The emerging Internet of Things (IoT) industry is a collection of complementary products and services that enable efficiency and cost optimization across multiple industries. While it does not have a vertically oriented value chain, it spans multiple industries and markets such as industrial automation, automotive, healthcare, environmental monitoring, etc., and the use cases within these industries are very diverse. At the front end of the application are also the end nodes or sensors, which monitor the environmental conditions and pass the data into the chain. These end nodes will be decentralized in various industries. The architecture of the design processor is basically a mainstream exclusive market such as arm, and the share of RISC-V is very small, so under normal circumstances, the market rarely analyzes the relevant field profile. But as the Huawei incident continues to burn, the U.S.-China trade war and the global international environment have begun to change, and the technology sector has begun to re-examine the relationship between these two.
The rise of the custom processor
Remembering the discussion from a long time ago about the future of the processor wars, COTS (commercial off-the-shelf or technology) processors aren't suited to building these endpoint nodes because the latter are application-specific. Instead, companies generally favor custom processors because it provides the flexibility to assemble only the required components. These components include analog sensors, DSPs, or proprietary IP, among others. In addition, a custom processor can significantly reduce BoM cost and chip size, thereby minimizing power consumption. It also has the added benefit of helping companies differentiate their products from those of their competitors. Overall, the low cost of entry and ubiquity of the IoT industry will encourage many startups and small companies to build products for cold applications. Also, by customizing the processor, these companies can further optimize costs.
The Wrong Moore's Law
In addition to the huge boost that the proliferation of IoT devices can give to custom processors, another influencing factor is the dubious existence of Moore's Law. For more than five decades, Moore's Law has been a self-fulfilling prophecy. Semiconductor companies have worked hard to make this law true whether or not the market needs high-performance processors. So there have always been innovators and early adopters eager to use products based on leading process nodes. However, it will take time for the mass market to align these new products. Moore's Law ensures that technology plays a dominant role with high performance, low power, and reduced cost.
However, the economic equilibrium guaranteed by this law is currently failing. The balance of costs does not hold as leading process node designs become complex with long lead times to commercialization. The quest for cost optimization is forcing the industry to look for alternatives, as shrinking nodes are no longer economically viable. In fact, customized processors are the answer as they can significantly reduce BoM costs. Billions of end nodes do not need leading process nodes, having custom processors on mature nodes is enough.
ARM's response
ARM has a monopoly on the smartphone processor market. There is no dominant architecture in the embedded and IoT space, and ARM is poised to fill this gap as it has powerful CPUs and IPs that offer a wide range of features, performance and price options. With a unique licensing business model, specifically the DesignStart license on the Cortex-M0, ARM enables custom processor designs at a low cost with lower risk. The program is very useful for startups and small companies as they have access to proven architectures and IP at low licensing costs, and combined with an extensive IP ecosystem, software support and silicon partners, can dramatically reduce the time to market.
So how do we further optimize the cost of custom processors?
RISC-V
Open source software (OSS) has played a critical role in democratizing the software industry.One of the most popular examples of OSS is a Linux operating system.OSS enables innovation and differentiation at a low cost of adoption. This allows small companies and startups to build products based on OSS such as Linux. A large community of developers supports software development so there is no risk of vendor lock-in or proprietary technology becoming obsolete. The collective effort of the community ensures a large ecosystem while benefiting all users.Linux has gained immense traction in a variety of applications such as embedded, PCs and more. As more and more users start using Linux, adding more features and utilities, the network effect can also be put to good use.
RISC-V extends the open source movement to the CPU ISA. it is an open source ISA, license free and royalty free. It is also because RISC-V has no license that the ISA can be used to build custom processors with zero licensing costs.RISC-V is gradually building an ecosystem. At Embedded Electronics & Industrial PC Applications 2017, RISC-V demonstrated its vast ecosystem through FPGA solutions, security IP, debugging infrastructure, and more.
Few ARM customers have started using RISC-V to design custom processors. Now, SoC design companies can develop custom processors at a lower cost without paying license fees. With some NRE investment, these companies can develop SoCs and manufacture them in fabs. As a result, the processors will also be less expensive than those based on ARM IP. On the face of it, an ideal candidate is likely to become the dominant ISA in the IoT industry. with custom processors and zero licensing costs, RISC-V looks a lot like that winner.
A look at RISC-V "free"
Linux has been very successful in billions of product deployments. While there is considerable effort and expertise required to use Linux in commercial products, the benefits will far outweigh the man-hours. Linux can provide very good flexibility, while the large community provides a good ecosystem for the operating system and extensive support for peripherals, third-party software, and so on.
However, there is still a big difference between open source concepts and chip design due to the fundamental differences between software and hardware. Unlike software, which takes time and effort to develop, hardware involves tangible components that need to be paid for by someone, and secondly, you can rework the software many times after testing the hardware and emulator. For a fairly small cost, you can reduce the time and effort. However, errors in the hardware can also cost you a million dollars! Multiple iterations of the processor can dramatically reduce costs. Overall, hardware design is more complex than software development.
Let's consider the case of open source RISC-V. In a SoC, the CPU IP is only one part; many other physical IPs and peripherals are needed. Therefore, a huge IP and EDA ecosystem is needed around the CPU IP. But you can only get CPU IP without license credentials; yet the ecosystem around it has disappeared.IP vendors should see a viable business case to add support for RISC-V to their portfolio. Assuming there is a strong community supporting RISC-V, it provides all the IP and tools needed to build SoCs. But the question remains whether companies building custom SoCs will take the risk of using community-supported ISAs? Failure could lead to multiple flows, which would add significant cost. Overall, designing SoCs is complex and requires good expertise in a number of areas including implementation, physical design, and packaging.
With ARM ISA, most of the issues mentioned above are mitigated. Instead of the community support provided by an open source ISA, you have access to proven IP, a robust ecosystem (software, cloud services, security solutions, chip vendors, foundries) and committed support. This reduces design complexity significantly, though it still requires some specialized SoC design to build custom processors.
Who will build RISC-V-based SoCs?
The idea of open source is disruptive because it gives companies with limited budgets a level playing field to compete with the big boys. While the concept of an open source ISA is revolutionary, it may not have a disruptive impact on the democratization of chip design.
In my opinion, it is unlikely that small companies and startups looking for some niche application in the IoT space will invest the time, effort, and money to build custom processors based on a community-supported ISA because they will have to verify that the entire system meets their specifications. Instead, using a licensed ISA is a safe bet because they have access to a validated system complemented by a robust ecosystem.Multiple streams of SoCs can add significant cost. Mature ISAs have some initial cost, which is a good starting point, but it's not a free-for-all for a mature ISA.SoC design is not a core component for them, so hiring a diverse chip design team may not be a pragmatic decision. Due to the widespread adoption of ARM across the industry, the design part can be outsourced to smaller companies that specialize in ARM-based SoC design. EDA tools and fab costs are high. eda vendors and fabs already support ARM-based IP; they should also see the economic benefits of adding support for RISC-V. Until RISC-V reaches critical mass adoption, it's like a chicken-and-egg situation. Multi-homing increases costs for any company, whether it's a fab, an EDA vendor, a design house or an application developer. Low volume operations can attract higher rents. So all of these cost overheads must be considered when building a RISC-V based SoC.
Market leaders in SoC design will certainly develop RISC-V based SoCs as it can increase purchasing power by replacing ARM. However, I don't believe these companies will be interested in working with low-volume customers who need customized processors. Its huge overhead makes selling millions of standardized SoCs make a lot of business sense.
In summary, in my opinion, RISC-V in its current state cannot significantly disrupt the semiconductor market structure. One of the key advantages of the open source movement over licensing entities is that it minimizes the barrier to entry into the market by providing a good enough foundation. While RISC-V will provide the flexibility to build custom SoCs at a low cost, the ecosystem is not yet ready to embrace it. The entire semiconductor industry needs to synchronize to make RISC-V successful.
Conclusion
In this international climate, I believe that RISC-V would have been smart enough to anticipate the above issues and capitalize on the windfall, and many of the issues would have been solved internally. In my opinion, RISC-V should focus on one segment, such as IoT end nodes or something else, and then provide a compelling complete solution for that segment, as well as the overall ecosystem, rather than focusing on the entire IoT and embedded industry. Once they achieve mass adoption in one segment, it's easier to spread to other segments because new users have a good case study or example to work from.
What more does ARM need to do to be considered a leader in embedded and IoT? I don't have any answers for this because from an external perspective, ARM looks pretty good right now with a large installed base and a good upward trend for the future. Extending the DesignStart license to other Cortex-M IP would be a good option for further applications though. However, the main core should still be a strong ecosystem of OS support, cloud services, security, IP, debug toolchain, EDA, silicon partners, etc. All of these play a vital role in building custom processor-based products at low cost.
Low cost and customization are often mutually exclusive. Any ISA that targets both ends of the spectrum will play a dominant role in the IoT industry. Of course, with the establishment of the RISC-V Foundation, there have been a number of companies and research organizations that have joined the RISC-V camp to explore the viability of RISC-V in the future. Currently participating companies are IBM, NXP, Western Digita, Pfizer, Qualcomm, Samsung, Google, Tesla, Huawei, Alibaba, and more than 200, and the U.S.-China technology war may be the catalyst for the rise of this model, many years from now to look at the history of science and technology to this moment to mark a heavy one.