Microchip releases VectorBlox 3.0 SDK for low‑power FPGA AI inference

Microchip Technology has launched the VectorBlox™ 3.0 Accelerator Software Development Kit (SDK), offered free to developers. The SDK and associated CoreVectorBlox IP streamline the optimization, compilation and deployment of convolutional neural network (CNN) models on PolarFire® FPGA and SoC platforms, enabling multiple vision or sensor‑based AI functions on a single low‑power device. By leveraging sparsity‑based model compression acquired from Neuronix, the solution skips zero‑valued operations, boosting inference speed while cutting power consumption.

Shakeel Peera, corporate vice‑president and GM of Microchip’s FPGA business unit, said, “As AI models continue to grow in complexity, compression is becoming essential for deploying intelligence at the edge.” Vito Fortunato of Planetek Italia highlighted the SDK’s role in real‑time Earth‑Observation processing on the AI‑eXpress‑1 satellite, noting its radiation resilience and reliability for continuous low‑Earth‑orbit operations. The SDK integrates with Microchip’s Libero® SoC Design Suite and supports mission‑critical aerospace, defense and industrial applications that demand high performance, efficiency and secure boot.

In a related release, Microchip also introduced a high‑speed Edge Vision reference design that pairs a PolarFire MPF100T FPGA with CoaXPress connectivity, providing a ready‑to‑use platform for industrial, medical, robotics and automotive vision systems.