Edge processing ramps up as computing cores continue to break through
Время обновления: 2021-12-01 14:19:12
Machine learning technology based on edge computing is becoming more and more important today with the rise of the Internet of Things, where the processing of important data needs to be closer to where the data is initially located. In the past two years, the theme of artificial intelligence development has been very clear, in the landing application of domestic and foreign manufacturers are grasping the deployment. Among the many AI technologies, both the upper layer of algorithm applications, or products, ultimately rely on the underlying arithmetic power guarantee, that is, the "core".
Whether in the new generation of wearable devices, or in the field of robotics, edge processing solutions seem to be more suitable for these applications that require real-time, secure, and autonomous response. The different environments in these diverse scenarios have different power and performance requirements for the chip. How to ensure the energy efficiency, security and connectivity to cover different scenarios is a major challenge, while the execution of deep learning tasks on the chip to obtain high-quality data is no easy task. In the core of edge computing, how do the mainstream manufacturers break through?
NXP cross-border series MCUs
NXP has an edge computing platform, EdgeVerse, which includes a comprehensive portfolio of processors, microcontrollers, and signature software products. i.MX RT series crossover MCUs are part of the EdgeVerse edge computing platform, a large family of products that combines the Arm Cortex-M core, real-time processing capabilities, and MCU availability.
From the original RT500 series with the Cortex-M33 core to the current RT1170 with the M7 core. the crossover series of MCUs are outstanding from performance to scenario coverage, and it can be said that this crossover product drives the convergence between application processors and MCUs.
RT1170 crossover MCU set a new record with a speed of 1 GHz. RT1170 dual cores are 6468 CoreMark, Cortex-M7 with a main frequency of 1 GHz and Arm Cortex-M4 with a main frequency of 400 MHz.
RT1170 breakthrough cross-border MCU combines extremely efficient computing power, a variety of media functions and a variety of real-time functions. The RT1170 is also equipped with large-capacity, low-latency on-chip SRAM memory with up to 2MB of SRAM with a 512 KB TCM for the Cortex-M7 and a 256 KB TCM for the Cortex-M4. This makes the RT1170 extremely responsive in real time, with latency as low as 12ns.
The series also reduces power consumption as much as possible, with the integrated DC-DC converter reducing the dynamic power consumption of the RT1170, while the chip itself has a low-power mode at 21 MHz. In addition to high-performance computing, real-time and power consumption level, this cross-border MCU integration is also high enough. The already powerful multimedia performance is realized with GUI and enhanced HMI, accelerated by OpenVG graphics and up to 500MHz main frequency.
The security aspect is reflected in the entire EdgeVerse, with secure boot and encryption engines and AES decryption all reflected in this crossover MCU. It can be seen that in the field of edge computing, NXP has made great efforts to build its own NXP edge computing ecology through the powerful performance of cross-border MCUs and a comprehensive supporting system.
RA edge computing MPU
RA in the edge computing using the old friends we know - RZ series. This time the burden of edge computing falls on the RZ/A series, a 32-bit Arm Cortex A9-based processor with up to 10 MB of on-chip SRAM that can buffer up to WXGA resolution graphics. In particular, the A2M under the series, with dynamically configurable processor technology, is well versed in embedded AI high-speed image processing.
The RZ/A2M features high capacity internal RAM (4 MB), a dynamically configurable processor to provide 10x image processing performance to support edge computing functions, and powerful connectivity via MIPI/LVDS/2-port Ethernet. the A2M provides enough bandwidth to process all the information in the central data center.
The RZ/A2M uses an Arm Cortex-A9 CPU of up to 1000 DMIPS, coupled with 10 MB of on-chip RAM, HMI on-chip hardware, and QSPI flash memory to provide higher performance than an MCU while reducing system cost. Combined with its high-bandwidth 128-bit wide parallel data bus, the RZ/A supports faster graphics and digital audio signal processing.
RA's layout on edge computing is more inclined to embedded vision applications, broadening its edge computing ecosystem with powerful graphics processing capabilities.
Write at the end
Edge computing effectively makes up for the shortcomings of latency and data privacy in data desensitization, data awareness and real-time decision making, and also reduces high IT infrastructure costs. In the long run, edge computing with a strong core will take IoT and other related applications to a whole new level.
Следующий: How small are the transistors on a chip?