Machine learning (ML) and neural networks will continue to be a hot topic in 2020. One of its uses is forecasting in everything from when an electric motor will fail to how stocks will perform in the current market. It might be handy to use ML to create this article. However, such forecasting rests in my hands this time around.
ML hardware acceleration for training and deployment will get lots of coverage. I’m more interested in checking out the increased use of ML on standard platforms, though, including applications running on Cortex-M and RISC-V platforms. Most vendors have an overarching approach to artificial intelligence and ML that allows applications to be written to a framework rather than a particular hardware platform. Available performance and capabilities will limit a design and functionality, but not whether ML can be used.
While artificial-intelligence advances will dominate much of the technology discussion in 2020, it’s hardly the only new tech that will affect development decisions in the coming year.
The consolidation of RTOSes by vendors and IoT solution providers is making it easier to build IoT applications, but it’s also pulling developers into walled gardens. Related support like smart-speaker integration and use of ML in the cloud also requires the use of unique frameworks and tools.
Networking technology will see as many advances as ML, possibly more. 5G will of course be at the forefront as deployment begins, yet changes are blowing in the wind. The IoT aspect of 5G will just be starting to emerge, but plenty of wireless solutions with different characteristics are equally important. This includes technologies like Sigfox, LoRaWAN, NB-IoT, and Bluetooth mesh. These will actually be more interesting to developers because their standards and components are more mature. Though it’s fun to look at the latest tech, it’s more useful to be able to use it.
Ethernet, PCI Express (PCIe), and other fabrics will see a host of new standards and products in 2020. PCIe is the base for standards like NVMe as well as CCIX and CXL. NVMe has become the desired storage interface for flash drives with M.2 sockets on most motherboards, but SATA and SAS still handle large populations of drives. And hard drives aren’t dead yet. Gen-Z’s memory-semantic fabric will continue to gain ground.
Power-efficiency improvements for microcontrollers and microprocessors will continue even as the top-end clock speeds remain limited. Multicore is here to stay in a big way with single-core solutions being a relegated to microcontrollers. Even there, it’s not uncommon to have a dual-core solution with a security or peripheral management processor alongside the primary core.
Security will be getting less lip service and more practical developer support. The standard software development kits will make it easier to take advantage of secure enclaves and encryption acceleration that once required direct interaction with the chip vendor or a security partner. The latter will still be useful as security really needs to be an end-to-end, start-from-the-base design project, not an add-on. Remember, secure boot only guarantees that the booted core is what you thought it was. It doesn’t provide protection for security bugs that are part of the system or application.
RISC-V continues to gain steam with some interesting platforms becoming available in 2020. Shifting politics and a desire for a more adaptable platform is driving adoption even as the instruction set continues to evolve.
Keep an eye on FPGAs. The tools are already making it easier to take advantage of their flexibility without needing to be an FPGA hardware expert. RISC-V FPGA SoCs will also be available in 2020. Developers that overlook the tiny, flash-based FPGA solutions are likely to be beaten by their competition that exploit these platforms.
That covered a lot of ground, but it’s what we do at Electronic Design as we track and present the latest technology for developers.
