Open Source Hardware

Open-source software has revolutionized many aspects of software development. But I believe it has a business-model problem. It is largely successful in supply chains where someone wants to "commoditize their complements".

It reminds me of Bruce Yandle's Bootleggers and Baptists. Bruce is an economist, and at one point headed up the FTC. The Bootlegger and Baptist way of looking at things is that both bootleggers and baptists want liquor sales to be illegal on Sundays. Baptists for religious reasons. Bootleggers to drive demand. Politicians, who want the lobbying dollars of the bootleggers, can use take the moral high ground by saying they are against alcohol for religious reasons, while pocketing the cash. In open-source software, the bootleggers are the large companies who make all the money selling servers, or advertising. The baptists are the open-source programmers who think they are changing the world by working for sweatshop wages in the noble cause of open-source. I wrote about this in more detail in Open Source in 2020 earlier this year.

As I quoted in that post:

The total amount of money being put into open source is not enough for all the maintainers. If we add up all of the yearly revenue from those projects in this data set, it’s $2.5 million. The median salary is approximately $9K, which is below the poverty line.

But it also affects all of us. Another quote from that post (about the HeartBleed vulnerability):

OpenSSL, a project that runs on 66% of all web servers, has just one full-time employee. One.

Open Source Hardware

Increasingly, there is a move towards open-source hardware. Of course, hardware doesn't have zero marginal cost so the actual hardware can never be "free" like software. But the design can be.

Facebook put the design of its servers into the public domain almost a decade ago since the more suppliers there are, the cheaper Facebook's data centers will be. It is known as the Open Compute Project. I wrote about that in my post Open Server Summit: How to Install 5,000 Servers Per Day about the Open Server Summit back in 2016. Microsoft, Apple, and Google are all contributors to the project.

Just a couple of months ago, I wrote about Rakuten's mobile network in Japan that is based on ORAN, the Open Radio Access Network. See my post Fourth 4G Network Goes Live in Japan. Just a couple of weeks ago, Nokia announced that they were adding support for ORAN. Here's a Reuters' report: Nokia to add open interfaces to its telecom equipment.

Another area at the intersection of open source and hardware is RISC-V. But note that RISC-V is not open-source hardware since it is "just" an instruction set architecture (ISA). You can build your own hardware, either open source or closed source, using RISC-V. without needing anyone's permission nor pay anyone royalties. Having said that, there are several open-source implementations available, some from academia (such as BOOM or Pulpino) and some commercial (such as Western Digital's SweRV or most of SiFive's products). For more on RISC-V from the summit at the end of last year, see my post What's Happening in RISC-V Land?

Open Source Chips

MOSIS came into existence in 1981 when the Mead & Conway revolution was getting underway (for more on that, see my post The Book That Changed Everything). MOSIS stands for MOS Implementation Service (with MOS being its usual metal-oxide-semiconductor). It combines primarily academic designs onto MPWs or multi-project-wafers. Normally, in semiconductor fabrication, the whole wafer is replicated with a single die. MPWs allow either more than one design to be put onto a reticle, or more than one reticle to be stepped onto the same wafer. This reduces the cost for any individual design. In fact, many test chips at the commercial foundries are done with the same approach before a process is in high-volume manufacturing or, perhaps, even finalized. MOSIS says that it has delivered over 60,000 designs over the years. For more about them, see their website.

Just a few weeks ago the FOSSi Foundation (Free and Open Source Silicon) made a big announcement. FOSSi regards its mission as:

Free and Open Source Silicon (FOSSi) are components and systems that are inside silicon devices (‘chips’). It is our core belief that building blocks that form such digital devices should be free and open (see our mission).

On June 30, they announced:

Did you ever dream about creating your own chip? I mean, a physical chip. One which you can hold in your hand, and which does exactly what you’ve designed it to do?

Until today, there were two major road blocks: you had to get access to a process design kit (PDK) from a chip manufacturing house (a foundry), and you had to have enough money to actually pay for the manufacturing. These times are over. Today.

Those two problems were fixed with two things:

• In a FOSSi talk, Tim Ansell of Google announced SkyWater PDK, the first manufacturable, open-source process design kit. What differentiates this PDK from previous attempts is the fact that it is manufacturable: with this PDK, you can actually produce chips with the SkyWater foundry in the 130nm node.
• That leaves you as chip designer only with one roadblock: money. Manufacturing chips is expensive—even for more than a decade-old nodes like the 130nm node, you need to spend at least a couple thousand dollars. ou know what? Don’t worry—Google and efabless have got you covered! They are providing completely free-of-cost chip manufacturing runs: one in November this year, and multiple more in 2021. All open-source chip designs qualify, no further strings attached!

Of course, 130nm is not the bleeding edge of semiconductor manufacture (it's about 8 or 10 process nodes ago). However, it is still used extensively, especially for mixed-signal designs and RF. It remains to be seen how much of an impact this has. In principle, you can design a 130nm RISC-V processor and fab it for free, but its performance is likely to be disappointing. But oak trees grow from small acorns.

Here's the video of the FOSSi talk announcing the SkyWater PDK (but beware it is over an hour and a half):

If you are less interested in chips themselves, and more interested in open-source EDA tools, seem my post OpenROAD: Open-Source EDA from RTL to GDSII.

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.