1950's style Modern Computer

What would be the characteristics of a modern computuer “powereed” by 1950’s style vacuum tubes. (ie just before transistors)?
UNIVAC be damned!. Take modern computer “archetecture” and apply it to late 1950’s tech. How big, power hungry, fast/slow would a typical 2007 computer be if it utilized vacuum tube tech?

I think it would be at least the size of a small city ( 5000 homes worth of power) and as fast as a toaster… per megaflop
your ideas/responses?

FML

Lots of heat radiators designed to look like rocket ship fins to show how fast it is.

indeed… 64 megawatts at least… grin

fml

The logistics of handling the error correction would squish any effective output immediately.

In other words, just idling, the fifty-five million tubes that would emulate a P4 could not even keep the cycles free of error to enough correct their own (mal)functions. The failure rate of vacuum tubes is a sick one compared to logic gates made of silicon.

yeah, my first thought was that you’d never keep all the vacuum tubes functioning long enough at the same time to get anything done, given as many as you’d need to duplicate, say, a 2.8 GHZ chip

We discussed this sort of thing not that long ago. Look at SAGE, the computer that ran US defenses in the 1950s. All vacuum tubes, and completely redundant (so they could fix one half while the other conti8nued to run. Heat fins alone wouldn’t cool it – you had big open racks of tubes filling entire rooms. Until a decade and a half ao, they had a section of this up that you could walk through t the now-defunct Boston Computer Museum.

http://en.wikipedia.org/wiki/Semi_Automatic_Ground_Environment

We also have to consider memory and disk.

Ferrite core memory of 1G would take up an area of about 3 acres if on a single plane.

The IBM 305 RAMAC 5M disk drive was about as big as a refrigerator. So for a modest 300G, you would need 60,000 of those puppies, taking up, let’s say (assuming 3’x3’x5’), about 30,000,000 cubic feet. If we assume a footprint of 3x3 then that’s about 12 acres of floor space.

The first computer that I wrote a program for, in 1963 as an undergraduate student, was already old-fashioned at that time. It was a valve-based computer called Utecom at the University of New South Wales (“Utecom” was short for “University of Technology Computer”, and University of Technology was the former name of UNSW.

Utecom occupied a very large room, and broke down regularly as one of its valves failed. You changed the “operating system” (for want of a better word) by removing and replaced boards wired up different ways. The programming language was called George, and was a reverse-Polish language very similar to that used on HP calculators. In fact Utecom was very similar to, and about as powerful, as the sort of scientific calculator that you buy for about $20: all you’d need to add to that would be a punched-paper-tape reader and writer, and you’d have the whole capability with today’s technology.

I imagine that to emulate a typical modern-day laptop you’d need to fill every building on a large university campus with the processing units. They’d also be the fastest way to emulate RAM, and I don’t know what you could use for a hard drive except more valves: anything else would be intolerably slow. You’d need to build in redundancy, since otherwise the system would fail each time a valve failed, and that would add more valves and some sophistication into the operating system.

“I am sick of all this motherfucking ferrite on this motherfucking plane!”

Looks like there’s another literary genre in the making here: valvepunk, based on the premise that semiconductors don’t exist. :cool:

The IBM Mark I is on display at the Harvard Science Center.

Babbage-Punk: All analog, all the time.

If you want to see the mother of all slide rules click here and here for pics.

It would be horribly slow. Ignoring the fact that it would always be broken, due to component failures, the propagation delays between components would kill its performance. A typical cycle time for a modern processor is well under a nanosecond. That limits the physical dimensions of the processor to a few inches. Scale the processor size up to the size of a large warehouse, and you will have to scale down the system clock frequency a similar amount.