commodity is $94.20; labor value of that commodity is $81.10 (using the unskilled wage) or $89.50 (using production worker compensation); income value of that commodity is $133.00; economic share of that commodity is $197.00;
    – D Duck
    Jan 8, 2020 at 9:12
8

My firm designed microcomputer boards in the late 1970s to early 1980s, and we often had discussions about whether a particular design was going to use static or dynamic RAM.

When you say "static RAM, because it's quite a bit easier to get to work", you also need to remember that the refresh circuits cost design time, chips, and board space. (No surface-mount: a pull-up resistor was 0.1 by 0.2 inch at minimum). You needed to have a certain amount of RAM before any extra circuitry is worth the savings on the memory itself. We didn't have the gate arrays and similar circuits of today, and the refresh circuits were tricky if they weren't to hold up the bus for the CPU periodically.

The Z80 was as far as I remember unique in having built-in support for dynamic RAM, and so you will see DRAM much more commonly than with other CPUs, but you still had to multiplex the address bus. As the Z80 had a separate IO space you didn't have to chop up the address space so much (though still needed for ROM).

One super-economic design we made was a Motorola 6809 board which had 64 Kbyte dynamic RAM, with about 8 Kbyte ROM mapped in on A15 for a short period after reset. We used software refresh: if the memory rows are the low-order address lines, you can refresh all of memory by accessing enough bytes in a row to hit all the rows. We put a timer on the non-maskable interrupt, and the NMI response was a block of 127 No-ops and a Return-From-Interrupt. (In fact it was a block of PAGE2 markers, which were only 1 cycle each, rather than 2-cycle NOPs: as an undocumented feature multiple PAGE2 opcodes behaved like a single one.) Brutal but small and cheap way to get lots of memory.

Improve this answer
1
  • A very helpful aspect of static RAM from a "getting it to work" standpoint is that systems using static RAM can be "paused" and single-stepped in ways that would fail with dynamic RAM. Even though something like a 6502 has a minimum clock speed of 100kHz, operation can be suspended for an arbitrarily long time long with /RDY, and the bus will remain static while /RDY is held.
    – supercat
    Mar 14, 2020 at 19:21
5

The Vic20 was not the last consumer product to use static ram on the main system board.

In the mid 90's a Socket 3 (486 class) motherboard was created by Ocean Technology octek.com - defunct. The HIPPO-DCA2 motherboard which required at least one 4MB 72-pin SIMM of something called DynamiCache RAM in the first 2 slots.

DynamiCache was a built from high density Static ram and was slated for 4/8/16 MB. However the slots on the DCA2 were too close together for the heat spreaders eventually needed on the 16MB module so the max you could use was two 8MB SIMMs.

At the time, the fastest DRAM available was 60ns FPM. DynamiCache operated at 15ns. As a comparison when DDR eventually reached consumers, it operated at 8ns.

Improve this answer

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .