Microchip has announced custom-designed firmware for its MEC1723 Embedded Controller (EC) to support NVIDIA DGX Spark personal AI supercomputers. Microchip says the firmware is tailored to optimize the MEC1723 for system management of AI workloads on the NVIDIA DGX platform, with a focus on performance and security in AI computing architectures.
Microchip notes that embedded controllers manage power sequencing, alerts, and system-level energy regulation. For DGX Spark, it says the MEC1723 also manages critical firmware operations, including secure firmware authentication in which firmware code is digitally signed and authenticated by NVIDIA to help maintain platform integrity.
Microchip also describes a root of trust for system boot, using cryptographic verification of the firmware via Elliptic Curve Cryptography (ECC-P384) public key technology. The company notes this is critical because the EC is the first device to power on and authorize secure system boot for the supercomputer.
On power and platform control, Microchip says the firmware supports advanced power management, including battery charging, alerts, and system power-state transitions to optimize energy efficiency. It also lists system-control functions such as key scan and keypad operations for user input.
On interfaces and performance, Microchip says the firmware adds new host-interface support by implementing packet command format processing unique to the NVIDIA DGX interface, beyond traditional byte-level data transfers. It also cites value-added integration that incorporates Embedded Memory Interface (EMI) and Static Random-Access Memory (SRAM) interfaces to improve overall system performance.
Source: Microchip
