Me too, if that's smartphone was in 1960s, then you would not be able to hold it, the computer would be big enough it could be the size of probably about 100 Mansions put together and the average kid could get lost in the system easily and the computer would have to be under the ground to make space for other buildings probably deep underground and the building would have to still be modified to drill the wires through the building to connect it up to the console that controls the extremely ginormous system Plus you would have to add on to the building for a staircase to go down underground into the computer for maintenance, and the computer would be extremely loud and it would have to be cooled by ice and the computer would have to have its own power plant because the computer would consume a large amount of power and the fans would have to spin extremely fast and the fans would probably be so loud that it could actually permanently damage your hearing to the point that you cannot hear anything at all
Amazing video! I would also love to see these kind of video's of other old supercomputers. Its great to hear an expert speak about the things he loves.
My dad got hired as a refinery technician at the first oil refinery in the States built from the start to be computer controlled. When he started in 1968 the computers filled a 10000 square foot room. When he retired 25 years later they had upgraded the computers. It was odd to see a single machine only slightly larger than a desk top unit sitting in the middle of a huge, empty room. As it's been ANOTHER 20 or so years I wonder what it looks like now?
One question about the mercury tubes: How strong are they affected by vibrations of the machine? I could imagine that (accidentally) bumping into the shelf or walking around could trigger errors. How did they work against that influence?
i just love this kind of videos about those very primitive computers, they always make me excited about building my own... I mean... they make it look so incredibly simple... it makes me wonder if it actually IS that simple
I was reading this book called programming, principles and practice using C++. It's mentioned in the book, the first program ever to run on a stored-program computer (the EDSAC) was written and run by David Wheeler in the computer laboratory in Cambridge University, England, on May 6, 1949, to calculate and print a simple list of squares. This was an interesting video to watch regarding this. Edit: Imagine running DOOM on this. LOL. Although a guy named Sandy Douglas did run an OXO game on this thing using a cathode ray tube as the display. I'm just saying maybe it's possible?
That's so cool, I bet in the 1960s the Apollo computer was big enough were a little kid could crawl in the computer or a small sized adult and there were computers that took up a whole entire room, and every computer back then had to be installed in a building by the computers manufacturer
The question is, did they develop the uniselectors from automatic phone exchange equipment, or the other way around, or were they merely an uncanny bit of convergent evolution? They even seem to step at about the same speed as a rotary phone dial...
I am oddly surprised and impressed that it had hardware for multiplication. I can recall when building simulated processors it was often easier (far fewer gates) to handle it in software instead.
neeneko well these where made by the government with big money behind it. It is really important and seeing these where already slow machines emulating would cost too much time
Could you add some more context to the video description? things like EDSAC's intended task and year(s) of construction. ... or just a more obvious link to the main video, and this one does not stand on its own.
Intersting, thanks. I think the racks he described as the actual 'computer' part of the machine would be called the 'arithmetic logic unit' (ALU) today. And the 'main unit' is executing what is known as 'microcode' today.
I hope there are some young people around, too, so the knowledge about this technology gets passed on to the next generation. It would be a shame if this went the way of Greek fire, Damascus steel and terra preta.
Penny Lane though we have modern inventions that are far better and more advanced (napalm, modern high purity carbon steel/tungsten carbide, terra pretta is currently being researched intensely) I agree with your sentiment as a young person.
Penny Lane There are many classic computer hobbyists. Though, individual hobbyists are often limited in scope to minicomputers (takes up a small number of 19" racks for even the largest setups, as opposed to a whole room) or microcomputers (see: the Altair 8800, or original Apple II; small personal machines that you could put on a desk in other words) instead of real "big iron" like the EDSAC. For reasons of economy, and for reasons of ease of maintenance. A PDP-11/20 being mostly SSI logic ICs, that you can find replacements for at reasonable prices. Of coruse there are museums for the "big" machines, and some hobbyists do own "big" machines. If you're interested, the Living Computer Museum in Seattle has many classic computers from the 1970s, and 1980s up and running for the visiting public to use.
+AdmiralCreideiki As a young electronics engineering student, who has spent years drooling over computers & circuits, I desperately want a PDP-8 or -11 to take apart and learn how to use and maintain... one of these days, I hope.
+Aurelius R You can find small PDP-11 systems for sale occasionally now, though they'll be more the later models like the QBUS PDP-11/23, or QBUS PDP-11/73. The "classic" PDP-11s with the blinkenlights front panels and mostly SSI logic construction are rather larger (due to their construction), and unfortunately *very* expensive sinc ethey are "collectible" now-a-days. PDP-8 machines have just become a giant pile of expensive because they too are now "collectible". If you wanted to take a look at what makes a PDP-11 or PDP-8 tick, the BitSavers documentation archive has a very big set of documentation for both the PDP-8 and PDP-11 machines (as well as of DEC's other machines). I don't know if RUclips allows pasting in URLs, if it does I can link you right to the page of PDP-11 documentation. Also, it is very possible to emulate a PDP-8 or PDP-11. The SIMH project provides a nice free emulation of both systems (amongst other systems) and it is easy enough to find software.
Root3264 Anything with a sufficient delay/decay that can be electrically driven and read can store things. People even used CRTs to store information back in the day.
Root3264 Yeah, I'm also surprised that this at some point apparently was the obvious thing to do. I had assumed something like magnetic-core memory has always been available.
Penny Lane im sure they had other options, say paper tape as a stack and registers or use relays to make a latch/ memory block. The problem is most probably they werent up to the valve's speed. Maybe if they designed some special relays with really high permeability and low reluctance they could have made way faster purpose specific relays but if they could do that they could just invent magnetic core memory instead. Not sure, but apparently they hadnt researched magnetics materials technology just yet.
@@laharl2k A bit late but from what I've understood from this and their previous videos they did use paper tape for temporary as well as long term storage. I think the delay lines were used by the computer part while actively performing computations (perhaps similar to a cache in modern processors) while the paper tape held the rest of the program or other variables that weren't needed at that part of the calculation.
zeos386sx how can something be said ignorantly in jest? If you're saying it in jest, you presumably know it's untrue, therefore meaning it's not ignorant to say it. Were I to say "I bet this could run minecraft" it's only ignorant if I honestly do think it could, being that I know it could not run minecraft, I am not ignorant. tldr: you're an idiot.
So the memory is held in pulses circulating through the mercury, into the wire through the crystal, and back to the mercury tube through another crystal? My understanding is that it cycles through that, and I guess those circuits amplify the electric signal coming through so that any lost energy is added back like that. If you held it while it was holding bits and connected to power, could you feel the pulses go through? If the memory is stored in the circuits, then I couldn't see what the point of the tube of mercury is for. I guess it's to slow the signal transfer down to the speed of sound instead of travelling quickly through a length of wire.
kindpotato Current RAM technology also needs constant refreshing. I doubt you could feel the pulses ... i wouldn't try ... this thing runs on valves, and valves are high voltage high frequency switching devices ...
detaart it should be reasonably safe to touch them. The metal tube would be at chassis ground. And the energy flowing through them is mechanical (sound waves), not electrical.
Modern RAM works exactly the same way. You need to constantly rewriting bits by shooting electrons back into the transistor to avoid losing the data. That's why when you shut down the power to the RAM it loses the data.
kindpotato The duration of the memory is the time delay required for the impulses to traverse the length of the tube, plus the time to regenerate the signal and feed it back into the mercury. This is why he uses the terms memory and "delay line" Interchangeably. I point out that when Andrew speaks of the pulses travelling at the speed of "sound," it is the speed of sound _in mercury_ they refer to (1450 m/s). You would not feel the pulses because they are carefully focused at the receiving piezoelectric crystal, and do not touch the tube itself. Judging from the Wikipedia article, correctly focusing the pulses was a significant chore. To minimize losses, the mechanical impedances of the crystals and the medium had to match (this is the reason mercury was chosen). To maintain this, the tubes had to be kept at a constant temperature (in this case, 40 °C). en.wikipedia.org/wiki/Delay_line_memory#Mercury_delay_lines
Would it be possible to create flip flops and latches with this technology? Having large tubes of mercury around seems pretty annoying to me when they leak.
Lttlemoi Well, it would be possible in theory, but in reality valve memory would be much more unreliable than a delay line, use massive amounts of power, and you would probably have 75-80 000 tubes to make it using such memory, which would be a bit excessive
Since they fudged the merc delay lines with wires, they could of fudged the vacuum tubes with a transistor inside, which requires less power, and more reliable
trailkeeper No, they don't work the same - the speed obviously being the big issue. Transistors are obviously smaller, and faster. Obviously I don't know the layout of EDSAC, but that can wreak havoc with timing. Pus, there's the whole reason for this. Some things you have to be realistic about - using your point, why couldn't they update EVERYTHING and just build a modern computer? The simple reason of the mercury delay lines is they are UTTERLY impractical. As they said, one box would be tens of thousands of pounds each, and extremely poisonous. It's no different to rebuilding an antique car, or renovating antique furniture - you want to remain as CLOSE as possible to the original, but there are some things that are either impractical or sometimes illegal, even.
Daniel Linger Ok, Id rather see the original tube version restored, but if they could not get vacuum tubes they would have to fake it a bit, but putting a transistor inside some container with pins like a tube would have. I understand what you are saying, its like putting modern wheels or a modern engine in a Model T car. Transistors are fast, but it would still use the same clock of the old computer, in fact a transistor works at both low frequency DC up to GHZ, billions of cycles, high frequency.
It's like a museum or does it have some actual uses? It's impressive but most of those things would be easier to simulate rather than build actual mercury tubes or mechanical machines.
Zachary Kaplan Bootstrapping (Booting): refers to the process of loading the basic software into the memory of a computer after power-on or general reset, especially the operating system which will then take care of loading other software as needed.
In the 80s there were powerfull mainframe computers sold by the IBM and some other companies, with 32 or even 64-byte words, 40 MHz clock and megabytes of available RAM memory. They were just insanely expensive and you can't compare them to low-end home market personal computers.
so, the echo device used by guiter players could be a .. delay line ? aha... that boot-selector pulse system - is that comparable to the dail-in system of phone systems - i know that it will generate the 0/1 effective a program that tells the system what to do.. first, next and so on - but - that ring pulse system looks so.. alike the phone number selector.. - just at glance.
Holy cow! Sound pulses in puddles of mercury for storage? Holy data corruption, Batman, at a simple clap of the hands. Seems like they'd only be usable in a studio.
sd4dfg2 idk if they ever will, the apple macintosh for example has aged better than command line computers as we still use GUI, until that changes. This looked as primitive in 1984 as it does today.
Vendetta042 I forgot where this quote came from, or exactly what it said, but it was along the lines of "science will soon progress to become so advanced it will be indistinguishable from black magic".
kindpotato I don't know for sure, but I would imagine that it was built to a spec - the people building it would have asked the physicists, mathematicians etc. who would eventually be the end users what they needed. 10 digits was probably the best compromise between what was technically feasible/affordable, and what the users said they wanted. Using two words to store one value wouldn't be feasible - there was about an 800 word limit on programs when it was first built (prior to upgrades). Every bit mattered.
kindpotato early digital computers usually had word sizes that were multiples of 6. Many of the current standards (like the 8 bit byte) were introduced by IBM in the 1960s and became industry standards.
+kindpotato Every bit costs a huge amount more to add. More tubes of mercury, more valves, probably an exponential rise in cost at the time for every additional bit.
ENIAC had 10 decimal digits, because a requirement was to be able to calculate with 8 decimals of precision (so at least 9 digits), and you want some extra up top to facilitate programming (making sure you never overflow a multiplication if you scale your numbers correctly). EDVAC, the successor, needed to match the 10 decimal digits to facilitate conversion, so needed at least 34 bits, plus 1 bit for a sign. It was to be a stored program computer, so the word length needed to match the instruction length at least roughly. They wanted a 1024-words memory, which required 10-bit addresses, and a 3-address instruction format, thus a 35-bit word would have worked fine (with 32 instruction opcodes). Except... the problem with recirculating memory is that the moment your computer ended reading the first instruction, a new instruction will be going through its terminals, leaving no time for execution, and forcing you to wait for the next time the next instruction comes out the mercury delay line. Eventually the EDVAC designers decided to use an extra address on the instruction format to set what instruction would come next (basically, EVERY instruction was a jump), because that made the machine quite a bit faster: you could set the next instruction to be exactly the address that was going through the circulating memory at the moment you ended your first instruction, saving time. This required a 44-bit word. Maurice Wilkes designed the EDSAC after learning about the early sketches for the EDVAC, but wanted to simplify the design while maintaining some speed and the same capacity. He cleverly made it a one-address machine and added support for half-precision numbers. You needed at least a 35-bit full-word to get the precision to match ENIAC, so a 18-bit half-word and 36-bit full word would have been great. But he had trouble using all the bits due to timing, so he sacrificed the first, resulting in 17 bit instructions (with one 10-bit address, 32 instruction opcodes, one length bit to select half or full words and one spare), 17-bit half numbers and 35 bit full-numbers.
14 commands the hardware could do. I presume they would include add, subtract, multiply, divide, select register, copy, delete, print (I'm not an expert). In modern systems we have complex programming languages with many commands that get compiled into code (or interpreted by a program that actively converts the commands) that runs on far fewer instructions that the actual CPU/GPU can perform.
Pod Theelder Yup - Western Electric telephone switches used rotaries straight up through to the #5 Crossbar system. It was mostly for sequencing uses then because the Crossbar systems brought in that thing called the Marker.
Joseph Victor Been done many many times. Google or RUclips searching should be able to show you some examples. Several fully featured computers were built with vanilla Redstone pretty early on after it was made available.
+Joseph Victor Instead of doing that, why not get yourself a soldering iron and actually build a REAL circuit? It'll last longer and you'll learn something.
Alexander Rusin The price in electricity is horrendous. I used to use an old IBM System 38 (?) which was all punch tape and teletypes with the toggle switches. The electricity bill was in the thousands.
@@crunchyfrog555 11KW, for comparison my gaming PC has a 750W supply and probably never uses more than 300W (if I had an RTX 3000 series card it would be a bit higher).
Your cameraman is useless; if he's not zooming in and out, in and out, he's wobbling all over the place and doesn't dwell on a subject long enough for us to take it in. Get a new cameraman please, or teach this one BASIC skills. Super videos, thank you :)
I find it fascinating being able to watch this on a pocket sized smartphone today.
Me too, if that's smartphone was in 1960s, then you would not be able to hold it, the computer would be big enough it could be the size of probably about 100 Mansions put together and the average kid could get lost in the system easily and the computer would have to be under the ground to make space for other buildings probably deep underground and the building would have to still be modified to drill the wires through the building to connect it up to the console that controls the extremely ginormous system Plus you would have to add on to the building for a staircase to go down underground into the computer for maintenance, and the computer would be extremely loud and it would have to be cooled by ice and the computer would have to have its own power plant because the computer would consume a large amount of power and the fans would have to spin extremely fast and the fans would probably be so loud that it could actually permanently damage your hearing to the point that you cannot hear anything at all
ok?
Hand working analogue electronics in jacket and tie - that is some epic class right there!
WOW! I haven't heard anything of EDSAC since Computer Studies at school in the 80s... and here it is!!!
The mercury thing is nuts, how do people come up with this stuff. Peoples brains are amazing.
Well, some people's brains.
Freddy Mercury's brain
Water/alcohol tube memory works the same way and is much more environmentally sound. Clearly someone allowed a mercury vendor to write the RFP.
Amazing video! I would also love to see these kind of video's of other old supercomputers. Its great to hear an expert speak about the things he loves.
The boot device mentioned at about 9:50 looks like an old analog telephone switch known as a Strowger switch.
There always seem to be new Computerphile episodes that I haven't seen yet and must watch! Such a great channel!
Anyone who has programmed in assembly language can really appreciate this explanation.
My dad got hired as a refinery technician at the first oil refinery in the States built from the start to be computer controlled. When he started in 1968 the computers filled a 10000 square foot room. When he retired 25 years later they had upgraded the computers. It was odd to see a single machine only slightly larger than a desk top unit sitting in the middle of a huge, empty room.
As it's been ANOTHER 20 or so years I wonder what it looks like now?
There's probably a lonely raspberry pi computer sitting in the middle of that room.
elfoxy
Maybe.
One question about the mercury tubes: How strong are they affected by vibrations of the machine? I could imagine that (accidentally) bumping into the shelf or walking around could trigger errors.
How did they work against that influence?
i just love this kind of videos about those very primitive computers, they always make me excited about building my own...
I mean... they make it look so incredibly simple... it makes me wonder if it actually IS that simple
The delay line memory stuff is really interesting:
en.wikipedia.org/wiki/Delay_line_memory
Utterly fascinating. Love it.
Love the EDSAC series of videos!
I was reading this book called programming, principles and practice using C++. It's mentioned in the book, the first program ever to run on a stored-program computer (the EDSAC) was written and run by David Wheeler in the computer laboratory in Cambridge University, England, on May 6, 1949, to calculate and print a simple list of squares. This was an interesting video to watch regarding this.
Edit: Imagine running DOOM on this. LOL. Although a guy named Sandy Douglas did run an OXO game on this thing using a cathode ray tube as the display. I'm just saying maybe it's possible?
That's so cool, I bet in the 1960s the Apollo computer was big enough were a little kid could crawl in the computer or a small sized adult and there were computers that took up a whole entire room, and every computer back then had to be installed in a building by the computers manufacturer
When I was working in a telephone exchange in the 1980s, we had selectors like you can see from 8:20
The question is, did they develop the uniselectors from automatic phone exchange equipment, or the other way around, or were they merely an uncanny bit of convergent evolution? They even seem to step at about the same speed as a rotary phone dial...
ok?
It's like looking at the first PC ever being built by a bunch of college age nerds. It's beautiful, especially the veterans there.
Very interesting. Thank you !
But could I make a suggestion? Can the cameraman have a microphone for him too, so we can better hear his questions ?
It's insane that this was once state of the art and now my TI calculator can probably do the same stuff EDSAC could do.
I am oddly surprised and impressed that it had hardware for multiplication. I can recall when building simulated processors it was often easier (far fewer gates) to handle it in software instead.
neeneko well these where made by the government with big money behind it. It is really important and seeing these where already slow machines emulating would cost too much time
Memory cost was higher than the cost of the whole rest of the computer, so they had to save on code size.
2:31 pizza electric crystal. Really relaxed pronunciation of piezo.
kindpotato mmm pizza
MichaelKingsfordGray *dies*
Lol yeah with mercury topping
Could you add some more context to the video description? things like EDSAC's intended task and year(s) of construction.
... or just a more obvious link to the main video, and this one does not stand on its own.
Bbonno Try here:
bit.ly/whatisedsac
and this is the main video: ruclips.net/video/C97MtJWDNX8/видео.html
>Sean
***** thanks, much appreciated
Just been reading about Mercury delay lines. Very interesting stuff.
Great video, nice work on the outro, love it!
It's like "Trainspotting" for electronics history buffs!
Intersting, thanks.
I think the racks he described as the actual 'computer' part of the machine would be called the 'arithmetic logic unit' (ALU) today.
And the 'main unit' is executing what is known as 'microcode' today.
You're almost correct except that it doesn't use microcode, but hard wired control
Even the idea of microcode would have cost thousands of dollars at the time :).
I hope there are some young people around, too, so the knowledge about this technology gets passed on to the next generation. It would be a shame if this went the way of Greek fire, Damascus steel and terra preta.
Penny Lane though we have modern inventions that are far better and more advanced (napalm, modern high purity carbon steel/tungsten carbide, terra pretta is currently being researched intensely) I agree with your sentiment as a young person.
Penny Lane
There are many classic computer hobbyists. Though, individual hobbyists are often limited in scope to minicomputers (takes up a small number of 19" racks for even the largest setups, as opposed to a whole room) or microcomputers (see: the Altair 8800, or original Apple II; small personal machines that you could put on a desk in other words) instead of real "big iron" like the EDSAC. For reasons of economy, and for reasons of ease of maintenance. A PDP-11/20 being mostly SSI logic ICs, that you can find replacements for at reasonable prices.
Of coruse there are museums for the "big" machines, and some hobbyists do own "big" machines. If you're interested, the Living Computer Museum in Seattle has many classic computers from the 1970s, and 1980s up and running for the visiting public to use.
+AdmiralCreideiki As a young electronics engineering student, who has spent years drooling over computers & circuits, I desperately want a PDP-8 or -11 to take apart and learn how to use and maintain... one of these days, I hope.
+Aurelius R
You can find small PDP-11 systems for sale occasionally now, though they'll be more the later models like the QBUS PDP-11/23, or QBUS PDP-11/73. The "classic" PDP-11s with the blinkenlights front panels and mostly SSI logic construction are rather larger (due to their construction), and unfortunately *very* expensive sinc ethey are "collectible" now-a-days.
PDP-8 machines have just become a giant pile of expensive because they too are now "collectible".
If you wanted to take a look at what makes a PDP-11 or PDP-8 tick, the BitSavers documentation archive has a very big set of documentation for both the PDP-8 and PDP-11 machines (as well as of DEC's other machines). I don't know if RUclips allows pasting in URLs, if it does I can link you right to the page of PDP-11 documentation.
Also, it is very possible to emulate a PDP-8 or PDP-11. The SIMH project provides a nice free emulation of both systems (amongst other systems) and it is easy enough to find software.
+Penny Lane ..the NHS, public education, etc.
reminds me of HEX, needs an Anthill Inside sticker
Back in the days, someone on the EDSAC team suggested an EDSAC watch, but the idea turned down.
The memory blows my mind
Love it! Any estimates on a completion date?
i recommend renaming to "inside the EDSAC computer"
A five foot long tube about 2 inches in diameter to store 32 bits. How big would a gigabyte be?
loperspest You've got to suspect that an 8 gigabyte pen drive would have been really BIG!
loperspest Probably exceeds any reasonable limit to be considered "portable".
31.250 million tubes ,488 billion dollars worth of mercury
Approximately the size of yo mama
Mind equals blown.
very impressive! 36 bits!
Thank you sir
Wow, I never knew they used Mercury to store stuff back then! (:
Root3264 Anything with a sufficient delay/decay that can be electrically driven and read can store things. People even used CRTs to store information back in the day.
Well, I meant mercury for basically everything, it was just eye-opening for me to use something other than vacuum tubes ^^
Root3264 Yeah, I'm also surprised that this at some point apparently was the obvious thing to do. I had assumed something like magnetic-core memory has always been available.
Penny Lane im sure they had other options, say paper tape as a stack and registers or use relays to make a latch/ memory block. The problem is most probably they werent up to the valve's speed.
Maybe if they designed some special relays with really high permeability and low reluctance they could have made way faster purpose specific relays but if they could do that they could just invent magnetic core memory instead. Not sure, but apparently they hadnt researched magnetics materials technology just yet.
@@laharl2k A bit late but from what I've understood from this and their previous videos they did use paper tape for temporary as well as long term storage. I think the delay lines were used by the computer part while actively performing computations (perhaps similar to a cache in modern processors) while the paper tape held the rest of the program or other variables that weren't needed at that part of the calculation.
They have been working on this for 5 years. I wonder what is the state of this machine in 2018.
Amazing Machine!
Hmm, what would emulate those delay lines I wonder.
I want a speed-up knob on my pc...
It's probably more meant to fine-tune the clock speed to match the wave propagation through the delay lines so that everything stays in sync...
You probably have one. It's called BCLK. I wouldn't turn it up too high.
@@Triantalex ok 🥰
me as a programmer
im obsessed with technology like thia i dont know why
maybe i was finding a way
Why didn’t they use oil instead of mercury?
This thing is probably still faster than a Xbox One.
tjpld that is about the most ignorant thing ever said, even if only in jest.
zeos386sx its a joke you idiot
Juan Paolo Gillera That's why I said "even if only in jest"
zeos386sx how can something be said ignorantly in jest? If you're saying it in jest, you presumably know it's untrue, therefore meaning it's not ignorant to say it. Were I to say "I bet this could run minecraft" it's only ignorant if I honestly do think it could, being that I know it could not run minecraft, I am not ignorant. tldr: you're an idiot.
tjpld That'l teach you to make a joke amongst nerds
Cool. Just yesterday I visited CSIRAC in Melbourne. The only surviving computer of this generation.
I know you are Brits, but could you make a video on the Zuse Z3? :)
So the memory is held in pulses circulating through the mercury, into the wire through the crystal, and back to the mercury tube through another crystal? My understanding is that it cycles through that, and I guess those circuits amplify the electric signal coming through so that any lost energy is added back like that. If you held it while it was holding bits and connected to power, could you feel the pulses go through? If the memory is stored in the circuits, then I couldn't see what the point of the tube of mercury is for. I guess it's to slow the signal transfer down to the speed of sound instead of travelling quickly through a length of wire.
kindpotato Current RAM technology also needs constant refreshing.
I doubt you could feel the pulses ... i wouldn't try ... this thing runs on valves, and valves are high voltage high frequency switching devices ...
detaart it should be reasonably safe to touch them. The metal tube would be at chassis ground. And the energy flowing through them is mechanical (sound waves), not electrical.
Modern RAM works exactly the same way. You need to constantly rewriting bits by shooting electrons back into the transistor to avoid losing the data. That's why when you shut down the power to the RAM it loses the data.
kindpotato The duration of the memory is the time delay required for the impulses to traverse the length of the tube, plus the time to regenerate the signal and feed it back into the mercury. This is why he uses the terms memory and "delay line" Interchangeably. I point out that when Andrew speaks of the pulses travelling at the speed of "sound," it is the speed of sound _in mercury_ they refer to (1450 m/s).
You would not feel the pulses because they are carefully focused at the receiving piezoelectric crystal, and do not touch the tube itself. Judging from the Wikipedia article, correctly focusing the pulses was a significant chore. To minimize losses, the mechanical impedances of the crystals and the medium had to match (this is the reason mercury was chosen). To maintain this, the tubes had to be kept at a constant temperature (in this case, 40 °C).
en.wikipedia.org/wiki/Delay_line_memory#Mercury_delay_lines
Would it be possible to create flip flops and latches with this technology? Having large tubes of mercury around seems pretty annoying to me when they leak.
Delay lines (mercury, wire torsion, etc) are a lot smaller and take less power then vacuum tube memory.
Lttlemoi Well, it would be possible in theory, but in reality valve memory would be much more unreliable than a delay line, use massive amounts of power, and you would probably have 75-80 000 tubes to make it using such memory, which would be a bit excessive
Since they fudged the merc delay lines with wires, they could of fudged the vacuum tubes with a transistor inside, which requires less power, and more reliable
trailkeeper No, they don't work the same - the speed obviously being the big issue. Transistors are obviously smaller, and faster. Obviously I don't know the layout of EDSAC, but that can wreak havoc with timing.
Pus, there's the whole reason for this. Some things you have to be realistic about - using your point, why couldn't they update EVERYTHING and just build a modern computer? The simple reason of the mercury delay lines is they are UTTERLY impractical. As they said, one box would be tens of thousands of pounds each, and extremely poisonous.
It's no different to rebuilding an antique car, or renovating antique furniture - you want to remain as CLOSE as possible to the original, but there are some things that are either impractical or sometimes illegal, even.
trailkeeper Transistors are current driven, valves are voltage driven. BIG difference.
detaart One type of transistor is current driven, usually the BJT, but the FET is high impedance type, not much current needed.
Daniel Linger Ok, Id rather see the original tube version restored, but if they could not get vacuum tubes they would have to fake it a bit, but putting a transistor inside some container with pins like a tube would have. I understand what you are saying, its like putting modern wheels or a modern engine in a Model T car. Transistors are fast, but it would still use the same clock of the old computer, in fact a transistor works at both low frequency DC up to GHZ, billions of cycles, high frequency.
Daniel Linger In fact, when they are done with this computer with the tubes, they should make the exact same design with transistors.
Why those machines were build in shelf style instead of PCB board style? Also are those red black jacks used for debuging or something?
Because they are massive. Easier to work on them if they are on a rack.
PCB not invented yet, these recreate the original computer.
It's like a museum or does it have some actual uses? It's impressive but most of those things would be easier to simulate rather than build actual mercury tubes or mechanical machines.
+Globbi You *completely* miss the point of this.
Can someone explain the point that was made at the very end about "bootstrapping"?
Zachary Kaplan Bootstrapping (Booting): refers to the process of loading the basic software into the memory of a computer after power-on or general reset, especially the operating system which will then take care of loading other software as needed.
I wonder how powerful a computer would be if engineers used the same design philosophy they did back then.
She looked very interesting when I saw her. Also the machine opposite with all the spinning lights.
Today is Saint Isidore of Seville's day, the internet's patron saint. I'll pray for ye, guys. Ye are saving our culture, just like he did.
Yeah but can it run OXO?
But does it have dedodated wam?
So they had 36 bit computing back when this was built. Only 8 or 16 bits in the 1980s
In the 80s there were powerfull mainframe computers sold by the IBM and some other companies, with 32 or even 64-byte words, 40 MHz clock and megabytes of available RAM memory. They were just insanely expensive and you can't compare them to low-end home market personal computers.
36 Bit means the amount of data they were able to hold. Not adress-words, data words. Like a 36 Bit harddisk.
This would be called at best a 17-bit computer if it was built today.
so, the echo device used by guiter players could be a .. delay line ?
aha... that boot-selector pulse system - is that comparable to the dail-in system of phone systems - i know that it will generate the 0/1 effective a program that tells the system what to do.. first, next and so on - but - that ring pulse system looks so.. alike the phone number selector.. - just at glance.
Andrew Herbert talks like prof. Poliakoff
yes!
Why did the use mercury and not water or so?
+Stabacs McBass Because of the specific delay and impedance of mercury -- much more conductive and pure than water.
Interesting
Holy cow! Sound pulses in puddles of mercury for storage? Holy data corruption, Batman, at a simple clap of the hands. Seems like they'd only be usable in a studio.
Update please
This fascinates but I don't understand computing :(
But will it run Crysis?
What we use today compared to that is absolute magic.
And in a short time, today's computers will look just as crude and archaic.
sd4dfg2 idk if they ever will, the apple macintosh for example has aged better than command line computers as we still use GUI, until that changes. This looked as primitive in 1984 as it does today.
Vendetta042 I forgot where this quote came from, or exactly what it said, but it was along the lines of "science will soon progress to become so advanced it will be indistinguishable from black magic".
Malcolm Forde "Any sufficiently advanced technology is indistinguishable from magic". Arthur C Clarke.
diggers92 Wow, I was way off :P Thanks.
36 bits? That's so long why add another 4 bits. Why exactly 10 digits?
kindpotato I don't know for sure, but I would imagine that it was built to a spec - the people building it would have asked the physicists, mathematicians etc. who would eventually be the end users what they needed. 10 digits was probably the best compromise between what was technically feasible/affordable, and what the users said they wanted. Using two words to store one value wouldn't be feasible - there was about an 800 word limit on programs when it was first built (prior to upgrades). Every bit mattered.
kindpotato early digital computers usually had word sizes that were multiples of 6. Many of the current standards (like the 8 bit byte) were introduced by IBM in the 1960s and became industry standards.
+kindpotato Every bit costs a huge amount more to add. More tubes of mercury, more valves, probably an exponential rise in cost at the time for every additional bit.
ENIAC had 10 decimal digits, because a requirement was to be able to calculate with 8 decimals of precision (so at least 9 digits), and you want some extra up top to facilitate programming (making sure you never overflow a multiplication if you scale your numbers correctly).
EDVAC, the successor, needed to match the 10 decimal digits to facilitate conversion, so needed at least 34 bits, plus 1 bit for a sign. It was to be a stored program computer, so the word length needed to match the instruction length at least roughly. They wanted a 1024-words memory, which required 10-bit addresses, and a 3-address instruction format, thus a 35-bit word would have worked fine (with 32 instruction opcodes).
Except... the problem with recirculating memory is that the moment your computer ended reading the first instruction, a new instruction will be going through its terminals, leaving no time for execution, and forcing you to wait for the next time the next instruction comes out the mercury delay line. Eventually the EDVAC designers decided to use an extra address on the instruction format to set what instruction would come next (basically, EVERY instruction was a jump), because that made the machine quite a bit faster: you could set the next instruction to be exactly the address that was going through the circulating memory at the moment you ended your first instruction, saving time. This required a 44-bit word.
Maurice Wilkes designed the EDSAC after learning about the early sketches for the EDVAC, but wanted to simplify the design while maintaining some speed and the same capacity. He cleverly made it a one-address machine and added support for half-precision numbers. You needed at least a 35-bit full-word to get the precision to match ENIAC, so a 18-bit half-word and 36-bit full word would have been great. But he had trouble using all the bits due to timing, so he sacrificed the first, resulting in 17 bit instructions (with one 10-bit address, 32 instruction opcodes, one length bit to select half or full words and one spare), 17-bit half numbers and 35 bit full-numbers.
That's not inside a computer. I've seen Tron enough times to know it isn't like that.
;-P
250,000 dollars?
Better run 400,000 windows of Battlefield 4 and ArmA 3 on Ultra at 9,000,000,000 FPS.
Listen carefully and you can hear the characteristic clickity-click sounds of the WITCH computer in the background.
Hello
Hi
@@abstractapproach634 Hey there
hes holding a register
14 instructions?
+Alan Moy It's all you need :)
Alan Moy you can work with 4 opcodes to do anything
14 commands the hardware could do. I presume they would include add, subtract, multiply, divide, select register, copy, delete, print (I'm not an expert). In modern systems we have complex programming languages with many commands that get compiled into code (or interpreted by a program that actively converts the commands) that runs on far fewer instructions that the actual CPU/GPU can perform.
but... can it run minecraft ?
Sounds like an SxS telephone switch.
kd1s I believe that's what it is.
Pod Theelder Yup - Western Electric telephone switches used rotaries straight up through to the #5 Crossbar system. It was mostly for sequencing uses then because the Crossbar systems brought in that thing called the Marker.
I can hear all the audiophile weeping about those vacuum tubes being wasted on a museum piece :)
Minecraft nerd question: I wonder if its possible to make a very very basic computer with redstone?
Joseph Victor Been done many many times. Google or RUclips searching should be able to show you some examples. Several fully featured computers were built with vanilla Redstone pretty early on after it was made available.
+Joseph Victor Instead of doing that, why not get yourself a soldering iron and actually build a REAL circuit? It'll last longer and you'll learn something.
*nods and smiles* yup, I understood none of that
I could fit this in my living room, anyone know the price? haha
Alexander Rusin The price in electricity is horrendous. I used to use an old IBM System 38 (?) which was all punch tape and teletypes with the toggle switches. The electricity bill was in the thousands.
@@crunchyfrog555 11KW, for comparison my gaming PC has a 750W supply and probably never uses more than 300W (if I had an RTX 3000 series card it would be a bit higher).
@@grn1 Yes, I know that. That's the point.
40 instruction ROM? :) Not bad.
i know GITSAC :D
O my god! That thing is amazing! :D
Suck a fat one, Alienware... lol.
I bet this'll run CryEngine 3 just as fine as my trustworthy Commodore PET! :3
He's a bit inaccurate early on. Technically all computers are symbol manipulators.
piezoelectric crystal sounds like pizza electric crystal.
Mmmh pizza electric crystal
Your cameraman is useless; if he's not zooming in and out, in and out, he's wobbling all over the place and doesn't dwell on a subject long enough for us to take it in. Get a new cameraman please, or teach this one BASIC skills.
Super videos, thank you :)
Would they have used banana plug wires on the original? Those look suspiciously like banana plug ports.