Ah. In that case color me whooshed.
It seems like a computer would use a significantly higher number of tubes than audio equipment does.
Oh, it would. ENIAC, for example, had nearly 18,000 of them. You’d certainly need the guts from a LOT of amps–the exact figure is a metric buttload and a half.
More like 10[sup]9[/sup] metric buttloads.
The laptop I’m typing this on has a gig of dynamic RAM. That’s 8 x (1024[sup]3[/sup]) bits. Let’s assume that each bit requires one capacitor, plus other componentry. Bits are assembled in banks of 1024 bits, and each bank requires, say, 50 tubes for its support circuitry (read, write, and refresh). That means the memory would need 50 x 1024[sup]2[/sup] tubes. If each tube consumed 10 watts, that’s half a gigawatt right there. And then there’s the video memory and the processor…
To replicate the construction of a moderm computer would probabvly require billions of tubes consuming tens of billions of watts of electricity.
My computer runs off an 85-watt power supply. I think we’ve done well. 
Nope. One of the problems NASA ran into in doing refurbishment on the Shuttles is that the AP-101B computers used ferrite (magnetic) core memory for semi-volatile storage which was no longer available. IBM essentially ended up back engineering the semi-conductor memory AP-101S from the System 4/PI (from which it had originally descended) used on the F-15 Eagle and E-3 Sentry. This allowed NASA to continue to use reliable legacy flight control code and existing avionics hardware without modification (and the accompanying qualification and validation of new code) while expanding the “core” memory capacity, allowing somewhat more complex avionics code. (The Shuttle still has to load specific flight programs from the Mass Memory Unit, as its registers are insufficient to store multiple programs at once.)
Of course, you could make your own; the original ferrite cores were little more than small ferromagnetic rings threaded on a weave of insulated copper wires; many of these were hand-manufactured in China. The bigger question is where you’re going to get a suitable volume of functioning vacuum tubes. If you’re going to go this route, you definitely need to use [url=http://en.wikipedia.org/wiki/Nixie_tube)Nixie tubes for display.
Stranger
I don’t think so. In any case, such “can be confused as beinng serious and are intentionally false” postings are very rude.
We had a Univac 1000 in my college. It was the size of a large desk, it could do less than what a decent scientific calculator can do today.
I’ll just toss in that if you had some sort of unobtanium that allowed you to build something as big as you could ever want without worrying that it would collapse, you could build a full computer just using gravity and marbles that roll around and drop through holes (and then probably a lift to bring the marbles back to the top.)
You could make a pentium with just Renaissance technology if you made it big enough.
Not the “original original”, the one you replied to about old things being better than new ones…
on edit, sorry, that was QED
Of course, it might be a bit slower than the production model…
You could probably build it with Renaissance tech, but I doubt you could power it.
Yeah, I meant in terms of replicating the logic gates to handle addition, subtraction, etc.
Wow I didn’t mean to “whoosh” anyone nor have some folks get angry at Q.E.D. I was merely parodying the guitar player’s liturgy of ‘you can’t possibly get any guitar or amplifier today that will sound like the old ones.’ Actually, I recently bought an electric guitar for $100 and it is pretty damned good. (Okay it was made in China[sup]1[/sup]. Maybe 50 years from now people will say ‘those new Chinese guitars aren’t as good as the old ones because the lead paint in their finish gave it that characteristic tone that you can’t possibly get today.’)
Yes, as Q.E.D. pointed out, the majority of these 2 postings belong in that other thread.
Oh and for those trying to calculate the energy required for an all tube computer, don’t forget the cooling system would require a huge amount of energy.
We now rejoin the original topic already in progress. 
[sup]1[/sup] True story. $100 electric guitar made in China and it is great!!!
As my Romanian control theory professor was fond of saying: “You could have a slave to turn it.” 
I’m not going to dig through all the descriptions to find the specific volume, but one of the books listed here does tell you how to make your own vacuum tubes.
Vacuum tubes will drive printers and monitors just as well as ICs or transistors. Long term storage would be on magnetic tape or punch cards, just like it used to be.
When I first went to work the computer here occupied a room with several equipment cabinets. It was programmed in machine language. A pocket calculator now has as much computing capacity as that one did, lacking only in long term data storage and I/O drivers.
I asked my brother this, he said EACH little magnetic core doenut = a vacuum tube! So a memory board of 4065 doenuts, or “bits” would take 4065 tubes! Not much but a visual steampunk wall display, unless you buy a warehouse and build a BIG ONE! It would be fun to have the old SAGE computer in my back yard for tourists!
I don’t think this is true.
Each donut is a bit of memory. If you needed one tube for each bit, you could get rid of the core entirely. The reason core was used is because it was the cheapest, fast memory at the time.
The tubes are used for logic functions and registers.
Magnetic core memories had separate wires for each row and column, with each core being accessed only when the corresponding row and column were on. So assuming you needed one tube for each row and core wire, the number of tubes would grow with the square root of the number of cores. A 4096 bit core memory would only require 128 row and column address tubes, plus a few more for handling reads and writes. Nowhere near as bad as needing one tube for each bit.
The most common modern small signal tubes are dual triodes. With two triodes you can make a flip-flop - which is the equivalent of a single bit of static memory. You could arguably build a dynamic memory with a single triode per bit. This uses triodes as equivalents to the transistors in a modern IC. (Which is not all that unreasonable.)
Back when tube computers were built there were specialised tubes built that were much more interesting than simple triodes. You can build a tube that intrinsically holds more than a single bit value, and does so so long as it remains powered, requiring no external refresh (unlike dynamic memory). Tubes that held a single decimal value were easy. A Williams tube was the ultimate version.
Like modern systems, it matters how fast your memory is. You will want registers to be very fast, and you may put up with slower memory for less critical needs. One problem with delay line memory is that you have to wait, on average, half the delay line period to get your value out. Programs were very carefully optimised so that their cycle times were just right, and the next operand needed was just coming around in the delay line as it was needed.