A Guide to Choosing the Right IoT Processor for Your Company

The Internet of Things encompasses literally anything, from wearables to automobiles and everything in between. While form factors, connectivity, functionality — and myriad other factors — may differ from one “thing” to another, every connected device must have a processor. However, even processors can vary widely, and companies have to make a number of decisions when selecting one. 

“The processor is the heart of a semiconductor,” says Bill Ray, Gartner senior research director. “It does the mathematics. It runs the program.” As an example, Ray said a sensor in a thermostat may collect temperature data, but that data can’t go directly to a communications chip. “The processor takes the data and packages it and sends it to the communications chip. If the data hasn’t changed, the processor may determine that there’s no need to send it again,” he says. 

An IoT processor is not to be confused with a system on a chip, which “contains a radio chip, memory, processor and connectivity all bundled up into one piece of silicon,” Ray says. “Currently, with the IoT, we see a lot of variation in processors.” 

[Attend the RISC-V Summit to learn about the disruptive force driving the next generation of hardware, software and IP, Dec. 10-12 in San Jose, California. View the conference agenda].

That is due to the diversity of connected devices. “Today, there’s a chip in everything. Those new frontiers have different workloads than a server in a data center, or a desktop, or a cellphone. You have constraints on size, memory and performance demands as well as lower energy consumption. The level of innovation is going through the roof because there’s so many different places to stick a chip, and the barriers to entry are going down,” says Calista Redmond, chief executive officer, RISC-V. 

Most companies are experimenting with IoT solutions. According to Ray, they don’t know what connectivity technology to use, whether they need to have GPS functionality, how many sensors they need, etc. “We see separate processes deployed,” Ray says. “They’re produced in small quantities and used in small-scale deployments. When companies work out which apps will be popular, then they will start making custom chips for those apps, and those will be SoCs.” 

Selecting an IoT Processor

When choosing a processor, the main factor to consider is how much processing power you need, which is dictated by the software you want to run. “You have to write software for the sensor. If you go to 32-bit, you can use Linux and hire loads of cheap programmers. It’s easy and quick to develop for,” Ray says. “If you go with 16-bit, you have to hire someone who knows a low-level programming language, and that will cost more. In the long term, it will also cost more to maintain the software.” 

The software has an impact on power consumption, and that must be considered as well. “The more processing power available, the more complex OS you will write, and more functionality you have. That will increase power consumption, which can be an issue depending on the application,” Ray warns. 

In addition to processing power, developers should consider the processor’s instruction set architecture. “The instruction set architecture is the contract between the software and hardware,” says Tim Whitfield, vice president of strategy for automotive and IoT line of business at ARM. “In choosing architecture, you’re really choosing a software ecosystem.” 

He continues: “ARM has a huge ecosystem of software development tools, platforms that enable people to go through and take a concept or idea to a functioning system. Software is an expensive and skilled part of the design process. The more you can reuse and build on platforms that already exist, the quicker you can securely deploy your device into whatever system you want to deploy it into,” Whitfield explains.

RISC-V is an open source hardware ISA. “RISC-V has a simple small set of instructions which means that end users aren’t burdened with the last 20 years of instructions added into something,” Redmond says. “It’s fully transparent so they can see where they can add on the custom or OS extensions that fit well for them.” 

RISC-V is also focused on supporting developers with software. “Up to this year we were focused largely on hardware building blocks, and now we’re doubling down on the software ecosystem that goes around that, from the initial firmware code that talks to the hardware and the operating system, to the apps that ride on top of that,” Redmond says. 

RISC-V vs ARM

The processor ISA market is primarily dominated by ARM and RISC-V. “ARM really has the market in IoT architectures and is now being challenged by RISC-V,” Ray says. “RISC-V is enthusiastically being adopted particularly in China where licensing technologies from other companies is difficult. It’s definitely growing in popularity and we expect it to be a big challenger to ARM for low-end devices.” 

Both ARM and RISC-V, however, can support a broad range of use cases, from the smallest, most power-efficient endpoint devices to the most complex, high-performance device doing data processing and analysis at the edge.

ARM offers a variety of processors and licensing models designed to address the needs of any business, whether an established silicon partner, a more mainstream OEM or a startup. The company is no stranger to open source. Its embedded operating system, Mbed OS, is open source. “We’re open to looking at how we embrace open source hardware, but we have no plans to open source our processor architecture or processor design, and we don’t believe that’s what our partnership wants either,” Whitfield says.

Whitfield said the industry is moving to a fifth wave of compute, in which computing is moving to a distributed model. “We’re seeing an evolution where more processing moves out to the edge, and that means you need different kinds of compute engines and business models to enable people to innovate. Our partners talk more to us about how we can evolve technology to make sure we can solve the technology problems they have,” he says. 

Enabling innovation is a theme shared by RISC-V. “You don’t need to invest billions of dollars to make your own processors. Having building blocks from RISC-V accelerates the time to market, reduces development, reduces burden of having the engineers on staff to accomplish design points, and it alleviates some of the strategic risk, because you have so many participants in the community building off the same building block,” Redmond says. 

While RISC-V offers the flexibility to build your own processor, the model is not limited to that. “Our model is not to burden the end user with building all their own. Just like any walk of life, you can get concierge services. There’s an entire ecosystem to help support you, whether you want to do it all yourself or get help to accomplish a specific goal,” Redmond says.

Beyond the Processor 

Decisions regarding processors don’t end there. Once companies mature past the experimental phase of their IoT deployments and determine their requirements, Ray says they may either find a SoC that meets those requirements, or they may choose to make their own. “One of the things we’re forecasting is a significant increase in companies making their own semiconductors,” he says, citing Microsoft as an example of one company that has started making their own chips for endpoints. 

“Once production quantity increases and the IoT solution is successful, there’s a point at which it becomes economic to make a SoC that has exactly the functionality you need. There are two companies that will manufacture a chip and then you have your own processor – it lowers the cost and makes it more difficult for competitors, because you own the product and your own chip. It’s more difficult for people to make a copy of it,” Ray says.