8

I'm trying to get a feel for what it looked like when you designed a computer to have a lot of memory chips stuffed into it. To that end, I found this: https://commons.wikimedia.org/wiki/File:VAX_11_780_16mb_memory_board.jpg

If I'm counting correctly, that's 256 RAM chips on the visible side of the board. Assuming the use of 256kbit RAM chips, 256 x 256 x 1024 / 8 = 8Mbytes on the visible side, so if it's double-sided, that would be correct.

What are the physical dimensions of the board? Height, width, and depth in terms of how far apart they are spaced in the machine?

12
  • 1
    There are more than 256 chips. 8 submodules, 40 chips per submodule so total 320 memory chips of 256k bits.
    – Justme
    Oct 29, 2020 at 19:47
  • 1
    @another-dave I don't know what that board is and how much of the memory is used. All I see from the photo is that the memory chips are TMS 4256FML. Datasheet says they are 256 kilobit chips.
    – Justme
    Oct 29, 2020 at 20:03
  • 1
    @justme - I found a print set for a board using 256K RAMs, but it nets out at 4MB per board, 156 chips (see my answer)
    – dave
    Oct 29, 2020 at 22:49
  • 2
    I posted a link to this question in a Facebook group named Digital Equipment Corporation Alumni. Some field service engineers who maintained the 780 will probably see it. A few of them probably have one of these boards in their garage (yuk yuk). Oct 30, 2020 at 10:11
  • 1
    I've had no luck finding mention of 780 boards of more than 4MB. I wonder if it could be a 3rd-party board? Apart from that, maybe it's really for Venus (the 8600). Venus was an SBI machine like the 780.
    – dave
    Oct 31, 2020 at 1:44

4 Answers 4

8

In answering this question, I found that there were multiple generations of 11/780 memory board, and so sizes, capacity, etc., are quite hard to pin down. I still have not found a board larger than 4MB.

I did get an answer for physical board size, though not for slot width (board separation).


One MS780 memory controller can take 16 memory array boards. For the original release of the 11/780, the boards were the M8210 or M8211 (32Kbyte/8Kbyte capacity, respectively), leading to a maximum of 512 KB per controller.

The system supported up to 4 controllers, thus a max of 2MB.

So, for the 'get a feel' part, that's up to 64 memory boards. If we assume the boards are spaced at least 1" apart (I could not find slot-size info), that's over 5 feet of memory alone.

ECC was used, 8 check bits per 64 data bits.

The board size us 12.5 x 15.5 x 0.75 in; I assume the 'depth' includes the tallest chips.

The board numbers came from the MS780 memory system tech manual, see figure 1-1 on page 1-3.

Here's a link to an M8211 catalog entry in the Computer History Museum. When they say '64K module' they apparently mean 'bits'.


Later memory boards increased the board capacity and thus the system capacity. My 1980 "VAX Hardware Handbook" (includes the 780 and 750) has the following:

  • Per chip 16K bits
  • Per board 256KB
  • Per controller 4MB
  • Per system 16MB

The physical board size won't have changed for these though.


This 1984 print set for the MS780H memory controller talks about 256Kbit MOS devices, which squares with the OP's chip identification.

The memory board appears to be an M8374, a 4MB board (which nets out to 144 chips including 1 check bit per byte; the print set say there are 156 RAM chips).

The system might only support 2 controllers at this point though. It's not clear to me what is authoritative.

4
  • Great link, but ... err ... didn't rwallace ask for dimensions? <DuckAndRun />
    – Raffzahn
    Oct 29, 2020 at 18:55
  • 3
    @Raffzahn - Last time I studied physics, 12.5 x 15.5 x 0.75 in were considered to be dimensions. Though I confess that we didn't measure things in "inches" even back when I was in school. For the board sizes, I did at least convert the fractions to decimal values.
    – dave
    Oct 29, 2020 at 19:37
  • 2
    I did know that you might be over 30, but didn't expect you to be that old, still being trained in medieval units:)))
    – Raffzahn
    Oct 29, 2020 at 19:44
  • 2
    I grew up using imperial units in the real world, SI units in school, but now I'm living in a medieval country, and the ancient scrolls for that memory board use inches.
    – dave
    Oct 29, 2020 at 19:45
5

This is a VAX module but likely for an 11/780 successor, not an 11/780. The edge connector with approximately 150 signals per side and the ECL flip flops with 1987 date codes suggest that we are examining an 8 or 16 megabyte module for a VAX 8000 series or the like. 8 megabytes + parity is visible. Underneath is probably DRAM refresh, selects, and probably MOS<->ECL level shifting.

The 11/780 was not an ECL-based CPU and I can't imagine why we would see Motorola ECL 10k chips on an 11/780 board.

In any case, VAX boards were about 15" on the long dimension, which was called height because the modules were mounted vertically. Plenty of photos are available if you search for VAX backplane.

1
  • 2
    I believe you are correct. The picture from the OP shows a board with 3 edge-connector fingers. 11/780 memory plugged into the SBI (Synchronous Backplane Interconnect) which had 6 smaller fingers. This picture of a VAX-11 memory board at the CHM agrees with '6 fingers'. (VAX-11 was used for the /7xx machines, not later systems). So: does the OP want to know about 11/780 memory or about the memory pictured in the wikioedia lin?
    – dave
    Feb 11 at 16:23
3

To give you a sense of the inter-board spacing in these types of systems, it is useful to consider the main constraints. These include airflow for cooling, the physical space between the connectors on the backplane and component height.

  • The gap for airflow purposes was such that most folk could slide their fingertips between boards, but only at the expense of scratches caused by the sharper bits of the through-hole pins. After component height, this means ~10 mm.

  • The connectors on this generation of system were simple slide in affairs, much like a contemporary PCIe slot. Later generations (e.g. VAX BI backplane) had a much more elaborate cranked/clamped system. Signal integrity along the backplane demands the spacing between the connectors be minimised, but the physical constraints typically limit the pitch to ~20 mm.

  • To get a sense of the size of the components, those two large silver cylindrical capacitors by the legend card are ~0.5in in diameter (say 10-12 mm). Double sided board were rare because of assembly difficulties, but not unheard of.

So, the net result is an inter-board spacing of around 20 mm.

2
  • 2
    The depth of the original M8211 board was given as three quarters of an inch in the tech manual, which is a hair over 19mm. 20mm might be a little tight. OT: I once destroyed a DMR-11 removing it from an 11/24 (never let a programmer inside the machine!) because the tall ROMs got ripped off by wire spikes on the next board.
    – dave
    Oct 31, 2020 at 13:57
  • 1
    Good point, but your story kinda suggests that I was close :-) Oct 31, 2020 at 18:22
3

Other responders have already given the exact size from data sheets, but it's also possible to get a good estimate of the size of a thing from the size of its parts. This is a useful skill when we aren't able to measure the thing or have a spec sheet.

A somewhat visible feature on this board are the edge connectors, but we don't know their pitch. However, the middle edge connector is adjacent to some DIP chips. The edge connector fingers line up with the chip, whose width is three times the distance between fingers, and "skinny DIP" chips such as these are known to be 0.3" wide. So the edge connector pitch is somewhere around 0.1" (2.54mm).

0.1" is a very common standard edge connector width, especially for the era, but are we assuming too much? Edge connectors come in other sizes too. The nearest common size is 2mm, but that would imply the DIP chip is 6mm wide, and they don't come that small. It could be 2.5mm, but that's both unlikely given this was built in the USA and other parts use Imperial measurements so the board is probably designed around a 0.1" grid[0], and it's less than 2% different anyway so lost in the noise of the estimate.

From here, I could count the number of edge connector fingers, but that's a lot of effort and doesn't tell me the size of the gaps between the edge connectors. I measured the pitch of ten fingers (i.e. 1") in that picture as 253 pixels and the width of the card as 3794 pixels, and a quick bit of maths gives me ~15.61". The depth of the card is trickier because of perspective, but guesstimating gives me about 2900 pixels or ~11.93".

Realistically, boards come in more rounded figures, so I'd say it's 15.5" or 16" by 12". How well did I do? According to other replies, this board is 15.5" by 12.5", so not too bad: I apparently underestimated the perspective distortion but got close enough to get a feel for the size of the thing even if I get zero marks were I to have built a case to hold it.

If numbers in inches are still a bit abstract, this is approximately the size of an A3 sheet of paper.

[0] Even though I'm otherwise in a fully-metric country, I still end up using a 0.1" grid when designing boards because I'm mostly using DIP and other through-hole parts.

You must log in to answer this question.

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