Socionext recently partnered with Foxconn Technology Group and Network Optix to produce a new solution for edge computing, the BOXiedge. But what drives this multicore, edge-computing server and how does its SoC point to the trend in edge computing?
BOXiedge, a Development in Edge Computing
Cloud computing has its disadvantages: latency, security vulnerability, and unreliable network connection. These disadvantages have given rise to edge computing. Edge computing allows IoT devices to run AI algorithms that are trained on the cloud.
BOXiedge. Image used courtesy of Socionext
Socionext claims that BOXidege is one such edge computing solution. The fanless device includes a 24-core mini-server that uses 30 watts of power. The BOXiedge is designed for edge computing applications that require intelligence, including factory automation, retail, agriculture, and digital health.
By using many cores on a local system that is capable of running AI algorithms, users will not need to rely on a stable, secure internet connection. Sensitive data is also stored locally on the server, improving security.
A deeper dive into the processor that drives the BOXidege may illustrate how the virtues of edge computing are reflected in the hardware design.
SynQuacer SC2A11, the Chip Driving the Tech
The SynQuacer SC2A11 is a 24-core, 64-bit Arm Cortex-A53 clocked at 1GHz. The SoC includes three levels of cache with 4MB of L3, DDR4 RAM, PCIe, and LAN.
SC2A11. Image used courtesy of Socionext
It also includes various IOs including UART, I2C, and GPIO.
Each core has two processors—two levels of cache—while the L3 cache is used in the interconnects with all the cores. The SC2A11 is scalable, supporting 64 devices in parallel—a total possible core count of 1,536.
Designed for power efficiency (specifically, 5W), the SC2A11 is aimed at servers, industrial applications, and edge computing applications. The many cores are said to allow large amounts of data to be processed simultaneously.
Processing element of SC2A11. Image used courtesy of Socionext
The SC2A11 is housed in a metal lid BGA package, measuring 30mm x 30mm. This compact design makes it a useful option for applications requiring a large number of cores. Several security features are also integrated into the SC2A11, including OPAL and an ethernet processor. Socionext also offers an evaluation board for the device.
SC2A11 evaluation board. Image used courtesy of Socionext
The evaluation board includes the SC2A11 with multiple DDR4 memory slots and various I/O.
Will the Trend Toward Edge Computing Affect Hardware Design?
Microcontrollers are often geared for processing a single task.
The SC2A11, however, demonstrates that microcontrollers may take a new direction: multicore designs. While the SC2A11 is a dramatic example of a multicore SoC, it does show how designs are being influenced by edge computing.
Architecture of an intelligent video system for an edge AI server with Witness VMS. Image (modified) used courtesy of Socionext
The demand for more secure systems and lower-latency devices (such as those required in self-driving vehicles) emphasizes the utility of parallel designs. It’s likely that this focus on security and latency will influence future microcontrollers, including the integration of multiple cores and AI co-processors.