A nicely detailed presentation, just an hour long, by David Lovett of YouTube channel Usagi Electric, at VCF South West this year (2026)
These were two relatively simple relatively affordable computers from the days when the drum had a timing track which controlled the CPU. And the drum was all the memory you had.
When comparisons between the LGP-30 or -21 and the Bendix G-15 come up I wonder how they both compare and relate to the IBM 650. AFAICT from just looking at Wikipedia the 650 also had an all-drum main memory (at least in the initial and base model); it was released a few years earlier; it seems to have been maybe 2-3 times as expensive to buy, but “only” 2-3 times, and IIRC IBM customers tended to lease instead; and it sold significantly better, about 2000 units in total (all between 1953 and 1962) compared to about 500-600 LGP-21s and -30s combined and 400 Bendix G-15s. The typical users and applications don’t seem to have been completely different, either: Knuth on his 650 isn’t that different from Bob Hargraves on the LGP-30 in the basement at Dartmouth[1]. Were the LGP-30 and G-15 intended to be something like a lower-cost competitor to the 650? Or was there just some underlying reason why relatively(!)-mass-market drum-memory computers became viable in the mid-'60s? (Something to do with the floating-head research in the mid-'50s maybe?)
Of whom, more anon. ↩︎
Mmm, half the cost would make them attractive… not being IBM might limit sales. Also, IBM’s tactic of leasing might be much easier for many companies to justify.
I do wonder though if these machines might be better, in some way, for scientific computing? There’s this comment in your Smithsonian link:
Bendix had obtained the rights to the original US digital differential analyzer known as the MADDIDA; the G-15 was the company’s attempt (especially in the Model 15D) to incorporate the special-purpose features of the MADDIDA into a general purpose computer.
I can’t immediately see what the 15D improvement was.
I note that the LGP-21 was cheaper but also less performant (Bell says “by about a factor of 3.”) There’s a switch to transistors, a big decrease in electrical power, and a change from drum to disk - still fixed-head.
Edit: an interesting article here about using a G-15 for numerical control of the machining needed for aircraft manufacture, in 1958.
Edit: nearby, the G-15 marketing materials, aka Technical Bulletin:
Programming systems are available for either fixed point or floating point decimal input and output.
Double-length arithmetic registers are included to permit the programming of double-precision operations with the same ease as single-precision ones.
…
Computation time is reduced by incorporating arbitrary-precision multiplication and division in the design; either operation may be terminated as soon as the number of digits required for the desired accuracy is obtained.
Edit: and there’s an accessory for differential analyser usage:
The DA-1 provides the unusually large number of 108 integrators and 108 constant multipliers; hence, it may be used to obtain numerical solutions of complex problems involving high-order, differential equations or sets of simultaneous equations. Facilities are provided for the input of empirical functions. The set of 108 integrators is processed 34 times per second. Seven decimal digit precision is available. Output may be tabulated or plotted graphically by G-15 equipment. The accessory is the size of a standard filing cabinet and weighs 300 lbs. Power consumption is one kva.
Sadly, the person I knew who had used a 650 and many other early scientific computers passed away over the weekend. J. D. “Dan” Larson (1935–2026) got his PhD in experimental high-energy physics at Caltech and went on to develop particle accelerators at Oak Ridge, Argonne and other labs. He recalled using the Wolontis-Bell Floating-Decimal Interpretive System on the 650 some time in the late 1950s. I tried to interview him about his computer usage, but he wasn’t as interested as I was. I miss him.
Yes, that’s basically my thinking as well. I think—and this is just speculation right now—there may have been a bit of a foreshadowing of 1981, with IBM making the market and then others trying to figure out how to get a piece of it. I imagine that for people like Stan Frankel and Harry Huskey the 650’s bi-quinary arithmetic looked like ridiculous IBM flab begging to be trimmed off, while some potential customers (business users especially) were actually quite reassured by the promise of unproblematic decimal arithmetic. Hence maybe the G-15 materials trying to reassure prospects that no problem, we can handle decimal using software.
It’s a bit reminiscent of Bell’s own PDP-4, a “less for less” cut-price version of the PDP-1 which didn’t set the market alight either[1]. My feeling is that while looking at the history of small computers in the 1960s is often treated as a quest for the first personal computer, that frame tends to obscure the fact that in the '60s there were quite a few people who were keen to sell you a small computer built with single-user interactive sessions in mind (here’s another one…), but few people willing to buy. And while no doubt that was partly because some people didn’t like or didn’t get the idea of interactive computing, I think it was mainly because the price-performance tradeoffs made single-user machines look barely capable and/or wildly expensive. This was still very much the era of Grosch’s Law.
I’m sorry to hear that! Dr. Larson seems to have been an interesting friend indeed.
