The story of efabless
The semiconductor industry has been an impenetrable wall for the open source community, but one company (efabless) plans to change that. The goal of efabless is for anyone to be able to create their own chip using open source tools for free. As efabless continues to achieve this goal, they are setting their sights even higher and will open the valve of open source to the last place that still needs it.
The Beginning
Mohamed Kassem worked at Texas Instruments (TI) developing the most complex products in the world on extremely tight schedules. At TI, Mohamed was on the team that was closest to the silicon and the tools used in the process. The tools he was using to model a system so complex were inaccurate and had gaps in predictability. Being able to function with these gaps gave him a head start when working with open source tools as they would have the same problems he was facing in the licensed tools. With the knowledge of how these tools could be inaccurate, Mohamed looked at popular open source simulators and ran experiments to gauge how well they operated. With some more experience using open source tools, Mohamed visited Tim Edwards, drew a picture of a chip and told him that he wanted to make it with only open source tools.
As it turned out, Tim said that companies were already building products using open source tools, just not sharing this information as they were private companies. Knowing how widely used open source tools really were in combination with many experiments, both Tim and Mohamed knew the project would be possible. Mohamed did not know how well the open source tools would work on a complete chip, but he knew he could close the gaps that the tools left him.
“The main barrier in open source is that people don’t believe a chip can be made”
The next step was to go and ask for the support of the people who made the tools. Mohamed told them that he wanted to make a complete chip with their tools and wanted their support. He told them that they did not need to invest anything and that if there were any bugs, he would fill the gaps. With the tools in place, it was time to overcome the next hurdle. Convincing a fab (chip manufacturer) to make the chip.
Getting a Fab On Board
A chip on paper (or in this case, a computer) is not quite as impressive as having it in your hand. Therefore, in order to create this completely open source chip, Mohamed would need someone to manufacture it. For most, getting in touch with a fab is impossible unless you know someone, work for a major tech company, or can write a check large enough to get them to answer the phone. In this case, the check size opened the doors.
Mohamed was consulting for a client based in the middle east who was working on making a clean room with a large budget. After looking at the proposal, Mohamed asked if they would be able to increase the budget and try to get a well known company to build an actual fab (not just a clean room). The client agreed and Mohamed began contacting well known fabs, and telling them the budget that would be involved. Once this number made it on the desk of the executives, they immediately began to call, some the next day.
One of these executives was Rudy De Winter, the CEO of X-Fab (a semiconductor company in Germany). During his meetings with Rudi, Mohamed discussed his efforts to make a chip from completely open source tools and create a community of those wanting to do the same thing. De Winter was extremely interested and told Mohamed that he would offer assistance and that X-Fab would manufacture it. The infrastructure and connections were in place and allowed the Raven (chip made from open source tools) to be made.
With one chip built and interest from some of the most important leaders in the industry, Mohamed was confident that the efabless idea had a strong foundation.
Skywater and Google
With this strong foundation built, it was time to get a foundry on board to manufacture more than one of these chips at a time. Mohamed connected with a tech founder who was extremely interested in open source technology and wanted to know what it would take to create an open source chip. Mohamed explained that the biggest challenge would be to get a fab on board, and the talks with Skywater Technology (a small fabrication foundry based in Minnesota) began. However, after coming inches away from closing the deal, the investor backed away.
Despite the first attempt at making a deal with Skywater falling through, Mohamed still wanted to make it work. While attending workshops and showing off the Raven, Tim Ansell (Software Engineer - Google) said that Google may be able to help. After some nagging from Mohamed, Google agreed. The deal was that; Google will commit to buying shuttles (chip manufacturing run) from Skywater for the efabless project. In return Skywater would be willing to find the money and reasons to allow the open source project. Once both sides signed on the dotted line efabless had everything in place to change the world.
Efabless Becomes a Household Name
After sealing the deal with Skywater, the news was released that Google (in partnership with efabless) would be manufacturing open source chips for free. Once this news hit the press, efabless went viral. Major news outlets and influencers in the tech space quickly hopped on the bandwagon and began spreading the mission of the Google and efabless initiative. The valve to the semiconductor open source community was open (at least for the 130nm process).
efabless was able to begin launching shuttles that included hundreds of developers eager to use the open source technology. As the chips were manufactured and delivered to their creators, they worked exactly how they were supposed to (with some minor hiccups). Proving not only that a chip could be made with open source tools by an experienced engineer, but that those with a much less impressive background could also design chips on efabless.
With this early success efabless quickly became a household name in the chip industry. It is clear that the future of open source chip technology will include and be led by efabless and its partners.
Present and Future
Education is one of the most important parts of efabless, as shown by the tutorials and available documentation on the website. However, Mohamed’s goal is to have enough good and well maintained educational material, that someone could from no experience at all to designing a chip. With some key partnerships, it is likely that this goal can be achieved in the near future.
Another key in open source education is partnering with universities across the world. Already, some schools are teaching open source tools and many have already been through the efabless process. Mohamed is continuously making relationships with some of the most prestigious universities so that students and faculty can have them in their arsenal.
The future of chips, without a doubt, includes the continued improvement and use of open source tools. Mohamed sees this already in successful companies that use only open source tools to create their products (like Antmicro). The commercial aspect of open source is important, as ideas cannot succeed for very long without money. However, it is difficult to get the commercial benefits if you are just an individual trying to make a chip for free. This is where the CLEAR FPGA comes in. The CLEAR FPGA is a crowd funded FPGA by chip ignite, the proceeds cover both the manufacturing and delivery of the product. The CLEAR FPGA will hopefully serve as an example for those who want to make a chip for free and also be able to put it in the hands of those who supported them.