newsfilter
← Latest News
GlobeNewswire·

Microchip Advances Neural Network Implementation with VectorBlox™ 3.0 Accelerator SDK

MCHP
  • Microchip Technology released the free VectorBlox 3.0 Accelerator Software Development Kit (SDK) & CoreVectorBlox IP to simplify FPGA-based AI implementation for aerospace & defense sectors.
  • The integrated toolchain streamlines optimization, compilation, & deployment of convolutional neural network (CNN) models on PolarFire FPGA & SoC platforms.
  • VectorBlox 3.0 leverages sparsity-based model compression from the Neuronix acquisition to reduce compute demands while preserving accuracy by skipping zero-valued operations.
  • The solution supports efficient execution of vision-based CNN models on a single low-power FPGA, enabling consolidation of multiple AI workloads & reduced power consumption for always-on edge applications.
  • Planetek Italia utilized the platform to validate real-time Earth Observation processing, including object detection & semantic scene analysis, on the AI-eXpress-1 satellite deployed in 2025.
  • The Spacecraft Pose Network v2 (SPNv2) enables autonomous navigation, rendezvous, docking, & debris removal using vision data on SEU-immune PolarFire FPGAs & SoCs.
  • AIKO validated the synergy between PolarFire SoC & VectorBlox by deploying the clear_CHARLES suite for onboard cloud & ship detection on power-efficient space platforms.
  • VectorBlox SDK v3.0 is supported by Libero SoC Design Suite & available to customers at no charge.
  • A free webinar on achieving two times faster CNN inference with sparsity-aware acceleration is scheduled for July 16, 2026.