Everything else is very experimental at best. These days the FOSS FPGA tools are finally getting some traction, with Yosys and Nextpnr. But AsicOne tried to make an ASIC with open source, and faced endless troubles.

For ASIC there is basically QFlow, which is quite old, but used successfully to tape out a chip in the past, and there is OpenRoads, which is very new, experimental and ambitious. There are still major gaps in these tools, so in the end you inevitably have to sign an NDA and use proprietary tools and libraries.

And that's just talking about DIGITAL semiconductors where you compile HDL to pretty much generate the transistors from foundry cell libraries. So you have to sign an NDA to get the cell library, but you can at least release your code.

For analog chips, you can't do anything. An analog design highly depends on the parameters of the transistors you use, so before you even BEGIN designing, you have to sign an NDA to get the transistor models and you can NEVER open source an analog design.

The small dot of light at the end of the tunnel are projects like Minimal Fab, who make more accessible fabrication lines with open transistor models.

The crazy thing is that back in the days there were lambda rules, which were open rules anyone could use to design and model with. But with sub-micron devices, these scalable rules no longer scale, so fabs started producing secret models for their specific process.

I'm hopeful that after FPGA, and digital ASIC, analog will be next to be revolutionized.

Yes you can release your code, but you can't release your netlist or GDS-II. There's no guarantee that somebody else will be able to take the same HDL and close timing, even with the same foundry libraries (say if they are using a different tool, or different options). You'll also need things like clock-gating cells, memories, IOs (at a minimum) and those are foundry specific, so those would need to be abstracted out in some way.

> For analog chips, you can't do anything. An analog design highly depends on the parameters of the transistors you use, so before you even BEGIN designing, you have to sign an NDA to get the transistor models and you can NEVER open source an analog design.

Now this is where I disagree. Sure you can't open source your analog GDS-II, but maybe that's not the way to go. In my opinion what you want to do is build a foundry independent PDK for a generic 28nm, 40nm or whatever node using PTM models. A well designed analog circuit needs to be relatively independent of specifics, otherwise it's not going to work across all corners (this is more true for modern nodes than the kind of nodes the old textbooks talk about) and it'll be difficult to port to another process. So there's a good chance that analog circuits built for 'generic 28nm' or 'generic 40nm' could be ported to any foundries process (of course the PDK needs to be well designed). Yes you won't be able to push things to the limit as the DRC will be wider, but analog rarely needs to go to the limit. You could probably take the same approach for digital, but that's a lot more open source stuff to build.

Check out OpenRAM and FreePDK45 for academic projects taking this approach. Unfortunately FreePDK45 is only available to those with an academic email (despite being called 'open source'), which makes me very sad.

I talked to someone who worked on AsicOne, and he said that even if you make your own PDK and draw your own transistors and everything, you'll still have to sign an NDK to do the sign-off and what not. I'm not intimately familiar with the whole process myself, but from what I understand it is basically impossible to have an open source analog design that you can actually manufacture. (sure, you can make a theoretical toy thing, but if you can't manufacture it, who cares?)

You will need to run foundry DRC decks, but the company you're taping out through will do this for you (I presume that you're not big enough to deal with TSMC directly). This is because a design that fails DRC could actually break other people's chips if you're sharing a wafer.

Of course if you really want to know that it'll work you need to also run foundry LVS and stimulate corners with foundry spice models and foundry PEX. But if you're gutsy you could skip this, if you believe you've put enough margin into your PDK corners.

Certainly there is zero need to redraw your transistors. Transistors are transistors, a few layers (od, poly, contact, implant, ...), there's no magic, no magic sauce. The foundry wants an SVG with overlapping rectangles (of some minimum size), nothing more.

It’s very cost prohibitive to run a batch through a semiconductor fab. You are not going to request one offs unless you have a few millions to spare.

It’s not like running make after cloning some sources.

Maybe for a killer home router with a 100% open hardware+software platform that can route 8 ports of 5Gbit/port at line speed that would more trusted than random commodity hardware.

Or something to handle say 8 security cameras and use machine learning to handle all 8 streams for not just motion detection, but also identifying what person/object is in each stream.

I'd certainly pay a premium for smart devices in my home that I knew that I could trust, wasn't spying on me, doesn't require cloud connections, used open APIs/standards, and wouldn't die with the next time a company dies, gets bored, gets greedy, get purchased, etc.

It's fun to have access to HDL code but it's simply not possible to actually turn it into chips as an OSS effort.

FPGA-stuff is the way to go I think. The HW cost will be much higher but you can muck around with it as a lone engineer in your home with a turnaround time of minutes.

This page talks about the specific features available for hardware offloading:

https://help.ubnt.com/hc/en-us/articles/115006567467-EdgeRou...

faster iteration and development and more competition means more innovation and cheaper prices and more choice for consumers.

The primary benefactor of open source has never literally been the consumer because only an exceedingly small niche of people makes use of the source directly. The biggest benefactor is the ecosystem as a whole.

For consumers the great thing about open source has never been that they can run make in a terminal. It's been that people can run make in a terminal who can then produce products for consumers.

See Venezuela and Adobe. See the trade war with China and Huawei's position...

Interfaces are uncopyrightable, according to everyone but the Court of Appeals for the Federal Circuit. And the case where the latter said it was copyrightable has been appealed to the Supreme Court specifically to overturn that ruling, and the Supreme Court will probably do so.

Edit: You may be right that it's only patents afterall and not copyright. This was Intel's briefing:

"However, there have been reports that some companies may try to emulate Intel’s proprietary x86 ISA without Intel’s authorization. Emulation is not a new technology, and Transmeta was notably the last company to claim to have produced a compatible x86 processor using emulation (“code morphing”) techniques. Intel enforced patents relating to SIMD instruction set enhancements against Transmeta’s x86 implementation even though it used emulation. In any event, Transmeta was not commercially successful, and it exited the microprocessor business 10 years ago."

surely most of the x86 patents have expired, yet i don't see any interest in open source implementations.

The idea that open sourcing the logic is going to make other companies competitive with a financial behemoth like Intel is really a stretch.

All the more reason to make it profitable for fabs to exist within the US.

I'm particularly interested in this space. Anyone more familiar with it care to comment? Is there a push for more US fab capacity? Any startups? Any political push to change laws to make it more favorable to start one?

What stops Fab Facilities existing in "western" countries at the moment?

Someone told me a major reason for the move was because the chemicals used in top-tier facilities are basically banned in the west. Not technically banned but considered so dangerous (they're all carcinogenic) that the cost (health and safety, insurance, compensation) make them uneconomic.

Is that right?

Does it matter whether fab facilities are near to final manufacturing locations? Will Apple\Foxconn buy chips from a San Fran shop if they then have to be shipped to China for inclusion in the device? Since most devices are put together in SE Asia, the delay might be killer...

https://en.wikipedia.org/wiki/Foundry_model

Additionally there is this: https://en.wikipedia.org/wiki/List_of_semiconductor_fabricat...

Note as a result of this comment I did some digging and it does seem like Global Foundries does do wafer-share: https://www.globalfoundries.com/design-services/multi-projec...

RISC-V is great; a universal commodity ISA is welcome and will remove some impediments. But RISC-V isn't going to reduce the IC business to simple integration and fab. Specialized devices are already indispensable and will only become more critical as the general purpose CPU performance curve continues to flatten.

Transistor A whispered to Transistor B, while pointing at Transistor C: "I may be biased, but he looks a bit saturated to me."

Instead of pushing electrons with a voltage difference, to signify 0 or 1, a light wave of red or blue, can be used instead.

Anyone here a photonics expert?

The project goal was to outperform ASICs on a very well-defined, highly regular problem, and achieve it without the capital cost of typical ASICs, so that specialised circuits could realistically be built and used for different problems. So, not a general purpose CPU, but something that can compute.

OPA is potentially an extremely high bandwidth signal processing process, which doesn't involve converting to electrical signals, through transistors and then back to optical the way that some photonic designs do. It is more like the way optical communications amplifiers work, directly amplifying the modulated light that is passing through.

We are talking >1THz bit rates per logic element, and it's also quite an energy efficient process (despite limited OPA conversion efficiency, because you can recycle some of the light that hasn't converted), so it was worth exploring.

In the process, no insurmountable technical obstacles were found during the time of the project, but we ran out of time and money.

But it was surprising to find that, despite the superficial promise of photonics, it wasn't obviously a lot faster, or faster per Watt, than the best silicon electronics after all. This is because silicon transistors are pretty fast and efficient these days, and because you can fit a huge number of them in an area much smaller than the wavelength of visible light. You can confine light too, and there has been some published progress at nanoscale OPA elements, but it's a much more complicated structure and process (plasmons etc) than OPA in bulk materials, and nanoscale OPA may be just as difficult to manufacture as nanoscale transistors. Also, quantum: Just due to light being quantized as photons, there comes a point where to carry enough information at high data rates, the power density needed is an issue.

Sorry, I didn't answer your question :-)

My guess: Actually making a photonic CPU is economically and motivationally constrained rather than science constrained at this point, even though there's plenty of R&D still needed to do it. The motivation isn't that strong because the benefits aren't that obvious, and I think if they were obvious, the big commercial labs would have shipped a working prototype already.

At the long haul level, they use a transport layer tech called SONET. Rather than demuxing the optical signal into bits, then back into optical and out another pipe, they wanted to switch using fancy mirrors. Frankly, I'm not sure of the advantage. Maybe some performance? Security? I'm doubtful though.

There's at least some cool work going on in the GPU/TPU area.

Here are a few predictions, some of which have already occurred, that may result due to the death of Moore's law:

Already Happening:

- More custom chips (squeezing the last bit of performance)

- More reliance on the cloud, to push off processing power where there are more economies of scale

- The rise of traditionally "second tier" processor manufacturers (e.g., AMD, ARM) to be head-to-head with traditional leaders (e.g., Intel).

- A greater amount of chip manufacturers R&D dollars spent to each dollar of revenue.

Starting to Happen

- China and developing countries catching up with chip technology (when the leaders are no longer growing exponentially, it's easier to catch up)

- Governments imposing their will on chip developers (when there is less competition over performance, other factors like trust and national origin will start mattering)

- Trade secret theft, i.e., in the past if you stole Intel's designs, you would get one good chip, but that theft would be obsolete in 18 months. Now, it gives you a much longer advantage.

- An societal shift from utility to branding, i.e., as all goods start becoming equal, branding is the main differentiation.

- Living standards catching up to western and U.S. standards.

Further Afield:

- No significant technological improvements for decades.

- Stagnant per-worker/capita productivity.

- Economic growth becoming far more tied to population than individual productivity.

- Economies/governments fighting to increase their population (e.g., through legal immigration or by force).

- Government's power and control becoming based more on population rather than ideals or innovation (e.g., China).

- More monopolies ... in a dynamic and innovative society, a small smart company can defeat a larger slower one. In a stagnant society, that won't work and the competitive advantage can only be obtained by consolidation and economies of scale.

- Social unrest ... in an exponentially growing society, there is room for every generation to become wealthier than their parents, but in a flat society, on average half will become richer and half will become poorer.

If CPU and computer hardware become commodity, it would certainly be a new era of tinkering that is upon us. Perhaps more pressing problems than computational power f CPUs and transistor cost will then be attacked.

In industrial times, you would be right. However since the "information revolution", a huge amount of productivity gains can be directly attributable to transistors. Sure it started with improving simple calculators, but then came spreadsheets, industrial CAD models, instantaneous global communication, remote teams, global branding and a whole lot more.

In fact, it's pretty fair to say, that nearly every standard of living improvement in western countries since roughly 1980 can be attributable to the transistor.

Very interesting statement.

Out of cusiosity, is there any way you could back it up e.g. chemical or biological research was not very digital for a long time after the 1980s. Of course, digitization was transformative in the end of the 90s and beginning of the 2000s, and it seems to me that it is just now where we are in that exponential growth of exploiting IC technology.

I quickly checked your impressive background, and I guess you have seen a lot of the IC sector in terms of innovation and technology. I believe that you are making a sound call here.

At the end, everything is intertwined. Advances in material science go back to IC design and production, and vice versa.

Honestly, I therefore think the reverse development will further continue. Big brands will have an even harder time convincing people to buy their devices. Of course the economy might become more stagnant, on the other hand people might have to work less and be able to invest more time into other ideas - even if those aren't that profitable.

There are also already stagnant economies, for example Japan. There seems to be no correlation between growing/stagnant and inequality in terms of the Gini index: https://en.wikipedia.org/wiki/List_of_countries_by_income_eq...

Specifically what about lower-freq-more-cores becoming more and more common as we improve programming parallelism?.

But hey, valley of silly cons, who cares?