DARPA Has an Ambitious $1.5 Billion Plan To Reinvent Electronics (technologyreview.com)

The Defense Advanced Research Projects Agency (DARPA), which funds a range of blue-sky research efforts relevant to the US military, last year launched a $1.5 billion, five-year program known as the Electronics Resurgence Initiative (ERI) to support work on advances in chip technology. It has now unveiled the first set of research teams selected to explore unproven but potentially powerful approaches that could revolutionize US chip development and manufacturing. From a report: The ERI's budget represents around a fourfold increase in DARPA's typical annual spending on hardware. Initial projects reflect the initiative's three broad areas of focus: chip design, architecture, and materials and integration. One project aims to radically reduce the time it takes to create a new chip design, from years or months to just a day, by automating the process with machine learning and other tools so that even relatively inexperienced users can create high-quality designs.

"No one yet knows how to get a new chip design completed in 24 hours safely without human intervention," says Andrew Kahng of the University of California, San Diego, who's leading one of the teams involved. "This is a fundamentally new approach we're developing." William Chappell, the head of the DARPA office that manages the ERI program, said, "We're trying to engineer the craft brewing revolution in electronics." The agency hopes that the automated design tools will inspire smaller companies without the resources of giant chip makers, just as specialized brewers in the US have innovated alongside the beer industry's giants.

Comments

Uh.....

[phantomfive]

"No one yet knows how to get a new chip design completed in 24 hours safely without human intervention,"

What on earth is he talking about?

Re:Uh.....

[drinkypoo]

This is how you get skynet. Do you want skynet? DARPA apparently does

Re:Uh.....

[optikos]

eliminating the thousands or tens of thousands of timing violations. ASIC development goes a compilation process of source code not entirely different than software. The challenge of ASIC design is locating everything just right so that distance is globally nearly-minimized (local optima but near the global optimum) to get all the electron pulses to arrive where they need to arrive before it is too late. Plus, each logic gate costs a time delay (as well as occupies space exacerbating the distance problem). So another trick to solve timing violations is to simplify the design is some locality to lessen the depth of gates that a signal/calculation/operation must traverse, when viewed as a directed-acyclic graph (DAG). Sometimes space (# of gates) can be bloated up to decrease the depth of the walks of the logic-gate DAG (but then that increases area on the die, which exacerbates the distance problem).

Re:Uh.....

[Austerity Empowers]

Not to mention that the tools that area already automated to *assist* with this stuff are buggy as fuck, and EDA companies software development practices make Microsoft look good.

But more power to DARPA, the more that can be automated, the more that can be accomplished. Chip design is still so expensive that only a few people with very deep pockets can participate.

Re:Uh.....

[Anonymous Coward]

Do you want skynet?

Yes, but only if the Terminator comes for drinkypoo.

Re:Uh.....

[DontBeAMoran]

If you're going to do an Archer-style quote, at least do it properly:

"Do you want Skynet? Because that's how you get Skynet."

Re:Uh.....

[110010001000]

He is saying "we are going to spend taxpayers money on an idea I have". DARPA programs are full of great ideas.

Re:Uh.....

[Anonymous Coward]

If you can do a complex chip design from HDL to mask in 24 hours I think every chip manufacturer in the world will want to hire you.

Re:Uh.....

[Anonymous Coward]

>Not to mention that the tools that area already automated to *assist* with this stuff are buggy as fuck, and EDA companies software development practices make Microsoft look good.

Hmm, that depends. I find that Altera Quartus is extremely good software for ASIC design.

Re: Uh.....

[Type44Q]

I don't know but obviously it's bullshit; I didn't see any mention of "AI" anywhere.

Re:Uh.....

[4wdloop]

And that problem (of optimizing DAG) is at least NP.

Re:Uh.....

[eminencja]

tens of thousands of timing violations (..) as well as occupies space

Size matters but timing is everything!

Re:Uh.....

[Anonymous Coward]

And this is a bad thing? 1.5 billion hardly rates it's on line item on the country's budget. And the bulk of these funds are paid in the form of grants to research Universities and private companies that are attempting to innovate. This removes the pressure of having to generate a ROI to fund their work.

And by the way only around 50 percent of US citizens pay any federal taxes. Those complaining their federal tax dollars are being misspent are usually not paying any federal taxes to be misspent. They are complaining about how other peoples tax money is being spent. This is the result of income levels, tax breaks, and special provisions in the tax code. This does not include SS or Medicare withholdings which are already earmarked for specific expenditures. And 20 percent of Americans earn 53.4 percent of the total U.S. income but pay 67.2 percent of total income tax.

Re: Uh.....

[trevc]

Or RaspberryPi or 3D printing. Very disappointing.

The 'Matter Compiler' approach

[Rick Schumann]

I'm beginning to think that we're reaching the limits of what we can do with the laser lithography method of silicon IC creation. For instance look at the problems Intel is having with 10nm fabrication right now. Perhaps the way forward is straight out of science fiction: a matter compiler/3D printer-like approach, where an integrated circuit is built up an atom or a molecule at a time? Pure imagination on my part, but is it really out of our reach?

Re:The 'Matter Compiler' approach

[bobbied]

I'm beginning to think that we're reaching the limits of what we can do with the laser lithography method of silicon IC creation. For instance look at the problems Intel is having with 10nm fabrication right now. Perhaps the way forward is straight out of science fiction: a matter compiler/3D printer-like approach, where an integrated circuit is built up an atom or a molecule at a time? Pure imagination on my part, but is it really out of our reach?

Doesn't seem that far out of reach theoretically. However, as in all things, the practical cost effective reach seems a long way off.

A lot of things *could* be done, but we don't do them because they are too expensive or better/cheaper/faster options exist so we use the other options.

Re:The 'Matter Compiler' approach

[Nidi62]

Doesn't seem that far out of reach theoretically. However, as in all things, the practical cost effective reach seems a long way off.

A lot of things *could* be done, but we don't do them because they are too expensive or better/cheaper/faster options exist so we use the other options.

Isn't that precisely what DARPA is for, though? Fund research that is cutting edge, highly speculative, or too long term for private companies to undertake? A lot of the technology and techniques that DARPA develops eventually makes it's way to the civilian market as well.

Re: The 'Matter Compiler' approach

[Anonymous Coward]

AMD seems to be doing just fine with 7nm, so not sure how Intel sitting on their hands relates to this.

Re:The 'Matter Compiler' approach

[Anonymous Coward]

There you go. This system they are developing might be able to be generalized into other design related fields in the smaller scales. No more leaky facades, constricted parking or inadequate toilets in 100 million dollar opera houses would be nice.

Re:The 'Matter Compiler' approach

[bobbied]

Oh sure, but DARPA has a limited budget and needs to be throwing it's money towards the ideas that have the greatest possibility of paying a dividend on the investment. So if there are other more promising ideas that are higher on the cost/reward estimates, they will be funded first.

But my post wasn't about DARPA. It was about the development in technology in general.

Re:The 'Matter Compiler' approach

[Anonymous Coward]

The (main) problem with sub 14nm nodes is that you need to get everything right on each step in the mfg process - the old "good enough" standard for some steps really just doesn't apply anymore, as a flaw anywhere kills the yield. Also seems like the 10nm node isn't going to be a winner, more likely to see smaller future nodes be more successful.

Re:The 'Matter Compiler' approach

[Nidi62]

Oh sure, but DARPA has a limited budget and needs to be throwing it's money towards the ideas that have the greatest possibility of paying a dividend on the investment.

Good point there. Imagine what they would already have come up with if we took the money wasted on the F-35, LCS, or, going back even further, Land Warrior/FCS and had given it to DARPA. We might have flying, or at least autonomous, cars.

Re:The 'Matter Compiler' approach

[AHuxley]

Whats the math and physics say about getting so small that things stop working as needed? How many more nm smaller can be see as useful for the costs for a mil/gov project?

Re:The 'Matter Compiler' approach

[shess]

I'm beginning to think that we're reaching the limits of what we can do with the laser lithography method of silicon IC creation. For instance look at the problems Intel is having with 10nm fabrication right now. Perhaps the way forward is straight out of science fiction: a matter compiler/3D printer-like approach, where an integrated circuit is built up an atom or a molecule at a time? Pure imagination on my part, but is it really out of our reach?

If you're having problems implementing high enough manufacturing standards to support your current goals, then the solution isn't to switch to a system which is even *harder* to implement.

Re: The 'Matter Compiler' approach

[NicknameUnavailable]

Intel is backing the spectre/meltdown fixes into the "10nm" line to avoid spooking investors. It's a multi-year (at least) redesign because they have to completely reengineer their platform in order to get the backdoors fixed (without breaking the intentional backdoors.)

Re:The 'Matter Compiler' approach

[bobbied]

LOL You do realize that DARPA has been after the autonomous vehicle concept for over a decade now right? It's been going on so long now that I've seen multiple TV programs on the "science channel" about it and at least one "National Geographic" and one "NOVA" special too.

Re:The 'Matter Compiler' approach

[Nidi62]

LOL You do realize that DARPA has been after the autonomous vehicle concept for over a decade now right? It's been going on so long now that I've seen multiple TV programs on the "science channel" about it and at least one "National Geographic" and one "NOVA" special too.

I know they have. Just imagining what billions more in funding could have done. Throwing money at problems doesn't solve them, but it can go a long way towards it.

Re:The 'Matter Compiler' approach

[cdibbs]

Maybe they should start by making computing reversible. https://spectrum.ieee.org/comp... If they don't, then they'll run into temperature problems, regardless of the IC creation method.

Re: The 'Matter Compiler' approach

[Anonymous Coward]

That would be TSMC then. AMD is fabless

Re:The 'Matter Compiler' approach

[dcw3]

The D in DARPA stands for Defense, so your flying cars will have to wait. And DARPA doesn't get the money from all those other projects you mentioned. If I'm reading you correctly, you're arguing that we didn't need those weapons mentioned above. I'll disagree with you, while agreeing that we paid much more than necessary...FWIW, I've been in the defense contracting business for more than 35 yrs. DoD in general needs to have the way in which it procures weapons completely revamped. DARPA has had plenty of success stories (below are just a few), so I have no issue with them spending money this way.

https://futurism.com/nine-most...

Re:The 'Matter Compiler' approach

[Nidi62]

If I'm reading you correctly, you're arguing that we didn't need those weapons mentioned above. I'll disagree with you, while agreeing that we paid much more than necessary...

F-35: would have been much better spent upgrading existing airframes. the aircraft was so full of problems they had to delay retirement of the A-10 which is probably one of if not the best ground attack aircraft ever built, might as well have just kept them and upgraded them. The F-22 production lines could have been kept open. F-16s are literally breaking apart but they are already proven designs with established production and logistical support, could have just created a new block. It's trying to do too many things at once when what we already had worked plenty well enough.

LCS: Under-armed, under-armored, under-manned, propulsion systems that are inherently flawed and prone to failures, hulls prone to cracking, and over-reliance on contractors for support. Billions of dollars spent to retire frigates, only to end up with "replacements" having to be redesigned so much that they are now also classified as "frigates". Just like the F-35, jack-of-all-trades in the military gets you a very expensive and operationally compromised mess.

Land Warrior is the only thing that made a little sense. Networking soldiers is a great idea, especially in urban environments. Integrate real time intelligence from small, unit level UAVs and you can own the battlefield. The technology just wasn't mature enough at the time, but at least now that it is they should at least have a good idea of which ends are dead ends.

Re:The 'Matter Compiler' approach

[tlhIngan]

I'm beginning to think that we're reaching the limits of what we can do with the laser lithography method of silicon IC creation. For instance look at the problems Intel is having with 10nm fabrication right now. Perhaps the way forward is straight out of science fiction: a matter compiler/3D printer-like approach, where an integrated circuit is built up an atom or a molecule at a time? Pure imagination on my part, but is it really out of our reach?

Actually, modern ICs are done at the atom level buildups - to create the strained silicon that we use today requires subjecting the wafers to vapor deposition that literally grows the surfaces atom by atom.

The problem with 10nm is not the photolithographic process - we've outdone ourselves in this technology. It's the fact that at 10nm, transistor gate lengths are starting to be measured in the realm of atoms - with gate insulation thicknesses being a handful of atoms (which can easily lead to leakages). Add to this quantum effects - tunneling and the like and you get leaky transistors that consume a lot of power and generate heat.

Also remember that the only circuit elements that use the smallest dimensions are transistor dense ones - memory blocks. Things like general logic in a processor a wire-constrained (you are limited by how much wiring you can stuff between transistors and not the transistor size themselves), which is why they always fab tons more transistors than are needed, but still way larger than minimum size.

Thus, leaky transistors are in places where you don't want them to leak - generally memory cells where leaky charges mean bit flips.

Re:The 'Matter Compiler' approach

[dcw3]

I'm well aware of how good the A-10 is/was. You'll be hard pressed to find any company that can upgrade a plane that's been out of production that long, by a company that no longer exists.

The F16 is/was a great plane as well, but it's gotten pretty long in the tooth now. And why are they breaking apart? Planes go through "phase" on a regular basis where the plane essentially gets rebuilt. It's why aircraft like the U-2 just hit 60 years old, but none of the airframes are anywhere near that.

As for LCS, F-35, etc. Virtually every major product like these are initially delivered with major bugs to work out. It's the nature of the business and the way that the Pentagon does requirements, funding, and scope creep, as well as some shitty contractors, that end up causing huge budget overruns.

Re:The 'Matter Compiler' approach

[Areyoukiddingme]

Perhaps the way forward is straight out of science fiction: a matter compiler/3D printer-like approach, where an integrated circuit is built up an atom or a molecule at a time? Pure imagination on my part, but is it really out of our reach?

Neither science fiction nor your imagination. It's called a nanoassembler, and it was described fairly formally for the first time in Engines of Creation by K. Eric Drexler in 1986. People have been working on the concept ever since. It's still at the level of attempting to build the tools required to build the tools required to build the tools required to build the nanoassembler. Optimistically. It could be even worse than that. Shoving individual atoms and molecules around into the desired order is for all intents and purposes not a mechanical process. Electromagnetic forces are everything, at that scale, and atoms are very nearly unreal.

Re: The 'Matter Compiler' approach

[raftpeople]

But remember that node comparisons between companies are apples to oranges. Intel 14nm is approx the same as other fabs 10nm. In addition, they aren't exactly sitting on their hands, they made a choice in their process approach that is turning out to be probably a poor choice and they are struggling, but that's not the same as sitting on their hands.

Re:The 'Matter Compiler' approach

[Anonymous Coward]

A lot of things *could* be done, but we don't do them because they are too expensive or better/cheaper/faster options exist so we use the other options.

I see a paradigm shift here: Creating one, custom, chip on a day, with nearly 100% yield (not mentioned, but implied) for a dollars or hundred dollars worth, instead of 10000 or 100000 units batch with final production costs of fractions of a cent per unit. The things have not been done like that ever since 1940s-1950s when first contact transistors had been made. From then on, the goal has always been mass production of electronic devices.

That task is not suitable for industrial processes we used so far, just like some optimal solution for one-off chip production would be ridiculous if attempt was made to scale it up to industrial level production volumes. The difference would be like the difference between a 3D printer and an injection-moulding automated line.

Hackers/Makers will be grateful for this.

Craft Brewing?

[Anonymous Coward]

This sounds more like the AnheiserBuschification of electronics. Less human care and oversight. More volume and automation. This ends with the Bud Light vat-wash of technology.

Re:Craft Brewing?

[Plumpaquatsch]

This sounds more like the AnheiserBuschification of electronics. Less human care and oversight. More volume and automation. This ends with the Bud Light vat-wash of technology.

Hey, maybe it will be more like "Dude spends a lot on brewing equipment and then uses it once (if at all), and the result of that one attempt tastes like shit." AKA "home craft brewing".

China steals this in 3...2...1...

[Anonymous Coward]

We pay for it, they take it.

Re:China steals this in 3...2...1...

[Anonymous Coward]

Nah... all this hand waving is just an attempt to keep China from raising electronic component prices any higher due to recent tariffs.

Re:China steals this in 3...2...1...

[Anonymous Coward]

Nah... all this hand waving is just an attempt to keep China from raising electronic component prices any higher due to recent tariffs.

Thank you dear professor of Chinese sainthood culture.

Intel's backup

[Anonymous Coward]

You can always trust the government to do the wrong thing!

FPGA

[bugs2squash]

Better open source programmable logic support and better devices would spur a great leap forward. icestorm is a great effort but it is limited to devices that seem elderly and affordable field programmable devices don't seem to be advancing very quickly.

Re:FPGA

[Austerity Empowers]

FPGAs have advanced a lot in 20 years, but they are the moral equivalent of running your code under Java. That's not at all what they do, but their ability to configure on the fly comes at a very high cost in terms of frequency, power and area. A custom chip is always going to be faster, smaller and cheaper (COGS wise).

It's true that FPGA mfg's could do more to enable other tools, but their motivation is very weak.

Re:FPGA

[Anonymous Coward]

Just like there is a place where Java is appropriate, there is a place where FPGAs are appropriate. Especially because doing raw custom chip design is getting more and more difficult (and therefore expensive) the smaller the chip features are.

Re:FPGA

[Agripa]

It's true that FPGA mfg's could do more to enable other tools, but their motivation is very weak.

Their motivation is negative. Practically no FPGAs are sufficiently documented for third parties to create programming tools.

Another AI winter here we come

[Anonymous Coward]

One project aims to radically reduce the time it takes to create a new chip design, from years or months to just a day, by automating the process with machine learning and other tools so that even relatively inexperienced users can create high-quality designs.

And they don't think the existing foundaries are working towards this goal already? It's very lofty and doomed to failure. Better work on little bits and pieces of the puzzle instead...

Re:Another AI winter here we come

[Anonymous Coward]

Foundries will ultimately profit on this, this is just a way of speeding up prototyping phase and cutting the costs of development. But, it is not in most foundries' business model to do their ow research.

Right now, chip "manufacturers" go through design and thorough verification phases before they knock on a foundry door and deliver the masks for photolithography they believe is OK (but will not know for sure before an expensive batch is done and specimens tested). The goal here is to get to testing phase in a day, without excessive cost, and then iterate until the design is acceptable, make masks and send them to foundry. With wider base of customers, foundries will have more work.

This is Socialism!

[Anonymous Coward]

It will never work to just throw money at it, as we all know from the Internet, government cannot produce anything of value.

Re:This is Socialism!

[Anonymous Coward]

Correction - not "as we all know from the Internet", should be "as we know from morons that endlessly praise capitalism despite its many flaws". Internet itself has very varied sources and conclusions on the subject.

You mean like

[goombah99]

The development of a human brain from a single cell?

Re:You mean like

[Anonymous Coward]

Exactly! Why have we not developed a computer that works like the human brain cell? What unique materials could we use to approximate our existing chip technology? Open research like this is needed. Too many research grants are based on expected results.

Not with this one

[evanh]

Given all the design work is done on the workstation by typing a lot of HDL code and doing spice simulations and maybe even some layout work, the particular focus of this one is purely a software investment.

The Internet

[hduff]

It would be nice if they could re-invent the Internet . . .

So terrific for the economy

[rbrander]

Basic research! Nothing more likely to "fail" in the commercial sense, and so hated by free market companies that hate risk (all of them); nothing more likely - longer term - to come up with the big finds that create whole new economic sectors.

Their "ARPAnet" idea wasn't even supposed to make money, that's the funny bit.

With just a little luck, some of this research will end up creating whole new economic opportunities, which will result in a few people becoming billionaires, who will probably, with tiresome regularity, turn out to be libertarians who don't believe government can do anything useful and attempt to pay no taxes.

Ah, those public bureaucrat-scientists struggling for grants: America's true Job Creators.

(Juuuust kidding, of course. America's real job creators are consumers: without people putting butts in seats of the restaurant, neither the cooks&waiters, nor the restaurant owner, nor his banker, have any jobs.)

Re:So terrific for the economy

[cyberchondriac]

America's true Job Creators.

(Juuuust kidding, of course. America's real job creators are consumers: without people putting butts in seats of the restaurant, neither the cooks&waiters, nor the restaurant owner, nor his banker, have any jobs.)

Not entirely, supply and demand play off of each other. Sometimes people are more enticed to consume more when there are more choices, better prices, shorter wait lines, or overall better service.
A fairly successful standalone restaurant owner can decide whether he wants to expand or not. He might be content with a well running business, though that will hit a ceiling as people weigh the quality of the food and service against increasing wait times and crowding, not to mention maximum occupancy. So maybe some just won't go out to eat tonight. OTOH, he could be an ambitious rascal and open another restaurant, and then another, hiring 3x as many employees, and enticing more customers to part ways with their money. This also boosts the related economies such as food wholesale, transportation, insurance, building contractors, etc.
(Or he could overestimate his market, go under and blow the whole thing)

Re:So terrific for the economy

[Anonymous Coward]

Edison, Tesla, and the Wright Brothers were all government employees working for the greater good of the people's collective under the guidance of Great Leader.
Oh wait, that didn't happen, in fact the communists were only able to "innovate" by stealing ideas from capitalists.

Not suspicious at all.

[Anonymous Coward]

Backdoors for days.

Unfortunately ... useless.

[Anonymous Coward]

Since this is coming from the US military, it's pretty much useless, unless you are the US military.
Everyone else won't touch it with a two mile pole, due to the 100% guarantee of there being back doors in there and that being then entire point of this program.
(Cue the US military sock puppets trying their cheap dated memes on me. Sorry, "conspiracy", "tin foil hat", and similar thought-terminating straw man arguments do not work anymore since Snowden. I have seen TAOs happen in real life, and know about dopant-level hardware trojans you freedom-hating anti-democratic fascist fucks.)

No one ever said...

[reanjr]

No one ever said DARPA wasn't ambitious.

Translation

[Anonymous Coward]

"relatively inexperienced users" = low-paid workers. We don't need those experienced, and high-priced, engineers.

Hey Siri, print me a homicidal AI bot

[nospam007]

I guess that's what they want.

Blue sky or green field?

[Tony Isaac]

What they need is people who think outside the box. An idea shower might help going forward. This new research will push the envelope, drilling down to new ideas. This will be a win-win scenario, impacting the bottom line. The game plan should be results-driven so we can all hit the ground running. When all is said and done, we might find we have re-invented the wheel, forcing us to go back to the drawing board. It's a no-brainer!

Craft Brewed Chips

[lys1123]

"We're trying to engineer the craft brewing revolution in electronics."

I for one am looking forward to running a computer based on The Great, Big Kentucky Sausage Fest Chip