Why Semiconductor design is set to look very different
If 2020 propelled the semiconductor market to new levels of demand, just wait for what 2021 and beyond have in store.
Fabs are already at capacity and semiconductors can be found in more products and systems, powering everything from personal devices to self-driving vehicles. Despite the impact of the pandemic on the global economy, International Data Corporation (IDC) says that demand for semiconductors remained strong, fuelled by the growth in cloud computing and devices to support remote work and learning.
Its own research bears this out. According to IDC’s Semiconductor Applications Forecaster (SAF), in 2020 worldwide semiconductor revenue grew to $442 billion, an increase of 5.4% compared to 2019. IDC has now forecast that the semiconductor market will reach $476 billion in 2021, which would mean a 7.7% year-over-year growth rate.
There are some clear trends within the industry too. For instance, PWC has predicted that the market for AI-related semiconductors will reach $30 billion by 2022, representing an AGR of almost 50%. At the same time, there is very much still a place for traditional system on chips (SoCs). Memory chips are expected to carry on holding the biggest market share through 2022, and silicon chips will dominate for the next couple of decades.
New technologies and players
Other factors shaping the future of semiconductors stem from technology directions, such as a burgeoning focus on open source hardware. As that gains momentum, it is going to change how organisations think about design, and will encourage a more collaborative and partnership approach to development.
The IoT is driving demand for more cost-effective semiconductors and on a larger scale. Similarly, 5G enables the bandwidth to finally support vast, interconnected infrastructure, for example for transportation, combining inputs from multiple API-connected sources. To achieve the scale required, standards-based reusable and sharable IP is going to be key to meet design requirements, but in a way that supports distribution and collaboration, while also meeting safety and security requirements. Open source could be a key to achieving that scale.
The nature of the players involved in the industry is shifting too, such as the introduction of vertically-integrated systems and non-traditional semiconductor firms, who may prefer to create their own devices and platforms for greater control. A good example of this being Apple’s M1, a processor designed for Mac, which furthers their commitment to building key functionality internally. This adds another interesting dimension to the industry, increasing competition and sources of innovation.
As much a necessity as a trend, semiconductor design teams have had to embrace remote collaboration on an unprecedented scale. Of course, to some extent that was already happening, but the pandemic made it a necessity for survival in many cases.
In these virtual environments, problems with workflow processes became impossible to ignore, with sharing of IP happening on a much wider scale. Addressing issues around collaboration and security has become a priority, and once organisations have achieved that, they will be equipped with experience that can be used to improve processes irrespective of location, whether in the design office or working remotely.
Regardless, the shifting nature of the semiconductor market reinforces some age-old challenges for designers: keeping pace with change and complexity, controlling costs, making sure a project stays on track and in line with requirements, and then meeting delivery deadlines. Many semiconductor product launches do not meet the original launch date. Contributing factors can include: the difficulty of collaboration across remote and dispersed teams, company acquisitions leading to design silos, not addressing management of exploding design data sizes, or keeping up with increasing complex design environments.
IP reuse versus over-sharing
IP reuse has long been talked about as the solution to keeping up with the sheer scale of developments required, avoiding unnecessary reinventing of the wheel, speeding up time-to-market and drastically reducing costs. A variety of assets can be reused include source code and binaries for software, plus hardware IP like arm processor cores.
The theory is sound, but successful management of IP reuse is another matter. Many organisations have multiple systems, including shared drives and source code management, to store and track all these files, but that makes it difficult to manage — let alone reuse — these files. Fortunately, there are a variety of tools and techniques that can help designers overcome those barriers, but those have to include a strong structure and control around reuse. While sharing of IP is the route to faster development, it can also bring risks around management and security. Traceability of IP and meta-data — who has access to what, where, and how they are using those assets — becomes critically important to the success of projects.
There is a balance required between sharing and over-sharing, which is why a combination of traceability, visibility, and access control are essential for modern semiconductor design. Of course, manual traceability is widely adopted, but to address large and complex projects, more organisations are using tools that automate much of the process, from requirements through to design and verification. This makes it easier to track which IPs have been used, where, and when, and in turn avoid costly respins in design. Changes in requirements can be surfaced and communicated earlier and more clearly. Traceability also supports better remote collaboration.
In addition, creating a ‘single source of truth’, or data management platform, can unite all the software and hardware components within a project. Even small files become simpler to locate and reuse, and there may be millions of those within an enterprise, not just the big complex designs. A single source of truth is typically based on using a version control system, which provides real-time and historic visibility into all the assets involved, while allowing users to carry on working with their preferred tools and systems.
A single source of truth is also critical to identifying configuration issues across systems, in hardware and software, that are masked by the complexity of the systems. In turn, that makes it easier to bring together collaboration across remotely located teams, both internally and externally.
Overcoming IP leakage
With comprehensive traceability and visibility in place, then it becomes more feasible to identify and prevent IP leakage, a perennial and costly concern in the semiconductor industry. Some of the root causes include dispersed collaborators, inadequate control over who can view and download IP, and users inadvertently exporting IP to an unauthorised source. With semiconductor markets becoming ever more global, mitigating IP leakage is of paramount importance.
Semiconductor design is going through a period of unprecedented change. The events of the last year, however, have given us a blueprint of how to handle the needed changes to collaboration and culture. The good news is that there are tools available to provide the needed traceability to facilitate IP reuse at scale while also providing the security and communication layers for effective collaboration.
Semiconductor design will look different going forward, but the changes facilitated by a global pandemic will help drive major technological innovations moving forward.