5
$\begingroup$

Incoming generic apocalypse, a bunch of people with low budget tries to quickly download publicly accessible blueprints, so in a few decades or centuries someone would be able produce again computer hardware without need of reinventing the wheel.

Instead of simple "just pick some 10-20 years old technology", I started to wonder whether blueprints would have uniform quality across board? I mean for example whether hypothetically RAM would be straightforward to replicate, while modern processor designs were too well protected as trade secrets.

So how modern should be realistically the designs? Secondly, should their quality be similar across the board, or are there components which blueprints would be clearly easier / harder to get, in a way that are likely to affect the future design?

I'm NOT asking:

  • how to store data for potentially long time
  • how to build a microchip factory with low population and tricky access to rare Earth elements
  • about software (except maybe firmware, if it would be a bottleneck)
  • how to build the most basic components (but what in long run would be the top components that one could build without effectively designing them from scratch)
$\endgroup$
  • 2
    $\begingroup$ I'm not fully certain that I understand what you mean by "computer blueprints". Actual blueprints are not used in computer manufacture, and I cannot think of any engineering artifact actually used in computer manufacture which would have a similar function with the blueprints used in the construction industry or in the mechanical industry. The point being that the manufacture of computer parts is high technology, almost fully automated, and if one can make the machines to make the parts one doesn't actually need to copy anything. $\endgroup$ – AlexP Aug 3 at 13:11
  • 2
    $\begingroup$ Perhaps by "blueprints" you are simply referring to data and instructions... things whose whole value is the information it caries rather than its material properties? $\endgroup$ – Cort Ammon Aug 3 at 15:45
  • $\begingroup$ If you don't do it within the useful lifetime of somebody who has done it previously and can show you all the tricks and skills that aren't on the recipe/blueprint/data, then you will need to fully re-invent that particular wheel. $\endgroup$ – user535733 Aug 6 at 1:34
9
$\begingroup$

Chip makers are paranoid when it comes to protecting their designs and processes.

A bunch of people with low budget has no chance to find on the internet some recent design which makes sense (on line scammers are another story).

That said, you cannot mix in the same computer a CPU from 2010 with a RAM from 1980, because they would be blind and deaf toward each other: different buses, different clocks, it would be a hell of a job to integrate them, and again, your people have low budget and no tools.

Last but not least, chipmaking is one of the most advanced technologies today. Just having the blueprints would not enable you to make a chip, unless you have also the means to produce high purity silicon crystals, slice and polish them and process them in the proper way using high purity dopants and metals.

In a post apocalyptic world you would be like if you managed to build a replica of the first transistor.

first transistor

A modern microchip contains gazillions of those.

$\endgroup$
  • $\begingroup$ It would be nice to know the scale of those photos. $\endgroup$ – Ray Butterworth Aug 3 at 13:12
  • $\begingroup$ @RayButterworth, in the linked page there is a photo of the researchers who made it while they are looking at it. It gives a sense of the dimensions $\endgroup$ – L.Dutch Aug 3 at 13:30
7
$\begingroup$

Modern "blueprints" are a closely guarded secret. They are also enormously complex and basically useless. By the time you can use them, you have more than enough development in place to recreate them. It would be better to develop some basic computing capabilities so that you could recreate a VHDL synthesizer. Then you can take VHDL files and synthesize them for the particular process your people invent (it won't have the same timing properties as the process it's imitating, so having a computer help with that will be a huge perk).

However, perhaps of interest might be the Intel 4004. It's a CPU from the 1970s with 3000 transistors. There are also pictures of its masks available online which Intel released during an anniversary celebration.

4004 mask

$\endgroup$
3
$\begingroup$

As others have pointed out, you will not get contemporary masks and all that without extensive effort. But that might not be useful anyway.

Short answer: you can not just produce everything again after an apocalypse

The manufacturing process is massively complex and if you do not have EVERYTHING needed, including maintenance knowledge for all the required machines that have been falling apart for 20 years, it is borderline impossible to recreate the process and just continue production as if nothing ever happened.

First of all:

What "blueprints" are we talking?

You have to distinguish between what type of "blueprints" we are talking about:

  1. Hardware description code that is the very basis for any chips functionality (in HDLs, Hardware-description languages, e.g. VHDL, Verilog)
  2. Masks for production
  3. A 'script' that details ALL the steps ('turns') of a chips production (including which masks to use for which 'turn')

The first one would allow you to start a new production from scratch as soon as you have access to higher calculatory power AND the synthesis tools to create masks and all that.

Second and third are necessary if one were to continue production, but that require you have access to a functional (and actually clean) clean room with all the machines fully operational. That is a highly unrealistic scenario.

Another thing to consider is the reason for Intels "Copy Exactly!" policy. Every clean room in the world (for a specific technology node) looks exactly identical. Identical setup, identical machines, identical everything. The reason is that Intel worked out that recreating a process that worked in one type of lab in another lab can take several years, depending on the amount of differences in machines and materials used. You can not simply take a 'script' and run it on different machines. It just does not work. You will never get nanometer precision with other machines and identity of chemical processes with even slightly different materials (as in acids that have minimal contamination, etc.).

It is basically impossible and such a recreation would probably take several decades without all the former experts helping you, even if you had all masks and a script.

So the only thing realistic is starting from scratch. But how?

You can NOT just produce smaller technology nodes with the HDL-code for the chips

The reason is that a compilation (code to logical functionality) and especially a synthesis (logical functionality to transistor mapping and wiring on a chip) require massive computational power and high-end software that is (to my knowledge) exclusively temporarily licensed. So even if you had the power and the software, after the apocalypse all the licenses will have run out and you will not be able to use the tools. You'd need to get your hands on the codebase for all the tools or get a developer version of the software that works indepentently of any license.

So also recreating 12nm technology is unrealistic even if you had everything

The only option left is to start the entire industry from scratch.

What would be needed to rebuild the semiconductor industry from scratch?

If you had secured all available knowledge about semiconductor chip production, computer engineering, and preferably a few engineers (of many differenty fields), mathematicians, chemists and physicists related to these fields, you will get a lot further than with any blueprints and scripts imagineable.

There are many things to be done:

  1. Recreating simple processes as used for very early processors
  2. Develop machines to allow for more complex processes
  3. Develop software to compile, simulate and synthesize processor designs
  4. Use the processors and machines you produced to compute more complex designs
  5. repeat 2., 3. and 4. indefinitely

With all the knowledge secured on what concepts are needed for chip design and production you will probably make very good progress. (depending on how many surviving engineers, etc. you have, and how much people have resources to do non-survival things post-apocalypse.)

The more early processor designs (including documentation) you have secured the quicker the production of more complex ones can start over again.


TL;DR

Recreating the production process with pre-apocalypse technology is very unrealistic, but the more knowledge and expertise you secured the quicker you will be able to recreate chip production capabilities after the apocalypse.

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.