
Industrial AI edge gateways typically offer an AI computing power-to-power ratio of 2-10 TOPS/W. High-efficiency NPUs can reach 15 TOPS/W. In high-efficiency designs, low-loss ferrite beads are required for powering the AI core. The PBZ1608E600Z0T is recommended, as it has a DCR of only 0.05Ω and a loss of 0.01W. It also features a 60Ω 100MHz impedance to suppress high-frequency noise. Simultaneously, the power converter MOSFET is equipped with an RC snubber network capacitor of 0.01μF and a resistor of 10Ω to reduce switching ringing, resulting in an AI chip voltage ripple of <1%. In actual measurements, the 15 TOPS/W device showed no efficiency drop under surge voltage of ±2kV.