# How long would it take to create a Windows 1.0 capable machine from complete scratch?

I've been thinking about this concept for a while and have not been able to figure out what could be a plausible answer. It seemed appropriate for my first question on this website.

Imagine an experiment where a group of approximately 20 humans are dropped off on a fairly small (~1000 km²) island that contains all necessary resources as they would appear in a natural environment. The group would then receive the challenge to - starting completely from scratch - build a fully functioning computer system capable of running at least Windows 1.0 with usable speed and then run it successfully as fast as possible. They would start off with no tools or resources. These are the rules and conditions that would be present:

• The group of humans would not need to worry about life supporting and maintaining issues such as food, clothing, weather conditions, natural disasters and hostile wildlife.
• The group would know exactly how to find and assemble any items involved in the process of creating the machine.
• The group consists of young and fit humans that would not experience any social issues within the group, and would not tire from consistent 12-hour working days.
• For convenience, we assume that the humans would not age or die during their participation in this experiment.
• The island contains any necessary resource in completely raw form. Materials like metals may be present in a higher-than-natural rate to ensure there is enough available for completing the challenge.

As I started thinking more about this concept, I began expecting the minimum time needed to achieve such a goal would probably be at least 5 years (Edit: way above). However I am probably underestimating the time it would take to obtain some of the necessary materials and build all of the advanced machinery that is used in assembling a fully functional computer system.

Could anyone suggest a reasonable time estimate for completing such an extreme task/challenge? What would be the biggest obstacles along the way?

As this is my first time ever post on any StackExchange website, feel free to point out anything I should be doing differently.

Edit: The challenge does not require building a version of the hardware that was actually being used to run and interact with the OS. As long as it gets the job done and the system is able to run at usable speeds, it could be built from any material and can be as big as it needs to be.

Edit: I have accepted Karl's answer as it portrays the most factual sequence of steps that would have to be taken in order for the team to achieve the necessary level of technology for building the machine.

• It would be entirely impossible - The island does not contain a copy of WIndows 1.0. – Zibbobz Sep 12 '16 at 12:47
• Why windows? Why not other OS? Why run any OS at all, instead of providing a machine with proper speed? We can measure memory or clock speed, you know. That's more reliable than checking whether some arbitrary OS will run at 'usable speed'. – MatthewRock Sep 12 '16 at 12:58
• I remember Windows 1.0 - claiming it ran would be enough to make me laugh... – Jon Custer Sep 12 '16 at 13:36
• If the humans live forever and never have to worry about survival, then really all you need is 1 person and a bunch of rocks – Benubird Sep 12 '16 at 15:05
• How I built a toaster from scratch – user3106 Sep 14 '16 at 9:58

This demands a full chemical industry developed. (Let's guess they can get at iron ore and coal somehow, and suitable material to make ovens, and you swiped a few axes, saws and shovels to start with. I cannot guess how long it might take to bootstrap those.)

Factories they need to build in chronological order.

• bricks, cement & construction supplies
• blast furnace
• machineshop
• steel production
• keep improving all previous production sites at all times
• glass factory
• base chemicals factory
• advanced machineshop (lathe, milling machine)
• polymer production
• copper&electrical cable production
• electrical power station (this needs a bit more thinking, might need to be quite large)

Now you're in ~1890!

• aluminium factory
• semiconductor factory
• build first discrete electronics to help with all machining
• make your first integrated circuits
• build first computer with ICs, start programming
• better computer, more highly integrated circuits
• develop programming environment for the final task

20 steps, let's say I forgot another five. You can probably do every task in a year or two, if you have 20 people for it and know exactly how, but you will run out of personnel very fast. All the earlier factories need to keep running while you build new ones, and you will need ever more people to keep maintaining and upgrading everything. The factories have to grow all the time to produce base material for all the new things you "invent". And you need ever more people to do the logistics&infrastructure and dig up the base materials. My guess would be 35 years and 20000 people, depending on how you get past the first steps. Perhaps half a million man-years. You have no chance with 20 people. ;-)

An open question would be how to power all this. Hydroelectric and coal could do the trick, but one would need an estimate of the amount of electrical and heating power needed. At some point solar power could come into play.

P.S.: Afterthought: The personnel requirements could perhaps be halved if you're really crazy and make this system to collapse with the target reached, i.e. no resources left, factories&infrastructure ruined of old age, etc.

P.P.S. I might add that I thought the people constructing everything still need to dimension everything, i.e. they know the general rules, formulas, physical constants, but don't have a readymade drawing for every machine. Giving them a huge stack of premade blueprints seemed like cheating to me, and impractical, because it'd be hard to know e.g. the exact mechanical properties of the stuff they produce, before actually doing it there. It'd be another 20 years of science&engineering today, to prepare plans for all contingencies. ;-)

P.P.P.S. Why all the factories? The 8086 is at the top of 20 years of integrated circuit development, and you need a lot of electronics already to build and test the machines that are used to actually make one 8086. The last steps can probably be more manufacture than factory, but I am sure you'll have to make dozens of the ICs each time before you get one that works (how would you know that your wafers' specs are sufficient, without building even more sophisticated analytics?).

P.P.P.P.S Why ICs? It is impossible to build a general purpose CISC out of discrete transistors and let it run at several MHz. A parallelised RISC supercomputer (like the CDC 6600 mentioned in the disc.), no problem, but we are talking binary compatible to the IBM PC. Further, millions of transistors for the SRAM would be a pain to build and assemble by hand, and the latencies in the long wiring (not talking about the capacitance and inductivity) would make it inoperational in an 8086.

• P.P.P.P.P.S: You don't need all those PS's. Just integrate all of that new information into your answer naturally. Otherwise, it reads like a newsfeed or forum. – Robert Harvey Sep 13 '16 at 19:09
• 20 people with all construction supplies ill, in their lifetime, barely build all walls for all factories you ill need to build to produce this computer – jean Sep 13 '16 at 19:32
• I like how you just handwaved mining, because that's what you're going to be doing for the first thousand years or so. – Mazura Sep 14 '16 at 0:46
• I think I've played this game but I can't remember what it's called... I think eventually you got to build a rocket? – CJ Dennis Sep 14 '16 at 12:46
• @CJDennis this reminded me of Factorio – BgrWorker Mar 21 '17 at 13:37

I am assuming they have all the knowledge necessary readily accessible, in form of books or a magic tablet which does not run out of battery.

Your estimate is very low. It would take very very long time to get to a point to generate electricity. You need electronic machinery to build smaller electronics. Hell, even building a solder iron would take a very long time. Imagine melting and casting copper into stone to create wires. Also you might need to dig oil to make plastics, as some cables will definitely require insulation. Before you get to any of this, you will need tools to dig the ground. You will probably have to work with stone tools until you get to a point to make iron tools.

To sum up, you will go through whole of industrial revolution and then some, where thousands of engineers were working during that period to advance the field. I am not including scientists as your people have the knowledge. All in all, I would guess it would take around 50 years, probably more.

• 50 years to make an electronic computer from a scratch? Come on, get real. Talk in centuries! – Youstay Igo Sep 11 '16 at 20:20
• It took centuries first time, but there was time to research, fight wars and stuff. These people have the knowledge and 12 hours per day. I would say 50 is not far fetched. – Cem Kalyoncu Sep 11 '16 at 20:57
• I was told long ago that if you wiped out all stored human knowledge, the same generation would make it back to the internal combustion engine by only what is carried in our heads, and the next would be in position to attempt the electric motor. And that's w/o the magical ability to avoid human need. From there we can only guess. – Joshua Sep 11 '16 at 22:09
• I was really surprised the conclusion of this paragraph was 50 years. Sounds like you were arguing for centuries or millenia. – djechlin Sep 11 '16 at 22:53
• @joshua Electric motor is actually extremely simple. And it also produces energy when put in a turbine or something similar. I'd say it's actually easier to produce and use electrical power than internal combustion engines. As long as you know some background basics. – darthzejdr Sep 12 '16 at 10:41

There's only a few goals that are actual requirements and they have some low-hanging fruit:

• CPU, the only unique part that we want to move away from is mechanical switches. Galena and a steel wire can make a natural contact junction as can rusty and non-rusty metal supposedly. One person has made some relatively simple transistors, you could probably do something similar to that with the right chemical knowledge the materials used might even be simple to manufacture. I've heard current breaking down a dialectric can sometimes for junctions as well.
• Battery, Lead acid kind. You would need to build capacitors to smooth the power. Lead, like the rest of your metals, would be in raw form (or near enough) and sulfuric acid can be made with choice minerals and/or iron or platinum.
• Clock, you could make an inverter to provide a clock pulse or fashion a piece of quartz into a thin wafer (almost impossible but you only need to get lucky once).
• Display, a little harder but you could make a motor that signals you in binary. Possibly set up a grid of them for a screen.
• Input, crossed wires provide a keyboard input when they contact. You would build vertical and horizontal rows so they'te not touching and then pressing on a junction would complete the circuit.
• Memory, can be made as core memory from raw metals.

If you have raw materials instead of ores and a source of coal everything can be developed. Wax, paper, and electroplating can be used to build a circuit board if you feel it's neecssary. A solder iron could be made from a leather wrapped iron rod with solder being homemade (lead and tin alloy?). You can cold work your way to the tools and material shapes you need after you cast your initial hammer head and anvil from iron.

If you spent about a week each on:

• collecting materials
• a kiln for your athracite coal to melt the iron, melt glass to make a case for the battery, etc.
• make a bellows and other small tools
• cast the anvil and hammer
• making blueprints/designs
• possibly one for papermaking
• possibly spending an additional few months for collecting special materials.

...and a few months assembling the created parts;

It still seems pretty reasonable to accomplish within a year given perfect knowledge, the right raw materials, and a decent set of able-bodies.

• @Falco RGB is not necessary, for long time people where perfectly fine with monochrome displays. – MolbOrg Sep 12 '16 at 11:35
• Problem is analytics. You might manufacture a lot of the necessary things, but you won't know how to get them to the necessary specifications, because you have no idea e.g. how pure your base materials are and what's in them. It's an ever growing zoo of additional methods, which is why i said, heck, lets do it properly, see my above answer. ;-) – Karl Sep 12 '16 at 14:52
• I think you underestimate everything, 192 kb RAM required for windows 1 requires 4.6 million transistors. You need to put them in close proximity to be able run your machine at a reasonable speed. Also non of the materials are ready you need to dig, refine and melt them. You don't have industrial heaters, you have to build them. The "home-made" transistors require tons of modern machinery that you would need to work on for a very long time. You won't need factory lines and stuff but building a decent electronics lab takes a lot of equipment (I know from experience). – Cem Kalyoncu Sep 12 '16 at 19:26
• @CemKalyoncu More like 10 million, actually. Usually SRAM uses 6 transistors per bit (FLIPFLOP plus 2 for read/write), plus a few more for addressing every WORD. The soldering iron alone warrants construction of an electrical power plant, making miles of wiring no less. ;-) – Karl Sep 12 '16 at 19:47
• @immibis If you are prepared to solder ~50 million joints by hand (3-4 legs per transistor + plus a number of further connections for each, makes 50 man-years at 24/7 work, 1 min per joint), you can also heat the iron in a fire, right. You have no modern solder with flux in the bore, btw. ;-) – Karl Sep 12 '16 at 22:33

Rather than reinventing most of industrial civilization from scratch, I think your engineers will be better off thinking 'big' and developing a mechanically actuated machine from simple materials, like wood and fibers. I'll continue the trend of your generous assumptions and assume that your operators will operate the system perfectly, so you don't break any of the components. I'll also assume they carve and position everything perfectly so that you don't have to worry about the inevitable wear issues with moving/sliding wood parts.

It will be painfully slow to harvest all those resources with stone age technology, but still quite a bit faster than reinventing all of mining and metallurgy, let alone everything you need to manufacture semiconductor technology.

I'm imagining a huge machine with:

• A 'monitor' composed of patches of mechanically rotated dark/pale plant leaves as 'pixels'. You'd want to miniaturize this at least somewhat to make it usable, so this would probably be one of the most delicate parts of the mechanism. You'll probably also want to settle for a relatively low resolution. I'm imagining a vast array of sliding horizontal straight poles that adapt to vertical poles to rotate their corresponding pixels. You'd probably need to lay them out overlapping in 3 dimensions to get enough density.

• Mechanical switches that transform translation of a pole into either coupling or decoupling two other poles (so they 'transmit' only when the 'base' pole is actuated.) Basically you have a logic system composed of latching relays, with signals actuated by translation (and likely adapted/routed via rotation in various places.)

• Timing and power is supplied by people pedaling wooden wheels, with clever mechanical governance to transform into a clock cycle. This is where I doubt whether 20 people could supply enough power/speed to run 'fast enough' for your purposes. If it's not enough to run it 'live', you could store power using lifted weights or flywheels, so you operate the pedals for, say, a day, then you get a few minutes of runtime. Scale up as desired.

• For memory bits, you can leverage potential energy from gravity to store (literally lifted) bit states, with a read and refresh logic.

This is still a ridiculously massive engineering project, even if it is 'low-tech' from materials standpoint. Still, the basic components are all there to execute logical circuits and thereby build a fairly powerful computer. It's hard to estimate the labor involved, but I'd say you're probably still looking at a decade or more, just due to the sheer number of elements required. And that's assuming everything goes perfectly, with no mistakes made in manufacturing all these ideal parts by hand with stone tools.

• actually not bad, people in minecraft did processors – MolbOrg Sep 12 '16 at 2:14
• Clever, but such a device would not be fast enough to usefully run a Windows 1.0 type operating system, which requires an execution capacity in the low millions of instructions per second (the Intel 80486 of the era averaged 40 million instructions per second). By contrast, Alan Turning's codebreaker, which was state-of-the-art in the 1940s, managed 300 instructions per second. Your mechanical wood-and-leaf machine, properly greased with animal fat, would peak in the tens of instructions per second, before breaking. – Mark Micallef Sep 12 '16 at 3:42
• @MolbOrg It's a common excercise in any game that allows that, yes (remember Dwarf Fortress catputers? And waterputers? And a dozen of other designs? :P). But it usually hinges on something quite unrealistic - ignoring friction, energy etc. A redstone-like approach would be a lot closer to reality than a mechanical computer - but even then, the computer would be way too slow, with way too little memory to come close to running Windows 1.0. – Luaan Sep 12 '16 at 8:11
• @MolbOrg FPGA is already a highly integrated circuit - that makes a lot of difference. The switching speed is very high, and the distances between the transistors are very short. The only reason FPGA's can't rival "normal" CPUs in raw processing power is that their multi-purpose gates are (slightly) more complex than most of those in a common CPU. And Windows 1.0 is already a significant resource drain on a CPU made from TTL ICs (like one I've made), that only becomes worse when you replace TTLs with discrete transistors. The OP asks for Windows, not useful computation. – Luaan Sep 12 '16 at 11:42
• @MolbOrg They couldn't rival a chip specifically designed for bitcoin mining - that's my point. FPGA's are great in their flexibility, which allows you to reconfigure them to be more specific than general purpose CPUs, or just make a relatively cheap "chip". The problem with hyper-threading and out-of-order execution etc. is that it's incredibly complex. It's fine to make a homebrew CPU with 5000 transistors, especially if you have at least basic ICs for things like "16-bit AND"; making a multi-core, hyper-threaded, OOE CPU with three levels of cache... a lot more complex. – Luaan Sep 12 '16 at 12:22

See the links to building the “MPX-16” from scratch in 1983. This is sourcing the integrated circuits that were available when such machines were built by IBM. You can see the overall complexity and scale of the design.

Now you just need to build a silicon wafer “fab” and create the perfect crystal wafers… well, even if you supposed chipmaking could be scaled down to a home darkroom kind of thing, the industry needed to produce wafers is well beyond you little band.

That will be true even for the roughest semiconductor transisters; e.g. the stuff the Apollo mission used.

Any earlier technology would not be capable of running fast enough (as specified). Oh, and you want a CRT display to go with that? Again, we need industries, not a small party of individuals.

• Fun fact: we've currently lost the industry required to create new CRTs. We know how to make them (for now, provided copyright doesn't obliterate some of the finer important details over time before final release into the public domain) but have no actual way to make anything much more advanced than a simple deflected beam monochrome display. Even the phosphors are hard to get now, so image persistence / tube lifetime would be way off a good NOS tube. The mountain of continuous industry required to just maintain our way of life is truly and terrifyingly staggering. – madscientist159 Mar 31 '19 at 7:51
• For example, what about thomaselectronics.com/manufacturers ? – fabspro Mar 31 '19 at 13:22
• @fabspro I'm basically using apex CRTs like Trinitron as a canary here -- simpler CRTs may still be made, for now, but they will also stop production over time. CRT manufacture is already stepping backward in terms of complexity, and that will only accelerate. There are a few sources for this, but here's just one Google picked up venturebeat.com/2017/03/03/… "When the last major manufacturer stopped making CRTs, they sold ... to a Chinese company that couldn’t properly reproduce the winding procedure" – madscientist159 Apr 1 '19 at 7:02
• @fabspro Trinitron was also arguably, in terms of visual quality, the best CRT ever created. I used quite a few different models back in the day (as I suspect you have as well, from your comments :)) and the quality difference was very obvious. Even losing that apex manufacturing bodes poorly. – madscientist159 Apr 1 '19 at 21:14

This answer focuses on the computer goal, rather than the process to create the computer.

"As long as it gets the job done and the system is able to run at usable speeds, it could be built from any material and can be as big as it needs to be."

If you drop the requirement for "usable speed" then a simulation could be done using anything as memory markers. Lay out a massive grid on the ground to represent memory, and fill it with some kind of markers that mean 1 or 0, or use scratches in the soil.

Upshot: No industry required beyond feeding and housing and caring for the workers.

Downside: Time - the computer will run at hundreds-to-millions of seconds per cycle, rather than millions of cycles per second.

If they know what to do then probably two major problems comes in to mind

Transporting resources and gathering them is one problem

Second problem is low number of those peoples.

first is a problem because island is actually big, and if resources are scattered, it means pretty big distances. Getting-mining resources even is they are rich in quantities is not an easy task, and it gets not easier over time as demands will probably grow. Just moving 1 tonne at distance 10km with wheels etc without roads might occupy them all for day or more. But you have dig that tonne first, and it is not pure so it means to get 1 tonne of product they have haul more then 1 tonne of ore. So where resources are, in with form they are, surface of that island, distances might be bottleneck factors. Moving those resources across production complex is also a problem, gravity is a ... do not know the word, heavy might be.

As for second problem, number of peoples, they have not just replicate and fast-forward stone-steam-electricity as it was done and scale it for 20 people, but it have to be done in the way specially designed for those number of peoples, needs should never exceed 20 peoples doing something at the time. With no automation - at steam era, you have to have peoples almost for everything, they should work, watch, control, oil things, check that water gets in the boiler (not all systems which are used for that are reliable, and they tended to break or get our of regime their work) - so basically for moderate size steam machine it needs 2 people - one feed it, one watch it will not blow up and that it still rotates(kinda).

This get us to energy problem - how much energy can produce 20 humans with tools. At any point of that path stone-steam-electricity - there will be upper limit how much energy they might produce, in therms of power.

So whole process should not exceed their power production capabilities, their controlling capabilities. Glass making may need 24/7/365 watch - so 2 people out of whole process, and if there will be more then 9 such process at a time - they will run out of peoples.

And candidates for multiple points are chemical processes, there is a lot of chemistry involved in producing chips, not only for used chemistry in production, but used to produce that chemicals which are produced. Purity of chemicals might depend on bulk production, just because in big jar impurities form jar itself are less percentage then in small volume production. Some chemicals store not so well because of instability of them, impurity will grow over time - so it might be impossible to produce all needed stuff and make check-marks - or you produce 10 of them at once in short time or you produce none of them - just kinda exaduration, but who knows.

Making the process which leads to end result for those 20 peoples is more challenging then just replication of what we have done. I'm very interested in looking at their model just for brief moment, very exciting.

Sorry, but I'm not ready to estimate time, as have almost no glue what to do father then steam era. And not sure do they need 20m high refining columns for chemicals - if case they need to produce just one piece of that equipment which runs linux.

Steam era, they probably might achieve it pretty fast, less then a year, if they have no problems to know what to do and skills needed to do that. For real situation with people (not robots) I would say no way for 5y, but robot style people, may be may be, I consider it as possible. But this number is as good for me as 10y or 15y or 2y.
50y? do not think so, or they can do it in less time, or they just can't.

• It's a small island why would transporting resources be a problem? – NuWin Sep 12 '16 at 5:01
• Well, a watermill would be a lot better option for power than a steam boiler, if they have that option (I'm not sure if there's a chance for a useful river on a 10 are island; could it ever be rainy enough? :D). Even wind might be feasible, depending on local conditions. A stirling engine would require very little maintenance, and isn't too hard to build, though the energy output is rather low. – Luaan Sep 12 '16 at 8:21
• @NuWin would not say it is small it is something like 33x33km square or 35km diameter round island - it is like big city, but without roads. Building roads is power intensive task, which is not very useful if you do not transport goods in bulk, and is problem for lot of countries even today, with more then excess of peoples. Something like that are reasons to notice it as a problem. – MolbOrg Sep 12 '16 at 13:31
• @Luaan because of those options and uncertainties energy production is out of scope of my answer, at least at the moment. Steam engine is more example, which was used for manufacturing tasks, and we can find some data about that and know problems using them at production. Good river would be definitely good help, wind not so much(unreliable, not lot of power for the work, nice needs wood and good to go), good Stirling is hard to build complexity is on combustion level of engines, toys are bad examples, take look at few kW stirlings. I would rather go with turbines - but it is my preference. – MolbOrg Sep 12 '16 at 13:39

First of all, let's look at the requirements for Windows 1, it is an 8 bit computer with 385k RAM memory.

So, if you have space, it is possible. DIY computers are really not something difficult to make.

Displayed here is a board with transistors, that together makes a 4 bit computing processor.

In simplified terms, this is basically it, the challenge is to make $2^3$ times more efficient and smaller the oscillator (crystal frequency) can be increased and so small it fits in inside the area of a coin. But that wasn't one of the requirements.

Now the problem is, to make it out of something that looks like this:

It would take a lot of space.

A fast google of DIY RAM Memory shows that something similar could be made, fairly easy, where the true challenge is to make it small and modular. But given the means to extract the resources the area to build it, (I mean Boeing production size buildings), it could be done.

Regarding the running speed, I'm afraid I'm not experienced enough in that area, to know what speed it would run but it would largely be controlled by the switching capabilities of the transistor but I could not find the datasheet of the "first transistor" but a general purpose transistor has a switching capability of about 300 MHz.

• RAM is actually quite a big deal. The problem isn't making a bit of memory (SRAM is very simple, really). The problem is making a lot of memory - even Windows 1.0 required almost 200 kiB of memory. That's huge. – Luaan Sep 12 '16 at 8:25
• The problem is that big size means slow operation. Making a CPU fast enough for Windows 1.0 isn't too big of a deal, even from simple-to-make TTL chips; it might even be doable with discrete components. However, making memory fast enough for Windows 1.0 is a problem - exactly because of the size required. Note how the silicon area of any modernish computer is dominated by memory (DRAM and SRAM). The propagation of electricity is too slow to maintain the speed you need. – Luaan Sep 12 '16 at 9:06
• @Molborg Yes, memory was faster than CPUs for a time (this was still true at the time of the Windows 1.0). That's why I'm noting that in Magic Mouse' computer, it would already be the other way around. And I wasn't talking about frequency on the cable, but on the CPU - note how I'm talking about latency in the communication. If the CPU doesn't have out-of-order processing, it must wait on memory on each memory operation, which puts a limit on how fast the memory must be not to impede the CPU. Not a big deal for calculation, very much a big deal for Windows. – Luaan Sep 12 '16 at 12:03
• No, you're missing my point. Just the electricity lag would be enough to discount anything above ~10 MHz (synchronous, of course - there's a lot more you can do with asynchronous, but that makes the whole system much more complicated). That's before it gets to the transistors, and assuming optimal addressing (in practice, you would need much more than just the 10 meters of wiring to address anything in the huge memory core). And how much memory could you fit in a cube of 10m? If it takes a cubic cm (generous) to store one bit, you get about 1 MiB - ignoring all the addressing circuits etc. – Luaan Sep 12 '16 at 12:09
• @Luaan you are confusing old components with making something "new from scratch" with primitive tools, we have advanced computers who can calculate and design the system, beforehand. "The group would know exactly how to find and assemble any items involved in the process of creating the machine." So the problem here is finding components in nature and assemble it, it says nowhere that they could not have CAD and or calculated strategy with them. – Magic-Mouse Sep 12 '16 at 12:15

Putting things together is information. The arrangement of metal that distinguishes a pile of coal and rusted iron from tempered steel is information.

The way we usually do this is through crude processes that generate gradients that rearrange the locations of the atoms in a favorable way, and we iteratively move towards the arrangement of matter we want.

This involves the application of energy to generate a entropic gradient of the right shape, which is the only way we know how to mass re-arrange atoms into a new form.

But it is just information. Some energy needs to be added to get some states from others, but the net energy difference after processing tends to be far far less than the energy used -- most of the energy is leaked off as heat, not captured.

This leaked heat is entropy -- loose information. The ordered energy we use to induce the changes stuffs some of the information into the new structure, and the vast majority leaks off as heat.

If all the humans know is our current crude methods of infusing stuff with the structure we want, then they'd basically have to reinvent industrial civilization. Time would be measured in generations not years, as they would have to breed a population sufficient to manage the industrial civilization needed.

If they instead had all of the information they need to make it efficiently, and the ability to exactly use that information, they could literally walk around and hit things perfectly with hammers and cause them to reassemble into the shape they need.

Remember, all that is required to uncrack an egg that fell off a table is the exact right set of taps, impacts and sounds. It is the lack of information, and the difficulty in doing the actions exactly (low energy, extreme precision) that makes this impossible. The easiest way for a human to uncrack an egg is to feed the cracked egg to a hen, or compost it and feed the food you grow to the hen.

This level of knowledge and precision in action is far beyond what any human could do, but you did say the had exact knowledge on how to do it. And mere mortals have social issues. Clearly you are not talking about mere mortals, as they have no social issues.

So if each woman produce 6 children per generation, 3 of which are women, and the population is half men, after 10 generations you have ~120,000 people. After 20 generations you have ~7 billion people. I'd expect it to be somewhere in that interval without perfect information.

With perfect information, they walk around tapping things and those things reform into the exact ingredients they need. They touch them together, tap them, and they bond together. Their actions look more like magic than industry.

• Mike Resnick "Oracle" - probably not the level of knowledge OP is talking about. Any way It takes time and do not negate laws of physics. – MolbOrg Sep 14 '16 at 2:51
• An interesting take on the problem! I suspect our (super-intelligent) humans would be intensely frustrated: They would know exactly what "taps" are needed in what sequence to build their device, but lack the sensory precision to place and direct those taps as needed, and they would lack the muscular precision to execute the necessary taps. TL;DR perfect information is useless without the ability of perfect execution. – Spike0xff Sep 14 '16 at 14:22
• @Spike0xff Well, you tap. And it doesn't work. So you tap somewhere else. With perfect information, you can find the tap that would most move the system towards the state you want given the accuracy of your tapping tool (the first step, of course, would be building a more accurate tapping tool). The humans spend decades making seemingly useless devices, then start tapping things with them, and get a computer a year later. :) – Yakk Sep 14 '16 at 14:26

They can do it in seconds.

"The group would know exactly how to find and assemble any items involved in the process of creating the machine."

If their brain is programmed with what ever knowledge they need for making anything. Just make 1 of the 20 people the computer. The only real thing required will be a language to interact with the "computer" so that any one of the 19 other people using it can figure out what is happening, and since they already know everything they need to do, they can just do it. The "computer" can encode any information it wants into sound and the 19 people can decode the sound in their head into windows 1.0 UI. A person should be able to process any high level UI level command within a reasonable time with training, and since these people know everything they need, they should be able to do it.

If above is not valid because they didn't create the machine. Then it would take 9 months for 2 of them to biologically create a new machine and then a couple years of training to get the machine programmed correctly.

• If the case had been that people had to make a "computing machine" I would disagree, but the question just asks for a "computer system"; ergo this is the best answer. Back in the day we even used to have jobs for people whose titles was "computer", they were responsible for calculating things. – GrinningX Nov 2 '17 at 23:48
• I think you underestimated the time needed to create a new machine. With only a couple of years of programming, the system would be very high maintenance. – zakinster Apr 2 '19 at 7:09

For the sake of completeness, let me give you the perspective of an Electrical Engineer who once designed integrated circuits and knows a bit about the history of computing.

Given the limitations proposed by the OP, it is IMPOSSIBLE to develop the technology necessary to build a computer running Windows 1.0. There IS NO LENGTH OF TIME that will change that. The mountain of technology is so large, the developmental basics so great, and the knowledge so specialized at thousands of points along the developmental path, that it's impossible.

Sorry

(I upvoted Karl's answer because it was well thought out, even though he's not familiar with some of the core technolgies... but as much as I liked the answer, this one needed to be provided.)

Any piece of technology that is newly developed is, at its time, the pinnacle, or the sum of everything and everybody that went on before, within the "light cone" of the involved people or logistic processes.

In the stone age, presumably, this kind of "light cone" was relatively slim - individual tribes probably re-invented the same technology (sharp stones fixed to a stick) relatively independently from first principles. Over time, travelers or raiders brought ideas into circulation; culture developed. By the time of, say, the old Greeks (Alexander), Chinese, the Roman Empire etc., the light cone was - for the most advanced bits and pieces of their time - probably their core city state (and then shone outwards towards the "barbaric" regions they occupied).

Fast forward today: the "light cone" that goes into our current products (computers, etc. - even in the 1980's) is arguably incredibly large. It is so large that we are incapable of, for example, determining the true economic footprint of most of our wares - we are not even able to measure/document the logistic and product chains going into T-Shirts, never mind electronics.

Sure, a lot of our complexity comes from the fact that we not only produce one piece of everything, but do mass production; so on the face of it it may seem you can shave a lot off if you don't need that. But this argument does not work for your question. You need mass production for a lot of what you are doing - for example even your single Windows 1 capable machine needs lots and lots of repeated pieces (resistors, whatever). So even for a component only one or steps removed from your end product you already need to figure out the prerequisites to mass manufacture stuff. Surely not on the scale we have on Earth.

You might argue that the knowledge of the goal and all the principles along the way would change the time line significantly; and in the ideal world (nothing gets forgotten etc.) it probably would. Still, as long as you have humans in the loop, you still need a lot of those, since every one is simply incapable of cramming arbitrarily much applied knowledge into their head.

Finally, it is not like our technology was created in a completely planned fashion - we have demonstrated that this goes horribly wrong. The fact that we have (many) millions of people working in science and industry also enabled the evolution-like approach we have today - we are trying and failing a lot, and this is inherently important in the system. This kind of trial and error is a feature, not a bug. You'd arguably need to replicate it on your island - i.e., for non-trivial topics, you'd need multiple teams trying to achieve the same goal at the same time.

And then we get to the crux: all those people need to be born, grow up, be educated, fed, clothed... and voila, there's our whole Earth economy as a non-optional pre-requisite.

So, as you are limiting your time scale to well within a single generation, it's not possible for systemic reasons (not just scale).