Variscite in Israel has developed a system on module (SoM) for efficient machine learning on edge devices using the NXP i.MX93 controller.
The VAR-SOM-MX93 is the industry’s first implementation of the ARM Ethos-U65 microNPU neural processing unit. It is aimed at applications in industrial, IoT, smart home, automotive systems, and the wearables markets.
Using a system on module simplifies the development process without having to worry about the board design.
The VAR-SOM-MX93 has a flexible energy management architecture for efficient processing using the 1.7GHz dual Cortex-A55 cores, 250MHz Cortex-M33 real-time co-processor and Neural Processing Unit (NPU). The NPU has 256 multiple accumulate (MAC) units operating up to 1.0 GHz providing 2 OPS/MAC for embedded machine learning.
The evaluation kit of the VAR-SOM-MX93 includes the SoM with Symphony Board. The kit also includes a Low Voltage Differential Signalling (LVDS) display with a touch panel and display bridges.
The SoM adds camera interfaces and high-quality image processing (MIPI-CSI2 and MIPI-DSI 1920×1200 24-bit) plus a range of connectivity options: certified dual-band WiFi, BT/BLE, dual GbE, dual USB2 and CAN-FD. Security is implemented via NXP’s EdgeLock secure enclave, a preconfigured, self-managed, autonomous security subsystem.
“As a company dedicated to be first to market with innovative solutions, we are proud to deliver the VAR-SOM-MX93, which provides customers with a high-quality, future-proof SoM that enables ML capabilities, security and energy efficiency required of today’s smartedge devices,” said Ofer Austerlitz, VP Business Development and Sales of Variscite. “This is the next generation of NXP processors and Variscite SoMs, for edge and endpoint computing that pushes the boundaries of what was previously possible.”
Both the evaluation kit and the SoM are now available to Variscite partners and alpha customers receiving priority fulfilment says the company. The VAR-SOM-MX93 is included in Variscite’s long-term longevity plan and will provide an expected availability of 15 years.