• Please review our updated Terms and Rules here

Tube computer size: ENIAC vs Bendix G15

C

carangil

Experienced Member
Joined
Jun 3, 2009
Messages
285
Location
Oakland, CA
My dad mentioned to me the first computer he used was a Bendix G15. I looked it up, and was surprised it was vacuum tube based, given its small size. Huge by today's standards, but when I think about vacuum tube computers, I think of building sized computers like ENIAC. the ENIAC is usually regarded and the first tube based computer. Why was it so huge and complicated? Why not start with a simpler design? Were they going all-out to try to make a megacomputer, or were they really incapable of coming up with a simpler or smaller design?
 
carangil said:
Why was it so huge
Click to expand...

ENIAC was built using very large components, octal-base tubes, for example.

The MIT Whirlwind was a multi-story building, using similar packaging density, built a few years later but much more complicated.
It was actually a stored-program computer.

The G15 used much smaller tubes, semiconductor diodes, and was a much simpler design (serial instead of parallel arithmetic, for example).
 
We have a Bendix on display at the VCF museum in New jersey. It is very amazing how small it is compared to the Univac across the isle or the Cray a few feet away.
 
Serial operation and lots of diodes. Drum memory also keeps the need for active circuitry down.

Look at the Litton 1601 for an example of a minimalist 1970s serial drum-based computer. Since it used commodity TTL and not much of it, it would be an excellent candidate for a retro PC. Earlier versions (made by Monroe) used discrete components and still were small. The PB250 likewise used very few active devices and used about 1500 Watts and could be placed on a table. Basically, you only need active devices when you need to implement the NOT function.

Too bad, the old stuff doesn't interest many VC people nowdays. History appears to start with the microprocessor.
 
Hi All;

"" Why was it so huge and complicated? ""

There are many reasons, some of which were the large tube Size.. But, it was also this was the first Electronic Digital Computer and they didn't know what We know Now..
They had to work out not only what it would do, but how to accomplish getting that done and it took time to build..
There was no Stored Program, type of Architecture, and the Military kept changing what they wanted it to be able to do..

There is a Fairly New Book out On the ENIAC, which I have.. It's Title is --
"' ENIAC in Action "" Making and Remaking the Modern Computer
By Thomas Haigh, Mark Priestly and Crispin Rope..
MIT Press 2016..

And here is the Video to go along with the Book..
https://www.youtube.com/watch?v=3Z8XrC3kLJM

THANK YOU Marty
 
Subminiature (T3 outline) tubes have been around since the 1940s. When I was much younger, I used to buy surplus computer boards with 6 RCA 6021 subminiatures (or equivalent), IIRC mostly connected as flip-flops. They were positioned on either side of the PCB, flanked by an aluminum plate with clips on either side--three to a side. Lots of mica caps and such on the board--and really, really cheap--maybe less than a dollar apiece. There was a good reason for this--when a board went bad, it wasn't repaired, just junked and replaced with a new one. So you got 6 tubes, 5 of which might be bad.

This was in the heyday of "Surplus Row" on South Michigan Avenue in Chicago--dark, dusty stores just chock-full of military surplus stuff. It was heaven.

Subminiature "hearing aid" tubes (e.g. HY245) are pre-WWII, as were the 7-pin miniature tubes (e.g. 9001, 9002).

So, no excuse for the "bigness" per se.

One of my old bosses worked on SAGE. He said that one of the first troubleshooting techniques was to first walk down the racks, looking for any tube filaments not glowing. He mentioned that said racks were a convenient unofficial way to keep your lunch warm...

There are still subminiature tubes being sold on eBay. They appear to be new. Obviously, a great deal of them were manufactured.

Not as small as nuvistors, of course, but pretty small for the time. I suspect that they weren't employed in the like of the G15 and LGP30 because they were often wired right to the circuit (no socket), so not easy to field-service.
 
The ENIAC is really not that similar to modern architectures. It's 10 digit base 10, and fully parallel processing. Each accumulator (all the units, actually) runs at the same time as all the rest. There were ten accumulators. The best way to think of the ENIAC is to think of it as 10 adding machines with ten operators all talking to each other, with several extra tools they can use. Programming was done by setting switches and plugging patch cables: nothing at all resembling what we normally think of as programming.

I've done quite a bit of ENIAC programming. There are some really good emulators. Each one seems to have some really big bugs though, at least, the last time I used them. It's been at least 5 years.

The trouble with the ENIAC was that it was severely I/O bound. When they converted it to a stored program design, they eliminated the parallel processing capability, greatly reducing the processing speed. But, it didn't actually operate any slower due to the I/O bottleneck.
 
Chuck(G) said:
Subminiature (T3 outline) tubes have been around since the 1940s. When I was much younger, I used to buy surplus computer boards with 6 RCA 6021 subminiatures (or equivalent), IIRC mostly connected as flip-flops. They were positioned on either side of the PCB, flanked by an aluminum plate with clips on either side--three to a side. Lots of mica caps and such on the board--and really, really cheap--maybe less than a dollar apiece. There was a good reason for this--when a board went bad, it wasn't repaired, just junked and replaced with a new one. So you got 6 tubes, 5 of which might be bad.

This was in the heyday of "Surplus Row" on South Michigan Avenue in Chicago--dark, dusty stores just chock-full of military surplus stuff. It was heaven.

Subminiature "hearing aid" tubes (e.g. HY245) are pre-WWII, as were the 7-pin miniature tubes (e.g. 9001, 9002).

So, no excuse for the "bigness" per se.

One of my old bosses worked on SAGE. He said that one of the first troubleshooting techniques was to first walk down the racks, looking for any tube filaments not glowing. He mentioned that said racks were a convenient unofficial way to keep your lunch warm...

There are still subminiature tubes being sold on eBay. They appear to be new. Obviously, a great deal of them were manufactured.

Not as small as nuvistors, of course, but pretty small for the time. I suspect that they weren't employed in the like of the G15 and LGP30 because they were often wired right to the circuit (no socket), so not easy to field-service.
Click to expand...

Those *look* to be about twice the size of the ones I have, which are about 5/16" diameter. But it's hard to tell, they may be the same size. The ones I have are Soviet manufacture. Unfortunately, I haven't seen them for twenty-five years; I may not actually own them anymore. I never knew the specs of them, but I did build a reasonable audio amplifier with one, which was a triode, using only 12V at the plate.
 
The first miniature tubes that I encountered were as a young tech in the Navy, late 50' and early 60's. The AN/ARN-21 TACAN (Tactical Air Navigation), manufactured by the Collins Radio Corporation, had state of the art interchangable "books" which could be easily removed and replaced in the chassis. These books consisted of various miniature tubes and you needed to demonstrate a good soldering technique. The ARN-21, with its mini-tubes, remained in military use even after being upgraded to the ARN-158 solid state version during the early 70's. As a note, those tubes were durable and highly reliable as they withstood countless carrier launches and landings.
 
Alright, this makes a lot of sense. I suppose with ENIAC, they built and tested it in pieces. It still seems crazy to build such a huge machine as your 'first computer.' But, now that I think if it, it's probably easier to lay the whole thing out with large, bulky, parallel logic, than it is to figure out how to serialize the operations down to something small like the Bendix. I read some more about how it works, with the long and short serial tracks on a spinning drum. Very fascinating.
 
Now that was more like what I would expect for a first electronic computer. It's not stored program, but neither was ENIAC when it was first used.
 
It's always struck me that Antanasoff would likely have added a third drum whose function was to push the right buttons if he'd had the time and resources.
 
carangil said:
Now that was more like what I would expect for a first electronic computer. It's not stored program, but neither was ENIAC when it was first used.
Click to expand...

But the whole point of the ENIAC was to not be what those earlier machines were. It was fully electronic, not partly mechanical.
 
On Antanasoff's machine, pretty much only the I/O was mechanical. Yes, it used rotating capacitive drums for storage, but that's not a big thing--computers still use rotating magnetic disks.

On most computers of the last 70 years, I/O is still mechanical, be it a punch, reader, tape drive, disk drive, printer, keyboard, or whatnot. ENIAC was no different. The ABC had an all-electronic core, unlike, say, Konrad Zuse's Z3 of about the same time period, which used relays for computation.
 
As to the G-15 being small relative to ENIAC, ENIAC was in essence a research machine. Whirlwind, the worlds first "real-time" computer intended to be part of a feedback control loop was as well, designed as more or less a large breadboard layout so as they discovered problems it would be easy to modify. Although I never saw Whirlwind in operation, I did as a new college student in the mid 1960's spend a little time writing ALGO on a G-15 (attached a photo to prove it :) I also have both a G-15 flip-flop module and a Whirlwind module (register driver type 1 serial number 3) in my collection (I hang the G-15 module from my Christmas tree each year as an ornament along with several PDP-8 R202's and DDP-516 modules :). I recall reading in one of the trade magazines (maybe Datamation or EDN) in the mid 70's celebrating the 20th anniversary of the G-15 and mentioning that it was the worlds first personal computer :) Behind the IBM 650, the G-15 and the LGP-30 shared second place as the most produced vacuum tube computers, all drum machines.
 

Attachments

  • Image44.jpg
    Image44.jpg
    95 KB · Views: 1
  • 100_1189.jpg
    100_1189.jpg
    93.3 KB · Views: 1
You must log in or register to reply here.
Back
Top