Moore’s Law, Supply Chains And Security

Big changes ahead, from design to deployment.


The debate about the future of Moore’s Law continues, while other parts of the industry look for alternatives. In between, supply chains are being pulled in multiple directions, with safety and security often in the middle.

All across the semiconductor industry, significant changes are underway. Some of these have been in the works for some time. Others are new or accelerating faster than anyone would have predicted a year ago.

“Moore’s Law continues to roll right along,” says Joseph Sawicki, executive vice president for IC EDA at Mentor, a Siemens Business. “After years of confusion between Moore’s Law and Dennard Scaling, it’s clear that Moore’s Law continues as we get new nodes that allow further integration, larger numbers of transistors on-chip at lower cost per transistor. It all continues. We are seeing roadmaps already talking about 2nm and 1.5nm process nodes targeting delivery 6 to 8 years from now. Dennard scaling will continue to present issues in terms of power scaling and consumption, and the need to optimize algorithms for chips to ensure that ICs don’t burn up and die.”

The companies going in this direction are not necessarily the same ones pushing it in the past. “Chasing Moore’s Law with decreasing feature-size, especially below 7nm, becomes an expensive value proposition,” says Vic Kulkarni, vice president and chief strategist for ANSYS. “This is an answer for the privileged few large, rich Tier 1 semiconductor and bespoke silicon system companies like Tesla, Microsoft, Google, Facebook, and so on.”

But that is only half of the industry push. “Even with Moore’s Law rolling right along, in 2020 we’ll continue to see more companies leveraging various 3D-IC technologies, particularly in packaging, mixing various types of dies – processors, analog, RF, memory and potentially sensor– in one package,” adds Sawicki. “This allows companies to both manage yield and integrate multiple technologies into one device.”

Many designs require both advanced packaging and advanced node technology. “As more semiconductor companies look for alternatives to Moore’s Law scaling and begin to fully realize the benefits of modern-day multi-chip/chiplet packaging including stacking, expect to see a dramatic reduction in single monolithic SoC design starts,” says John Park, product management group director for IC packaging and cross-platform solutions at Cadence. “Instead, expect 2020 to launch the age of ‘More Than Moore’, namely multi-chip/chiplet packaging. Foundries and OSATs will continue to battle for market share, driving cost — the key metric — lower for state-of-the-art multi-chip/chiplet packaging. This will greatly accelerate the transition to multi-chip/chiplet packaging and heterogeneous systems as an alternative to CMOS scaling. The evolution of Moore’s Law is about more than just the limitations of physics. Chasing the latest design node is expensive, hard, and requires large design teams. Big chips typically don’t yield very well, either. Fewer and fewer companies want to take on these challenges when a viable alternative in packaging is so readily available.”

There are plenty of high-end markets driving both of these trends. “High end markets such as high-performance computing, gaming, including virtual reality and augmented reality, and networking will be the drivers for 2.5D and 3.D design innovation,” says Anna Fontnelli, CEO of Monozukuri. “The year 2020 represents the mid-point in a five-year trend that should see a 27% compound annual growth rate between 2018 and 2023 for 2.5D and 3D designs, according to a consensus of industry analysts.”

The limitation for 2.5D and 3D has been cost. “In 2020 and beyond, we expect industry players to drive down the cost of silicon-package-system development and time-to-market,” adds ANSYS’ Kulkarni. “Chiplet-based architectures will meet the challenges in various vertical markets, including 3D-sensor cameras for autonomous, IoT/IIoT, efficient HPC-driven AI and compute applications, 5G base station designs, and so on. We see the acceleration and adoption of this ‘More-than-Moore’s (MtM) Law’ already creating an industry-wide initiative, which is expected to have significant impact in 2020. Under the auspices of Open Compute Project (OCP), industry leaders are energized to work on defining and developing a chiplet-based architecture, called ODSA – Open Domain-Specific Architecture. A loosely connected consortium of over 50 companies is now defining new interfaces, link layers, a marketplace and exchange, an open ecosystem of trusted chiplet providers, and providing an early proof-point for industry-wide adoption. In 2020 we will also closely watch the development of another initiative, AVCC, Autonomous Vehicle Computing Consortium, a group of automotive and tech industry leaders who will be defining a reference architecture and a platform to meet the autonomous performance goals within the power, thermal and size constraints of a vehicle.”

Fig. 1: Open interface for chiplet-based design. Source: ODSA

Supply chain
Chiplets and 3D integration require some changes in the supply chain. “Continued consolidation and collaboration across the semiconductor industry, along with derivative designs and systems that incorporate multiple vendors’ IP, will result in increased IP reuse,” says Simon Rance, head of marketing at ClioSoft. “Finding and keeping track of reusable IP will increasingly become more difficult due to its multi-faceted nature, i.e. revisions, configurations, legal terms and conditions, etc. Growth around IP reuse will compound the problems already facing IP vendors, such as IP theft and data leakage. We will see significant changes in the IP world, both from the user and the vendor side as IP reuse continues to grow in 2020 and beyond.”

International tensions also are rattling the supply chain. “With trade tensions and the cybersecurity attacks linked to hackers and state actors, leading industrial companies will be very concerned with the security of the supply chain,” says Jo Jones, technology communications manager for Imagination Technologies. “Manufacturers of integrated circuit (IC) products have always had to deal with a gray and a black market that could cost them billions of dollars. However, the issue of the security of the supply chain is a more fundamental problem. Large western industrial concerns, either the manufacturers (e.g. Siemens, Bosch, GE) or the services providers (e.g. utilities, cities), will demand a clear understanding of the origin of all the parts in their systems, all the way down to individual integrated circuits. Supply chains will have to be certified to comply with specific policies.”

Safety and security
There is increasing attention being paid to both safety and security. “In addition to the software security and trust zones, we see the increasing number of chip and system design starts with heightened focus on hardware-based security,” says ANSYS’ Kulkarni. “These are requiring innovative ways to mitigate side-channel attacks (SCA) on any device that communicates with its surrounding objects, may it be a satellite, a drone, a plane, a ship or ground-based defense infrastructure.”

Others agrees. “There is a growing awareness that security is critical for all SoC designs,” says Rupert Baines, CEO of UltraSoC. “This is seen with growing impact of ISO 21434. Because of this, security will be an overriding concern in 2020. it’s a rather complicated picture. ISO 26262 is concerned with safety, and what happens when things go wrong. The upcoming ISO 21434 standard is next and is focused on cybersecurity. SOTIF (safety of the intended functionality) is about making sure the system is safe when functioning correctly. An embedded analytics infrastructure can provide information and analysis that allow designers to more easily satisfy the functional safety, risk assessment, testing, reporting and traceability requirements of standards such as ISO 26262.”

Security is also becoming a national-level concern. “IoT security will become a national priority,” says Imagination’s Jones. “The U.S. is quite aggressive in this domain, and it is building partnerships between the government, private companies and universities to define policies and next-generation cybersecurity technologies (e.g quantum crypto). The Chinese are well equipped in this area with a centralized government working with universities and industry. The Europeans are not as well equipped, as they tend to be fragmented and work on a national basis. Interestingly, most nations will adopt the same technologies, and the winners will be the most adept at creating quickly efficient policies with well-organized management structures.”

Along with this are increasing privacy concerns. “Privacy concerns and proposals to expand protections around facial and identity detection will continue to rise and will drive innovation into technologies that mitigate data collection in favor of local processing,” adds Jones. “This will also lead to changes in the types of data processing. Instead of using facial recognition to detect known persons of interest, the technology will be expanded to detect expressions, emotions, pose, gait and other tell-tale signs that infer a person’s intent and flag appropriately. Big Brother is watching, and will soon know if you’re up to no good in real-time.”

The tensions between the United States and China are having significant impact on multiple levels. “The tariff wars with China and targeting of specific companies, such as Huawei/HiSilicon, may have poked a sleeping giant,” says ClioSoft’s Rance. “China will not be held hostage to the U.S. and its politics. They will redouble their efforts to become self-reliant in crucial technology, not only in semiconductors, but also in EDA software. While Chinese EDA software companies, such as Huada Empyrean, may not be globally competitive, the Chinese government will provide incentives to help improve their market share in China. U.S. EDA companies are likely to lose market share in China, at least in the lower end of the market.”

There also may be unintended winners. “China spends more on importing semiconductor chips than it does oil, illustrating its deep dependency on foreign chips,” says Jones. “The 16-month U.S.-China trade war hit chip companies directly. European suppliers have been put in a strong position to support both China and the USA in their technology advancements, creating some interesting opportunities in the market while also causing an acceleration of China home-grown solutions. In 2020, the importance of China as a customer base will continue to grow. The ready availability of funding in that region will also help this acceleration.”

The genie cannot be put back in the bottle. “The geopolitics and trade wars will get even more stressful in 2020,” says UltraSoC’s Baines. “This will impact other countries. Security is a genuine issue in the trade war, and it is not just an economic thing. There are genuine concerns when governments start obliging system providers to put unknown binaries into their software. As a consequence, China is investing dramatically more in local technology and IP. It’s happening, and there is plenty of evidence out there.”

Leave a Reply

(Note: This name will be displayed publicly)