Semiconductor Engineering sat down to discuss challenges at the edge, the impact of open-source, and how to attract new talent, with Simon Segars, CEO of Arm; Joseph Sawicki, executive vice president of IC EDA at Mentor, a Siemens Business; Raik Brinkmann, CEO of OneSpin Solutions; Babak Taheri, CEO of Silvaco; John Kibarian, CEO of PDF Solutions; and Prakash Narain, CEO of Real Intent. The conversation was part of the ESD Alliance’s annual outlook, which this year was held virtually.
SE: Advanced packaging is gaining momentum across the industry, but not everything is in place. What still needs to be solved?
Kibarian: There is lot of innovation in the package. From a reliability and a security standpoint, the riskiest part of the supply chain is the assembly portion, because now you have an individual die moving throughout the supply chain with much less traceability. We’ve been working with SEMI on standards for traceability, particularly around the assembly flow, and for leveraging ledger technology like blockchain for logging what happens with material as it moves through that part of the supply chain. There’s no single entity that can make all of that happen. By and large, the assembly vendors are very thinly margined entities. They’re not going to invest in the infrastructure that is used by a broad spectrum of customers. One day that same line is running assembly for an Apple chip, and the next day it’s running assembly for a Broadcom chip. There are no the chip suppliers in the system. Companies that eventually are going to leverage that data don’t own or manage the equipment. So collaboration throughout the industry is going to be very important. We’ve been promoting this for our customers to participate in the SEMI standards. We are active participants on this. We think it’s a super-important part of bringing collaboration for the manufacturing part of this. SE: One of the big opportunities going forward is the edge. But there is so much diversity and so much customization that you can’t develop one chip and have it play across a lot of different areas. What does this mean for the semiconductor industry? Segars: It’s a massively fragmented space, and ‘edge’ as a term isn’t very helpful because it is so broad. When we talk about edge, we think about things that sit on the network side, and then we talk about end points, which are the things connected to the network. And the amount of compute within those endpoints can range from absolutely astronomical in a self driving car, or really, really tiny in a sensor that’s monitoring water flow through a pipe and trying to spot when there’s a leak. The range of solutions that you need is just enormous. But there are some common threads to it. There are sensors, there is computing, and there is a desire and a need to do as much processing locally as possible, because the cost of the computing locally is way cheaper in a lot of cases versus firing up the network, transferring data into the cloud, doing the processing there, and then try to answer back. All of this creates a lot of opportunity for the semiconductor sector. We’re going to need more advanced, more energy-efficient computing devices that sit in those endpoints, and there’s just such a broad range of them. The design space right now is wide open. You need compute power sitting at the edge of the network, which all those things are connected to. There’s a really interesting software complexity problem that arises from that, where you have an application that is running in multiple places. It’s running in the cloud. Some of it is running at the edge of the network doing local processing. Some of it is running in the endpoint. You’ve got a coordination problem between all these different base stations that might have mini data centers in them, keeping the software up to date and keeping the interaction of one live with another. There’s a whole set of complexity that no one’s solved yet. I look at this and see a ton of opportunity and a ton of really hard problems. The world’s engineering community loves to go and tackle those things. So this future is not going to be boring. There’s a lot of really hard problems to solve. Sawicki: One thing I find interesting about about the edge is that success is going to be completely determined for any particular device manufacturer based upon what they can do in terms of compute per watt. Our industry has spent probably the last 15 years trying to enable high-level design techniques. This is really the first market that sees that as being critical. It’s being able to design at a SystemC/C level and really optimizing the nodal structure of your neural network and determining how that’s going to map against the characteristics of a particular software stack that’s going to layer on top of it. It becomes the critical item in terms of driving performance and driving down the cost. It’s the first time we’re actually seeing that technology start to have a very large take-up, because that whole thing about optimizing the circuitry for the task of that particular edge device is critical to having market acceptance. Taheri: Back in 2012 when IoT was coming together, edge had a very different definition than it has today, and that definition depends on the application. It could be a single sensor, such as a pressure sensor or humidity sensor that has all the algorithms embedded into it and wirelessly communicates with thousands of these things, all the way through to automotive, which can be considered an edge device. It encompasses many, many industries and applications, and depending on the architecture of the system, the edge varies based on compute, storage, wireless communication and power consumption. It’s always good to have anything that you could get your hands on that’s uses the least power, since it’s wireless and needs to be able to communicate these things. But more importantly, there are a lot of companies in agriculture or industrial markets that deal with robots, etc., and when they don’t see a solution out there they put it together themselves. That creates an opportunity in terms of how you provide IP, how you provide tools to these companies, and how you validate a system in order to make sure it operates correctly and reliably. If it’s if it’s a single sensor, or if it’s a gyroscope that sits in sits in your phone to do optical image stabilization, the application is different, the algorithms and power consumption requirements are different. And the key to all of these is to be able to customize. There are a lot of custom chips that are coming out that deal with certain segments of the market. Customization is the key, and how you address each customer’s needs with IP, tools and requirements is what’s coming up. SE: So we are looking at more fragmentation in the supply chain in order to serve these different markets. What impact will that have? Kibarian: We work with our customers to deploy analytics across their supply chain. People want a lot more diversity in supply chain. They want to have a more flexible supply chain than they had in the past. And the chip industry was set up primarily around getting to the new node fast. That was an anthem for what you wanted to do. You wanted to lock that all down and move really quickly. We’ve moved from a time where they had fixed nodes and fixed instances to really an agile-like method for the way we use technology. There’s many flavors of 7nm. There’s many phases of packaging. And there’s many possible ways you can implement that in the supply chain now. The industry is moving rapidly toward factoring flexibility into the supply chain as one of the elements in how they look at technology, and we think that’s going to be a bigger and bigger factor going forward for most of our customers. We already have very complex supply chain system interactions, which make it far more complex. SE: For a while, we saw the pendulum swing from hardware to software, where software was the cool place and everybody wanted to work for Facebook or Google. The pendulum has swung back, at least to the point where it’s hardware and software together. Is this helping to bring more talent back into the industry? Segars: There’s a bit of a shift going on. There was a period where software was king and it appeared that computing happened in the cloud. The challenge was, ‘What cool things can you do in software because you don’t have to worry about the efficiency of the hardware.’ With this growth of more edge devices, there is a need to drive efficiency. You can run these really complex software stacks, but the problem with customizing everything is that you get no software reuse. In addition, with software you need to worry about reliability and security. That is the worst combination you possibly could have. So there is a really interesting challenge around how you enable productivity that comes from complete standardization, and flexibility and efficiency that comes from complete customization. That inevitably requires getting people to really understand the hardware, computing and software together to solve those problems. But when you look at some of the results that can be achieved, it’s a fascinating technical problem that I think bright minds will want to go and focus on. Now, we probably could do a better job of marketing that opportunity, that kind of intellectual stimulation. But we are seeing people really interested in how you enable computing on a pinhead-size device and deliver massive power. That is an intellectually stimulating problem we see people wanting to go and solve. Sawicki: It ties back to a comment was made earlier about how systems companies are getting involved so much more heavily in the semiconductor space. The first signal was when when Apple first went to their 64-bit architecture. The claim was that it had nothing to do with things that geeks normally think about, which is address space and the size of programs you can put in place. It was done for power purposes. So here you had a company that was definitely the one that cool kids loved, which had just gone into the deep microarchitecture of their application processor to enable people to be able to get better performance out of their phones. Now virtually every one of the large cloud-based systems vendors — Microsoft, Google, Facebook — is doing hardware design as a key part of enabling what they’re doing. That clearly has an impact in terms of how people look at hardware. In terms of how that plays out internally, those are those are macro trends will help attract good people into the industry. Brinkmann: On way we can attract the best and brightest into this industry is with open source. Usually when you look at what attracts talent, it’s the ability to be creative in environments that are challenging, and where there is a lot of opportunity to make money and build careers. Open source doesn’t really sound like you have a lot of opportunity for making money and building careers. But if you look at the history and look at what is commercially viable, starting with operating systems, cloud infrastructure and other technologies, a lot of companies built their empires based on open-source ideas, and people made careers there building software. If you look at the last couple of years, we’ve seen a rise in the number of open-source hardware makers. There’s a lot of talent attracted to that. So that seems to work, at least partly. You also see Google, ST, TI, and others providing maker boards for promoting their chip architectures, and people jump on them and build their own devices at home. That’s sparking innovation. On the horizon for silicon is open-source hardware and the RISC-V community. It’s still very early, but it’s started to push ideas and innovation forward and attract some people. We’re seeing a lot of talented students in many universities looking at that. So, at least from that perspective, it’s already working. The goal here is to create a viable business model around open hardware, which is a challenge. One step is to create ecosystems around the technology, and there are some examples for that. Bringing the open source RISC-V cores to production level is one step that we’re very involved in. There are publicly funded projects in Germany and Europe with the RISC-V Foundation. These are things that help attract talent. That doesn’t mean the hardware will be free in the end, but it is an opportunity for us to really enable people to try out ideas out get interested in what we’re doing here. This is a big opportunity for us. Kibarian: Leaders in the industry for years have gone around saying Moore’s Law is dead. So if you are young and looking for a job, why would you go into an industry where the driving factor behind it is dying? We have done ourselves a massive disservice, without communicating what the real drivers are in our industry and what the economic opportunities are for people to make a living at this. There are a couple of really big factors that we don’t communicate enough. When you look at the fundamental limits, we’re still far away from the fundamental limits of performance per watt per dollar. When you look at the need for computing, it is always increasing. It expands infinitely, and there is tremendous innovation there. But we have not articulated what is the driver for our industry today, as we did for a long time with Moore’s Law. And going around saying this law is dead without saying, ‘Here’s the new king,’ does turn off people. We should do something about that. Related 2020 CEO Outlook (Part 1 of above roundtable) Impacts of the global pandemic and the rising cost of chip design. Chip Reliability Vs. Cost (Part 2 of above roundtable) CEO Outlook: Market shifts, higher productivity per engineer and the overhead and opportunities for security and reliability. Ed Sperling (all posts) Ed Sperling is the editor in chief of Semiconductor Engineering. Technical Papers AI Roadmap: A human-centric approach to AI in aviation March 9, 2020 by Technical Paper LinkNTSB Releases Report On 2018 Silicon Valley Tesla Autopilot Fatal Accident February 14, 2020 by Technical Paper LinkSupercomputing Performance & Efficiency: An Exploration Of Recent History & Near-Term Projections January 28, 2020 by Technical Paper LinkPlasticine: A Reconfigurable Architecture For Parallel Patterns (Stanford) January 16, 2020 by Technical Paper LinkCheckmate: Breaking The Memory Wall With Optimal Tensor Rematerialization December 5, 2019 by Technical Paper Link Trending Articles Manufacturing Bits: July 21 Intel’s next-gen MRAM; silicon oxide ReRAM; FeFETs. Startup Funding: June 2020 Sixteen startups draw $661M in June; big funding for chips in China; more quantum startups. RISC-V Gaining Traction Experts at the Table: Extensible instruction-set architecture is drawing attention from across the industry and supply chain. Universal Verification Methodology Running Out Of Steam It’s time to move up in abstraction again as a complexity overwhelms a key approach. The Race To Much More Advanced Packaging Hybrid bonding opens up a big improvement in die-to-die performance, but getting there is not trivial. Knowledge Centers Entities, people and technologies explored Related Articles Aging Problems At 5nm And Below Semiconductor aging has moved from being a foundry issue to a user problem. As we get to 5nm and below, vectorless methodologies become too inaccurate. China Speeds Up Advanced Chip Development Efforts underway to develop 7nm, DRAM, 3D NAND, and EUV domestically as trade war escalates. Making Chips At 3nm And Beyond Lots of new technologies, problems and uncertainty as device scaling continues. EUV’s Uncertain Future At 3nm And Below Manufacturing chips at future nodes is possible from a technology standpoint, but that’s not the only consideration. Metrology Challenges For Gate-All-Around Why future nodes will require new equipment and approaches. The Next Advanced Packages New approaches aim for better performance, more flexibility — and for some, lower cost. ML Opening New Doors For FPGAs Programmability shifts some of the burden from hardware engineers to software developers for ML applications. Scaling CMOS Image Sensors Manufacturing issues grow as cameras become more sophisticated.