Info 22-year old chip designer / inventor

[igor_kavinski]

This 22-year-old builds chips in his parents’ garage | Ars Technica

Zeloof started down the path to making his own chips as a high school junior, in 2016. He was impressed by YouTube videos from inventor and entrepreneur Jeri Ellsworth in which she made her own, thumb-sized transistors, in a process that included templates cut from vinyl decals and a bottle of rust stain remover. Zeloof set out to replicate Ellsworth’s project and take what to him seemed a logical next step: going from lone transistors to integrated circuits, a jump that historically took about a decade.

This guy is a super genius!

Haven't read anything this fun in quite a while.

 

[Hitman928]

Super genius is a bit much. It's much easier to replicate what others have done given you have the means to buy the necessary equipment as he did and access to all their papers of how they did it, which he does. It's very impressive to be this motivated and resourceful at his age, but ultimately it's a hobby that won't really lead him anywhere unless he wants to use it as a really cool resume booster if he gets the actual degrees he needs to make a career in device physics / chemistry. It's like that guy who used breadboards and discrete circuits to build a garage sized working processor. A neat hobby to kill a lot of time and learn some things, but ultimately you hit a hard wall of what you can do with something like this long before making anything interesting or useful.

 

[DigDog]

^what this guy said.
Looks good on resume', nothing else.

 

[igor_kavinski]

He's only 22, and he did all what used to be done by teams of people. I'm excited to see what he will come up with as he continues to refine his craft. By the way, not everyone can just read research papers and replicate what the authors describe. He does have an above average IQ.

 

[Hitman928]

He's only 22, and he did all what used to be done by teams of people. I'm excited to see what he will come up with as he continues to refine his craft. By the way, not everyone can just read research papers and replicate what the authors describe. He does have an above average IQ.

He's definitely intelligent and as I said, super motivated for his age, but he's going to find out real quick that there's not much he can do with the setup that anyone will really care about outside of the novelty of it, if he hasn't figured that out already.

 

[igor_kavinski]

What excites me is that he is a free spirit. He doesn't have the burden of having to deliver a product or working for a company under a boss with pressure to perform. He's doing this for fun and during his fooling around with materials, he could make some game changing discovery.

 

[Hitman928]

What excites me is that he is a free spirit. He doesn't have the burden of having to deliver a product or working for a company under a boss with pressure to perform. He's doing this for fun and during his fooling around with materials, he could make some game changing discovery.

Not gonna happen.

 

[Pohemi]

He's doing this for fun and during his fooling around with materials, he could make some game changing discovery.

He's recreating what engineers did 50 years ago, with 50 year old lab equipment.

It's pretty cool (and somewhat unexpected) that he was able to do it in a garage. How much further do you think he's capable of advancing the chip tech that he's assembling, though?

He's doing (what is now) basic chip production. He isn't a material engineer who found a new replacement for silicon. He has yet to produce a microprocessor capable of actual tasks (like 1+1).

 

[mindless1]

Frankly I don't think it takes that much intelligence and rather, somewhat of a ceiling to the intelligence a person could have, to reinvent the wheel, knowing the result, even if he devotes a lifetime to it (a fab costs more than a lifetime of income, if you are earning nothing because busy reinventing the wheel) could never be as good as a ready made product you can buy for the cost of lunch. Granted, you can get fab time, but ultimately rather than the discrete components he should be focusing on purpose specific architecture, design it then let someone else build.

It's only worth pursuing if he thinks he has some new angle, some new tech that isn't being leveraged in existing products. Of course then there is the hobby angle. There are a lot worse hobbies, but also a lot better ones.

 

[moinmoin]

Reminds me of Ben Eater's videos:

I'll say though that I can appreciate anybody doing such projects the hard way. Earlier hardware was way more approachable and malleable compared to nowadays' opaque and unfixable stuff which led to far fewer people toying with and hacking them. Recreating (some of) the basics (and publicly documenting the process) may well be the only way to keep the inner workings of tech approachable to, well not really the masses, but a little more than an already tiny minority.

 

[Red Squirrel]

That's pretty crazy awesome. Have to give him credit, most people would not even know where to start in making a regular PCB, let alone a semiconductor chip.

 

[eek2121]

Super genius is a bit much. It's much easier to replicate what others have done given you have the means to buy the necessary equipment as he did and access to all their papers of how they did it, which he does. It's very impressive to be this motivated and resourceful at his age, but ultimately it's a hobby that won't really lead him anywhere unless he wants to use it as a really cool resume booster if he gets the actual degrees he needs to make a career in device physics / chemistry. It's like that guy who used breadboards and discrete circuits to build a garage sized working processor. A neat hobby to kill a lot of time and learn some things, but ultimately you hit a hard wall of what you can do with something like this long before making anything interesting or useful.

I read this article the other day. My initial thought was that I wish companies would release better, cheaper tools so that chip design/development is more accessible. I bet if a company could hit the right price, they'd profit pretty handsomely. Sure, it's nowhere close to competitive with Intel, TSMC, etc. However, if you want to build a simple chip, you no longer have to go to one of just a small handful of fabs to do it. Even if you are just prototyping...

 

[IntelUser2000]

However, if you want to build a simple chip, you no longer have to go to one of just a small handful of fabs to do it. Even if you are just prototyping...

No real need to do this, because unless you want very specific functionality unique to your needs, you can already buy chips that has the functionality you need and it'll take you an hour to search and maybe 2 weeks to be in your hands all for less than $10 for 20 of them including shipping.

By the way what the guy did is good in that he can use his drive and persistence to develop himself. I am pretty sure if he has enough of an entrepreneurial spirit, some of the methods he can use for other ideas.

 

[Red Squirrel]

I do think chip making is something that will eventually be more accessible. If a company does not want to rely on a clearly very fragile supply chain it will be in their interest to make the chips themselves.

If I was making anything electronic that would be the direction I'd want to go. Start by outsourcing most of it but slowly bring stuff in house such as PCB manufacturing and assembly, and eventually even the chips. If you make your own custom chips you can also taylor them to your exact needs.

Still need to rely on raw materials supply chain though, but ideally you can probably get that in fairly large bulk quantities and always have a big reserve.

 

[Doug S]

I do think chip making is something that will eventually be more accessible. If a company does not want to rely on a clearly very fragile supply chain it will be in their interest to make the chips themselves.

Zero chance. It costs over $10 billion to build a leading edge fab, and only three companies are left that have the know how to make it work. It takes a LOT of scale to offset that expense, which is why even companies with bottomless wallets like Apple will never manufacture their own chips. If Intel is not able to operate as a foundry, they will be forced to throw in the towel like HP, IBM, and AMD before them.

 

[moinmoin]

Zero chance. It costs over $10 billion to build a leading edge fab, and only three companies are left that have the know how to make it work. It takes a LOT of scale to offset that expense, which is why even companies with bottomless wallets like Apple will never manufacture their own chips. If Intel is not able to operate as a foundry, they will be forced to throw in the towel like HP, IBM, and AMD before them.

I don't think we are seriously talking about everybody taking leading edge fabs in their own hands. It would be already sufficient to define the concrete needs, find out (in many cases) that when specialized the compute capability are actually negligible, and realize it on e.g. 0.5 µm.

But of course for such dedicated efforts to be worth this needs a mind set change away from the current wasteful belief in eternal growth (which constant deprecation circle allowed for cheap hardware with zero effort). We may well eventually get to that point when all the necessary efforts to prevent a climate catastrophe kick in.

 

[Hitman928]

I don't think we are seriously talking about everybody taking leading edge fabs in their own hands. It would be already sufficient to define the concrete needs, find out (in many cases) that when specialized the compute capability are actually negligible, and realize it on e.g. 0.5 µm.

But of course for such dedicated efforts to be worth this needs a mind set change away from the current wasteful belief in eternal growth (which constant deprecation circle allowed for cheap hardware with zero effort). We may well eventually get to that point when all the necessary efforts to prevent a climate catastrophe kick in.

If you are wanting to do something on 0.5um, it would be far more economical and environmentally friendly to just use an fpga on a more modern process.

 

[eek2121]

Zero chance. It costs over $10 billion to build a leading edge fab, and only three companies are left that have the know how to make it work. It takes a LOT of scale to offset that expense, which is why even companies with bottomless wallets like Apple will never manufacture their own chips. If Intel is not able to operate as a foundry, they will be forced to throw in the towel like HP, IBM, and AMD before them.

Who said anything about a leading edge fab? What if a company just needs something that is simple, but custom? The issue with current foundry companies is that you have to buy volume or they will ignore you. If I want to design and build 1000 riscv chips with an added custom instruction set I can’t do that at an affordable price.

 

[Hitman928]

Who said anything about a leading edge fab? What if a company just needs something that is simple, but custom? The issue with current foundry companies is that you have to buy volume or they will ignore you. If I want to design and build 1000 riscv chips with an added custom instruction set I can’t do that at an affordable price.

It's going to be a heck of a lot cheaper than building and maintaining your own fab, whichever node you pick. If you need a small amount of volume, there are MPW programs you can get on to drastically reduce the cost though at 1000 devices there may be other options (e.g. MLM) depending on the device size.

 

[Doug S]

Who said anything about a leading edge fab? What if a company just needs something that is simple, but custom? The issue with current foundry companies is that you have to buy volume or they will ignore you. If I want to design and build 1000 riscv chips with an added custom instruction set I can’t do that at an affordable price.

The design and mask costs (even at a pretty old node) are a far bigger problem than the foundry cost - and there are plenty of foundries willing to talk to you about such small runs (or stuff like MPC/MPW like Hitman928 mentions) during normal times when the world's fabs aren't all 100% full. It costs more but as I said the design/mask costs for a small run are a much bigger problem and won't get that done at an "affordable" unless you consider thousands of dollars per chip an affordable price for your qty 1000 run. Which you might if you are the DoD, and that's probably one of the reasons modern weapons systems cost so much.

That's why if you wanted 1000 custom RISC-V chips you'd buy 1000 FPGAs made on a more modern process. That would be the cheaper option even if you owned your own trailing edge fab. You still have to pay part of the design cost, but basically only the equivalent of VHDL programming part, not any of the layout/masking stages which is where most of the cost is.

 

[Borealis7]

I bet he's a virgin

 

[igor_kavinski]

I bet he's a virgin

The extraordinary life of Nikola Tesla, a virgin inventor who spoke 8 languages | Qrius

Something about being a virgin keeps the brain more creative?

 

[moinmoin]

Something about being a virgin keeps the brain more creative?

Putting that energy into what really matters.

 

[sandorski]

There is at least one SaltyNuts/Mayne/Others thread where this should be Posted to.