Okay, hypothetical situation: you are on a world much like this one, just without any trace of human civilization. You have unlimited manpower, knowledge of every aspect of our technology, and survival is not an issue. How long would it take you to build the computer sitting on my desk, using only the surrounding resources? How many different levels of tools would you have to go through to reach the necessary level of technology? (i.e. wood tools lead to bone tools lead to stone tools lead to bronze tools, etc.) How many of the levels of technology that humanity went through could have been skipped alltogether had we known of something better?
We had a thread similar to this, but it got lost in a server crash a while back. We weren’t trying to figure out how to build a PC, but a car, so it’s a parallel situation, not totally identical, of course.
If you’ve got the necessary knowledge, you don’t have to work your way through all the various levels of technology. Starting out with stone or bone tools, you can skip bronze and go directly to iron and steel. From that point, it’s a fairly simple matter of building the necessary equipment that you’ll need.
To elaborate, using stone tools, you can build yourself a machine shop of the same quality that they would have had in the late 1800s quite easily. These books detail much of what you’d need to do. Now that you’ve got a machine shop, you have to decide what you want to build next. Do you want to go for an electric generating plant? Or do you want to build more accurate tools so that your generating plant will operate more efficiently? If you go to the electric plant, you can then crank out electric motors, electrify your machine shop, and then begin to work much more rapidly. My best guess is that it would probably take you ten years or so.
The hard part of technological advancement is in knowing what to do, if you’ve got that much figured out, then doing it, is much quicker.
Oh. That’s not the impression that I got. I seem to remember most people thinking it couldn’t be done in a lifetime, if at all. The infrastructure needed is just too great.
There were at least two threads on this, both in General Questions, and I think one of them is probably still around. Search is wonky now, though…
It was a bit of a split decision, IIRC. Most of us who’ve had some experience in the metal working trades believed that it could be done in about a decade or so, with others who haven’t worked in the trade believing that it wasn’t really possible.
It has a lot of variables to consider, which makes the final outcome difficult to ascertain. For example, if it were me trying to build a car, I’d skip the complexities of an internal combustion engine and go with steam. Fewer parts to make, and a lot more simple to build.
If your goal is to build a PC as quick as possible, and you’ve got a powerful source of running water nearby, you can skip the electrification of your machine shop and run it off of water wheels. Not the most efficent way to run it, of course, but it saves you from having to build bunches of electric motors just to run your machines. You can save those for the equipment necessary to building your PC.
If your goal is to build a PC, then you don’t have to waste time building other things that are possible. You don’t even have to go through all the various steps of building an 8081 machine to get to a Pentium class computer. You already know how to build a Pentium class machine, you just lack the tools to do so. Once you have those, the rest is a piece of cake.
I’d say the toughest part to make would be the Integrated Circuits (or computer “chips”). Those would require a very high level of technology along with the accompanying “clean rooms”, etc.
very true. Also, the equipment used to manufacture large computer chips is highly automated using – you guessed it – more computer chips. So you’ll have to design some simpler computers before you have the capability to design the box sitting on your desk.
How necessary is all of that automation? How much of it is merely a means of labor saving? Since the OP gives us unlimited amounts of manpower, if it’s purely labor saving, then we can chuck it all.
I doubt that it is all labor saving, of course. But just because we do it all with automated machines now, doesn’t mean that it has to be that way. Remember, while you’ve got guys using stone tools to build a machine shop, you can have guys figuring out ways of building a PC fairly rapidly while skipping steps.
Take the etching, for example. Currently that’s done by a machine, before the silicon is dipped in the various chemicals which create the various layers, etc. of the chip. It’s done by machine because the work is so small and everything has to be so straight that a human couldn’t do it freehand. However, it might be possible to make something like a repeatograph (or whatever those things were called that allowed you to trace a drawing and have it come out larger or smaller depending upon how you set them up) which traces an engraved line on a plate that represents the circuit path for the chip, the repeatograph is designed to reduce it down to the necessary size for the chip. (Something like that is probably how they designed the first one, I’d imagine.) The chip could then be dunked in the chemical bath.
Not saying that this would work, or would let you get a Pentium chip out of it, but it’s a possiblity that isn’t considered today because it’s not necessary.
-
-
- Regarding “making a computer from sand”: in my highschool electroincs class days I got the bright idea of electronic crayons. That is to say, crayons that would draw in a substance with a particular electronic property–say, one would be “regular conductive”, one could be “P-crystalline”, another could be “N-crystalline”, and you could have different resistive crayons that would produce so-many-ohms per inch of linear distance. An insulating one would allow you to overlap or cross conductive paths over each other without connecting them electrically. Couldn’t figure out how to do coils at all, but supposed that capacitance could have been two conductive layers, separated by a “resistive” or “insulating” layer. With all these you could then draw out an electronic circuit on a piece of paper and apply electricity to it, and have it actually function, because we had to spend quite a while learning to draw schematics and that got rather boring. …The teachers there thought that wax with enough powdered silver or other conductive metal would conduct at least somewhat, but had no clue about how to make the rest of it function, and had no idea who to even ask. And this was before the internet really took off, so there was no easy way to bother experts scattered across the globe. And I never got far enough in chemistry to learn myself, and forgot most of what I passed tests with… …it would be kindof neat to see how large something useful would end up needing to be drawn though, like maybe a 286 cpu? anyone?
-
As someone who designs computer stuff for a living I have to say that the amount of labor saved by the various cad programs is enormous.
In the old days of designing electronics people drew the layout by hand on a large sheet of paper and they took photos of this to mask the various masks needed to manufacture the chips.
A modern processor has about 40,000,000 transistors. I don not think that it is feasible to get the masks for that without computer help. Even a large team of people will not be able to get all that done with out error.
Intel doesn’t always seem to get it right, either as it seems like every time they introduce a new line of chips they make an announcement sometime thereafter stating that they goofed and the chips are prone to errors if you’re doing really complex tasks with them.
So, another variable to add into this is how “good” does the PC have to be? Does it have to be a Pentium class machine, or an SGI or a Fujitsu Numerical Windtunnel or what? Does it have to work 100% correctly, or are there acceptable levels of errors allowed? What’s it going to be used for? If it’s just number crunching, then one could probably build a giant difference engine and be done with it.
Building a PC akin to what is sitting on your desks and a basic car are two somewhat different technological tasks. Unlike Tuckerfan’s assertion that you can “just build it” if you know what you are doing with a car, with high density IC manufacture each successive level of manufacturing density needs the hardware and software tools produced by the previous generation to move ahead. You really can’t just “skip” a level of integration.
To manufacture this insanely complex stuff you need large teams of highly trained people. Training a large population of engineers to even begin to accomplish these tasks would easily take the better part of a lifetime if you were starting from “scratch” with an agrarian population.
Well, those membrane keyboards on appliances are printed from silver ink.
Down the hall here at the UW there are people testing polymer transistors (FETs made of droplets of black ink-stuff dribbled onto the substrate electrodes.)
As for starting with sand; what most people don’t realize is that the pace of progress is tremendously slowed by simple disbelief. Come up with any radical idea (such as integrating several circuit components onto silicon, or even ideas such as vacuum tubes,) and nobody jumps on it. The inventors even have to fight to keep the ideas from being rejected. Even the transistor itself was invented by physicist JE Lilienfeld around 1925, but was ignored for another 20 years.
Remove the blockage caused by normal human disbelief, fill a world with people who aren’t threatened by new ideas, and what would happen? Maybe everyone would waste time on crazy ideas which aren’t workable. Or maybe, if our present level of scientific conservatism is way “overkill”, then with some reduced conservatism, the pace of scientific progress would increase by a factor of 10 or 100.
How so? Think about it for a moment, if a 286 machine is properly programmed it can produce high quality computer images, it just takes it a hell of a long time to do it, where as a Pentium PC can do it in next to no time. The question is: Which is longer the render time to have a 286 machine design a 686 machine or using a 286 to design a 386 so that you can then build it, so you can use it to design a 486 machine, and so on?
**
The OP isn’t quite clear if you are the only one with the knowledge or if everyone is. If you’re the only one, then it is going to take ages, but if everyone has the knowledge to do their task, then it won’t take quite so long.
Well, IC manufacturing is what I got my BS degree in, so here are some of the major issues I see:
-
Ability to produce single crystal silicon with a known level of impurities, both wanted and unwanted. The wafers have to be extremely flat for imaging purposes (fractions of a micrometer, if memory serves).
-
For imaging, multiple things have to fall into place. First, near perfect lenses. I mean, the irregularities of these lenses across the lens field puts the best photography lenses to shame. Second, a suitable photoresist (polymer+photoactive compound) is needed. The photoresist has to be able to stand up to the required processing, whether it is a liquid chemical etch or an etch using some type of plasma process. The photoactive compound has to react with light in the right wavelength. When further processing is finished, the photoresist has to be easily removed. After exposure, the development chemical should not be an organic solvent like acetone, for the safety of the workers and the environment. Third, the imaging equipment can’t be done by hand. At school, I used a stepper (step and repeat–image one chip at a time) dating from the early 80’s, and it still had tolerances on its stage in the micrometers. To get repeatable linewidths, the exposure time has to be automated. Overexposure is just as bad as underexposure. Also, for best results, the photoresist application should be automated for repeatability, though not totally necessary with older photoresists such as novolac-DNQ. With new chemically amplified resists, until the wafers are developed, they are particularily suceptible to contamination by chemicals in the air such as ammonia.
-
Suitable materials must be found. For example, there is no gas that will etch copper that has a gasous endproduct. That is why copper processing is a relatively new innovation. Also, if copper gets into the transistors themselves, the performance of the product decreases dramatically. Another reason is silicon has an easily formed oxide that is easy to work with, making it the main semiconductor of choice.
-
For 1 micrometer processing (probably cutting edge a decade ago, but don’t quote me on that), some steps still needed high tolerances. For example, masks were and are made by scanning an electron beam across a chrome covered quartz plate to react with a photoresist. Electron microscopes are often used for quality control in one way or another. If ion implantation (basically shoot individual atoms of desired impurity at the wafer) is used, the depth and dose of the implant has to be controlled tightly. The gate etch is the most critical, as it determines the transistor’s speed. The smaller the gate is, the faster the transistor is. When I was at Intel, that was the step my group monitored most closely.
-
Safety is an extremely important issue in making computer chips. There are gases such as silane that burn at room temperature. Hydrofluoric acid attacks the calcium and phosphorus in your bones, but you don’t immediately feel any pain when contaminated. It also looks like water and is the most reactive acid that exists. Liquid chemicals are heated to high temperatures. For example, sulpuric acid + unstabilized hydrogen perioxide are heated to about 150C. To create the various materials, there are many associated hazards. Furnaces are routinely at temperatures 900C and greater, filled with highly reactive gases such as hydrogen, oxygen, dichlorosilane (which has the scariest warning label I have ever seen). Vacuum is routinely utilized in a number of different types of process steps. One tool I used had its bell jar wrapped in duct tape for safety, in case it imploded. The ion implanter can easily have a voltage of 10,000 volts.
-
A high level of training is required to use these tools safely. When I started my co-op at Intel, I had a 40 hour week devoted to training, and that only allowed me on the cleanroom floor and watch, not operate a tool or repair it. For electrical work, you need more training, and each tool you use, you need to be certified for what you can do (i.e. operate, prevenative maintanance at specific intervals, etc.) On top of that, you need engineers to support the process when things go wrong outside the ordinary or to develop the next generation of process technology.
-
The level of cleanliness the chips are exposed to is critical. For example, cigarette smoke particles are the perfect size to make your half finished wafers trash. I’ve looked over liquid chemical pureness levels, and often times the impurity levels are in the parts per million range, even to parts per billion. Front end (forming the transistor) and back end (forming the metal lines) processes have to be kept totally separate, as the metal can kill the transistors, particularily with copper and gold. At one point, the fab I worked at had a separate changing room for those who worked with copper, to prevent contamination.
Well, those are the major issues I can think of that would make creating computer chips like those used now (I mean CMOS, which came into its own around 1980) difficult if you had to start at scratch. There are other, less pressing issues, such as yield management, but I have little idea what those guys do. I either worked on specific steps in a process, or a single process, beginning to end. I also did not mention what is required to make manufacturing cost effective. I mean, you have to do something when a single tool can cost several million dollars and have a year’s lead time for ordering (stepper, in this case.)
Please let me know if I did not explain something well enough. I know I tend to slip into jargon and assume too high a level of proficiency.
—Samantha, with a BS degree in Microelectronic Engineering from Rochester Institute of Technology
PookaKitten, enviromental concerns may not be that big of a deal. If it’s a pristine planet, a little contaimination (while in the grand scheme of things a bad thing, of course) wouldn’t be that much of an issue, unless of course, you’re massproducing the things. A one off job wouldn’t be that big of an issue.
Here’s a question: What about the display? Full color or black and white? A black and white CRT would be easier to produce than a color one.
Organic solvents are worse than you think. They don’t break down like the newer ones (TMAH, aka tetramethyl ammonium hydroxide), and build up in your fatty tissues and liver. That’s on top of the environmental problems. One of my professors was an expert witness in several trials involving people who got sick as a result of working extensively with these solvents in the early days. He’s apparently one of the best known people in the lithography field, particularily the chemistry end of things. There was a reason I used the oxygen plasma asher to remove photoresist, once we had one, instead of acetone. That and laziness. As one of my other profs said, “A good engineer is a lazy engineer.”
—Samantha
Organic solvents are worse than you think. They don’t break down like the newer ones (TMAH, aka tetramethyl ammonium hydroxide), and build up in your fatty tissues and liver. That’s on top of the environmental problems. One of my professors was an expert witness in several trials involving people who got sick as a result of working extensively with these solvents in the early days. He’s apparently one of the best known people in the lithography field, particularily the chemistry end of things. There was a reason I used the oxygen plasma asher to remove photoresist, once we had one, instead of acetone. That and laziness. As one of my other profs said, “A good engineer is a lazy engineer.”
—Samantha
Is the entire planet solely and whole heartedly dedicated to producing this ONE PC? If not, then economics come into play as well. Modern fabs cost a couple of billion dollars each. The only way to get enough money to build the latest generation of chips is to sell many millions of the previous generation. Thus, if you want to do this properly, you actually need to get an economy going which requires significant amounts of computing power.
Given an unlimited amount of man-power and full knowledge of the technology involved, why would you need to make an older convential PC in order to make a new PC? It’d seem to me that if you knew the design of all the eletronics necessary to put together a modern PC, you could forego most miniaturization until working on the final product and just use really bulky machines capable of producing miniaturized components. Who cares if the componentry necessary for running an IC-producing machine occupies several city blocks–the OP asked how fast, not how fast while maximizing efficiency.
I think you could skip a lot more historical technological steps than most people think.
The original intent of the scenario was indeed that the entire world be dedicated to building this one PC, and all the components thereof, with no other concerns at all. All those involved known exactly what they need to know, and they don’t worry about things like food or economics.