..and now we have Windows 8.
Still - it does you good to laugh and it's been an interesting journey :)
After three years of restoration by the National Museum of Computing (TNMOC), the world's oldest functioning digital computer has been successfully rebooted at a ceremony attended by two of its original developers. Harwell Dekatron The Harwell Dekatron fully restored (click to enlarge) The 2.5 ton Harwell Dekatron, later …
..and now we have Windows 8.
Still - it does you good to laugh and it's been an interesting journey :)
Windows 9 will have a DIP-switch and incandescent-lamp user interface.
What, Lego is making computers these days? What a topsy-turvy world we live in!
They have a visual program generator which is actually rather clever, though it hardly prepares kids for years of generating xml.
When debugging a computer was really pulling bugs out of the computer :)
Would be curious to understand what type of processing power this thing has in more modern terms... Are we talking 1980s calculator watch? Is there an El Reg unit for such a thing (picoFLOP/s maybe)?
Good on these folks for saving a piece of history, it really is amazing how far (or is it better said "how small") things have come.
The Decatron wasn't designed for speed, it was designed for reliability. It could calculate about the speed of a good human mathematician with a calculator, but it could do it for upwards of 80 hours at a time.
I read in a newspaper article about this that it's equivalent clock speed is 100Hz, so given the times quoted for various operations, it should be possible to derive the processing power.
Sorry if I made it sound like I was slagging the performance. I was just curious if there was some way to put it in perspective vs. today.
"wasn't designed for speed" does sound a bit funny, it was as fast as possible then and hardly very reliable for a long time.
One could compare it to a T-Ford with great respect for both.
If the article is right (15 seconds for a divide) then it was running at 0.066666 FLOPS, 6.6666 microFLOPS or 6.6666666666666666666666666666667e-14 petaFlops .
> 0.066666 FLOPS
To put this in perspective, the Harwell Dekatron described was built in 1949, so assuming it was running continuously ever since it has 63 years of runtime at that 1/15 flop/s.
And then compare with the 27PFlop/s Titan at Oak Ridge (top of the latest Top 500 list).
A few calculations...
Titan would take about 500nS to perform the same number of operations at the Dekatron in 63 years.
Five hundred NANOseconds.
Even a nVidia Tesla card, running at about 1TFlop/s would only need around 130μS.
Even a computer as slow (by today's standards) as a megaflop would only need a couple of minutes.
Does that conform to Moore's law though?
The correct answer is 0 (zero) FLOPS. The machine is fixed point. All the calculations that involve digits after the decimal point would have had to have dealt with the old fashioned integer way.
Nice to see old hardware being reborn! Well done, those men.
The decatron was a type of neon filled gas discharge tube with 10 or 12 separately connected anodes for a static discharge arranged round a common cathode. Between each of the anodes were intermediate anodes for transferring the discharge, these were connected to two circuits. In a gas discharge tube there can only be one active discharge, and this is stable. From any of the 12 anodes the discharge could be transferred clockwise or anticlockwise by applying overlapping pulses in turn to the two circuits supplying the intermediate anodes.
The tubes plugged into valveholders, each requiring connectors for the number of anodes plus three to be soldered.
Brilliant. They counted, you could read whichever anode was conducting, and of course this could be used for a carry, and you could see the content of every tube.
So the single digit in this machine was almost certainly decimal.
I used this in the first electrostatic copying machine built in the UK. It used a continuous length of paper on which the image was projected from an original lit by an electronic flash. The paper was taken through the process to be cut to sheets and delivered to the output tray. The problem was that at the time standard paper came in two lengths, 10 and 13 inches. We wanted to be able to adjust the paper size while running. The machine gave a pulse for every inch advanced, a chain of pulses marking the cut positions passed through the valves. We used 10 output decatrons, when set to longer sizes a spurious signal advanced up to three anodes and then received a reset from the previous tube.
Brilliant, thanks, that was my first thought... One *digit* per valve? Thanks for the clarification.
BZs all round & jolly well done!
Eniac claims to be the first digital computer. But then as it was designed to calculate artillery firing tables, any surviving WWII mechanical artillery, bomber or torpedo calculator, should qualify.
ENIAC is dead, Jim, although some bits are on display in a few places. WITCH, on the other hand, is clattering, flashing and doing sums.
As I understand it, Eniac was a plugboard calculator. Our American friends do not like it that we got there first . And, now that the ARM architecture is taking over the world, I suspect that some of them don't like it that we have also got there last.
Or as the Americans put it,
Go Cambridge! (And Manchester, and PO Research, obviously)
If only Manchester had San Francisco weather, history would have been very different. But it is said that at the original Royal Society lecture on the Manchester machine, looking at the printouts (on which zeroes were represented by / because they could not afford to have made a new ball for the teletype) someone asked if the /es were the rain beating down on the windows in Manchester.
ENIAC was an extremely odd architecture machine - never repeated! It was more like twenty or thirty simple arithmetic machines that could be physically wired so that the '"result" of any machine was the sum/difference/product of any two other machines. Almost like the boxes on a spreadsheet, except that you had to physically wire them together.
Like WITCH, the calculations were in decimal. However, I don't think ENIAC had the concept of "instructions" as such. It could iterate, but couldn't do the equivalent of branch out of a loop when a condition was met: i.e. it wasn't "Turing Complete".
Difficult to claim that ENIAC was a true computer therefore.
ENIAC was programmed with cables and switches. I suspect this machine was too, or maybe with paper tape.
However ENIAC was at least the earliest machine of its type. This one was not, and the Manchester SSEM (Baby) was operational a year earlier with its program stored in RAM. And it was much, much faster. Paper tape programming goes back to the Zuse Z3 during WWII.
This is the oldest original machine that still works, maybe, but it was not a first of any sort when it was new, other than maybe reliability.
ENIAC had an upgrade to make it run the instruction-cycle, but that was after other programmable machines were built.
I hope I can be corrected about this, but it has always looked to me as if the British had no option but to classify computing technology as secret in order to adequately protect it from enemies. Then once the job was done, so to speak, we (along with most of europe) were somewhat pre-occupied rebuilding the nation with a severely depleted and ill supplied population to give much thought to the future implications of computers and were too busy getting our own house in order to pass special legislation to declassify the technology.
Does a certain other nation, relatively unaffected by such issues but whose freedom relied hugely on the recent efforts of other, make allowances for this? Do they give Britain time to recover or, as they must also look after their own interests, suggest a mutual collaboration to ensure their involvement?
Or do they simply rob them of the opportunity before Britain was even in a position to realise it was there?
I can't see it having happened any other way, unless the US were supposed to go ahead with the development of computing and volunteer reparations later, then welched on the deal.
Like I said, I don't want it to have been this way and I'm happy if someone knows it to be otherwise. But it could rank as some of the most despicable behavior ever to befall a 'special relationship'.
Much like upgrading is revision on android I suspect :)
How old is Mr. Barnes? I salute him for still being around and sharp!
Imagine how he felt.
I get nostalgic (sometimes) seeing the bits and pieces of things I used to work on rattling about in "free!" boxes under tables at electronic swap meets. And here HIS masters-piece is, restored and working!
Sounds like me in my early years.
Heck, it sounds like me today, except for maybe the toilet breaks...
Glad to see it's up and running again. Now I really must get up there to have a look and smell the valves!
OK, that's really cool and certainly a testament to building something reliable and repairable. A tip of the mug to these folks!
"I remember when it was all valves round here"
"where it was used to process mathematical calculations for Britain's nuclear program"
This smells like numerically solving differential equations.
So, how do you apply this thing to the task at hand? Do you program it somehow? Is there a dude with a large table for which he has to fill in a column and whereby he gets the values from the WITCH?
"This smells like numerically solving differential equations."
Closed-form solutions are for wimps.
Ding Dong the WITCH is... not Dead!
Well done folks. So - did you buy an UPS for that? ;-)
"World's oldest digital computer successfully reboots"
That explains why my web host was down for a while this afternoon.
Why didn't they just put a microchip into a calculator and then connect a regular laptop to the calculator, feeding in the problems and reading back the answers. Idiots
Downvoted for schoolboy humour.
I have revived some old systems in my time, but this is truly something.
On a side note, WITCH is great, but why do the British make such an @rse of acronym's, TNMOC?
NMC would be so much slicker, though TNMOC is certainly not the worst, I know of a group who elected to use OLHSFPALOS as their abbreviation (whatever the smiley for aghast is)
Amazing it as rebuilt to working condition. But I do have one question......(and I know I know this is old but still worth it sometimes) Can it run Crysis?
....... urPrivate Plunge/Pirate Purge/Public Master Pilot ARG Team of Inquisitive Beings with SMARTR Souls.
Such news has one wonder at what Virtual Pandora would Bletchley Boffin Types have Invented and Discovered QEd into Systems.
Well, surely you don't expect Blighty's Treasured X Stations not to have grown into AI Virtual Base Space Stations, which for the pleasure of all sizes and persuasion, are programmed towards XSSXXXX Code Levels of Behaviour Erotic and Exotic ....... Immaculately Perfect with All Flaws.
Ok now that is awesome
Amazing it as rebuilt to working condition. But I do have one question......(and I know I know this is old but still worth it sometimes) Can it run Crysis? .... James O'Brien Posted Wednesday 21st November 2012 05:51 GMT
Does the Pope wear a funny hat?:-)
I've heard of this before and I suspect *cost* may have had something to do with it as well. At 2 triodes per bit Vs 1 dekatron per decade I suspect the benefits soon mount up.
Note when people talked about "decimal" or binary machines in this era they were usually talking about BCD.
This is a a *true* decimal computer with counting by 10's built (literally) into the hardware. And note that qualification as the oldest *working* computer.
BTW What's with the picture? Was it taken with a wide angle lens or is the frame work *really* slightly curved, Cray 1 style?
Thumbs up on the restoration. I think this could be the *only* computer to survive a nearby nuclear explosion *without* being inside a bunker.
From the edges of the roof tiles it looks like it is a fish-eye pic, and in actuality the racks are in a line.
As someone who built stuff with valves when a student, I can tell you that each bit is represented by a double triode ( a simple latch circuit) and that 4 B7G double triodes and their bases were a lot cheaper than one dekatron. The problem, of course, is that the triodes are not indicating. Dekatrons are self debugging. Incandescent lamps do not work well at the sort of current you need per bit for a valve-based computer (about 5mA at 100V). And good luck with making that octal to decimal converter out of valves.
On the plus side, you can warm your hands on a valve- based timer.
> This is a a *true* decimal computer with counting by 10's built (literally) into the hardware.
Still true of modern PCs and Macs! Some 8-bit BCD instructions are buried in the 8086 instruction layer of Intel-family processors.
>> This is a a *true* decimal computer with counting by 10's built (literally) into the hardware.
> Still true of modern PCs and Macs!
No, not true of modern computers.
> Some 8-bit BCD instructions are buried in the 8086 instruction layer
Yes, but those are BCD instructions. The machine itself is a binary computer.
The machine in question is a *decimal* computer. Each storage node stores decimal values, not binary ones.
nosiy thing it was as well
We Intend To Create Havoc?
Well done to the guys there. I saw it some months ago while it was being restored and I'm pleased to see it up and running. Dekatrons are cool beasties - look for them in the Nixie-display calculators of the sixties, too, like the Anita. http://www.anita-calculators.info/html/the_technology_explained.html
There's an older, operational machine in Fujitsu's HQ's basement that is older, operational, and with no restoration. You can walk up up to turn it on, and watch it work. Go to Kawasaki and take a look.
Congratulations to all involved!
Biting the hand that feeds IT © 1998–2017