Fab in India: India is trying to create an indigenous chip-making industry | Hacker News Reader

Fab in India: India is trying to create an indigenous chip-making industry

economist.com

123 pointsyarapavan51 comments

Context: The designs are based on RISC-V ISA (https://riscv.org). When India looks east, it sees Huawei, a Chinese tech giant, being cut off from American-made components as a result of the trade war. To the west, it sees its most talented engineers working in Silicon Valley. By pouring millions of dollars into Indian-made semiconductors, India’s government hopes to solve both problems at once.

The open source academic project mentioned in the article is 'Project Shakti' at IIT Madras.

Link: https://shakti.org.in/

Best of luck to India. I sincerely hope that RISC-V will grant them the ability to produce commodity grade (even if low end) chips that can be utilized with open source software system. The first such use may well be industrial systems where OS and spec are not the major concern.

I'm very excited about India's ambitions. It would not only be meaningful for India, but for many people around the world. We are lacking open chips, and project Shakti is on a trajectory to solve a part of that.

I wonder when GCC learns to use the source code of the hardware along with the source code of the software to optimize the binary with full knowledge of the system.

> I wonder when GCC learns to use the source code of the hardware along with the source code of the software to optimize the binary with full knowledge of the system.

Wouldn’t this just mean not having to rely as much on Agner Fog?

GCC uses intrinsic functions when available during compilation. It does optimize for a specific architecture. Is there something beyond this you are referring to?

We will never have competitive chips that are open down to the transistor and process level.

Down to the RTL, sure. But a truly open competitive chip, never ever.

You might get one on some ancient process from 20 years ago, which will be handily outclassed in every single metric except openness, which people make noise about but don't actually pay for (so no business case).

A ~20 year old process... that'd be 250nm or 180nm, which is Just Fine for many microcontrollers, if not the new 600mhz-1ghz Cortex M7's...

If the 180nm design is good enough, and migrated it to a new process, you'd get more chips per wafer.

Hence never competitive. "It's open!" doesn't make a business decide to buy a million and put them in a product - cost, features, availability, support do.

Nobody cares about "open silicon" - it's only spouted by people who don't understand actual electronics, or if they do, they certainly don't understand the economics.

It will never happen. If it did, it wouldn't be useful. It makes zero sense.

From a corporate perspective, I think you're absolutely right, this kind of thing will never be profitable.

From a state perspective, being reliant on other countries for your electronic may seem like such a threat to your ability to remain sovereign that it's worth developing your own, even if you are generations behind the latest and greatest.

True! But do you think they'd open up that process?

Hell no!

I hope they will be more involved in the open-source electronic design software than other countries. Software like LibreCAD[1] and FreeCAD[2], Yosys[3] and Symbiflow[4], Chisel/FIRRTL[5], OpenROAD initiative[6], Degate[7], and many others.

[1] https://librecad.org/

[2] https://www.freecadweb.org/

[3] http://www.clifford.at/yosys/

[4] https://symbiflow.github.io/

[5] https://www.chisel-lang.org/

[6] https://theopenroadproject.org/

[7] https://github.com/nitram2342/degate

I really would like an alternative to Shenzhen for cheap manufacturing. Does anyone here have experience with factories located elsewhere?

I regularly quote PCBAs in different countries. If you are looking for any place as cheap as China, I can't say that such a place exists. The only way that the math works out is by building the product yourself with a cheap SMT line. Using a domestic CM is a non-starter for a low-margin consumer product.

That's for ideological reasons that I am looking for an alternative. I am just wondering what would be the additional cost. Is that more than 2x the cost?

I've found the domestic premium too be anywhere from 3X to 10X. It really depends on how much is being done. We sometimes build antennas, which are single layer PCBs. The cost-delta at production volumes is about 3X. On a circuit board assembly I've paid 15x more to build in the US, where the turnaround time was significantly slower and production yield was worse. This is CM-specific, of course.

Have experience in Vietnam. Some clients been asking for manufacturing there, but often get surprised that it will both be harder and cost the more than in China. Still some persist, and do that.

You have to effectively import everything from China, and just do final assembly there. I see close to no point doing that when local inputs cannot reach even 10% of the value.

Do you know if there is any decent plastic molding industry in Vietnam? I have something that needs off the shelf electronics (may buy them in China) but will have to meet manufacturers in person for the final assembly and plastic casing, which on a small series is probably going to be a sizeable portion of the cost (and seriously, that's more about ideology than cost anyway).

Even ordinary quality level plastics parts can often be easier to import from China there, moreover now. Even things like plastic bathtubs is a challenge at the moment. Not many options for small businesses there other than to work with garage factories, with all resulting complications.

I will be glad to get in touch with you if you really need to make it there, and you are really ready to pay. For a project under $100k you will likely pay a fee from 10% to 20%. Disclaimer, this is not a binding commercial offer

For plastic mold injection it's less about the infrastructure and more about how competent and reliable the supplier is. Circuit board assembly requires parts to be sourced. With plastics it's possible to be more logistically self-sufficient. You should be able to find what you need in Vietnam, but unfortunately I can't point to any specific sources.

Just confirming your 10% figure. Across a dozen or so products I've been involved with, the actual labor/SMT assembly costs have been about 10% of the total BOM (bill of materials). Very roughly the cost per component-placement is about 1-cent (USD). My quoted labor rates are about $7/hour, though I see $15/hour at a lot of other factories (and warehouses).

https://outline.com/E8CzrA

What did make RISC-V a game changer? Why did not any open source chip enjoy such a success before?

The question is: What will they do after, when Trump is gone, and his policiey on Chinese imports anf foreign workers are gone with him?

I hope they hurry. Now’s the time!

Reply on news.ycombinator.com