AI processor

Lightspeeur® Series Intelligent Accelerator

with Matrix Processing Engine (MPE™)

The Lightspeeur® series accelerator delivers a revolutionary architecture that features low-power, low-cost, and massively parallel compute capabilities to empower Artificial Intelligence (AI) to consumer electronic devices, mobile edge computing, as well as cloud AI data centers. The Lightspeeur® artificial intelligence accelerator is the world’s first to run a mixed workload of speech, Natural Language Processing (NLP), and image and video processing with Convolutional Neural Networks (CNN).


Lightspeeur© 2801S uses 100% proprietary and patented technologies to accelerate CNN processing at extremely high speeds, while consuming very little power.

The architecture uses a unique 2-dimensional Matrix Processing Engine (MPE™) with AI Processing in Memory (APiM™). AI processing memory blocks reduce power requirements and improve performance.

  • 9.3TOPS
  • APiM™ architecture (uses Memory as the AI processing Unit)
  • MPE™ architecture up to 180,000+ Processing Elements @66MHz
  • ALU usage efficiency is 77% and higher running CNN networks
  • Support Specialized Floating Point Computation. Image Data (activation) mantissa 5 bit, exponents 4 bit; Filter parameter (weight) mantissa 12 bit, exponents 2 bit
  • Concurrent multi-cell in-situ data computation within memory
  • Support standard AI frameworks and pre-trained neural networks
Target Markets & Applications
  • Mobile Edge Computing
  • Smart surveillance/video cam/toys/robotics/home
  • Augmented Reality (AR) products
  • Virtual Reality (VR) products
  • Face detection and recognition
  • Speech and voice recognition
  • Natural Language Processing
  • Deep Learning enabled devices
  • Cloud ML/DL systems
  • AI Data Center servers
  • ADAS/Autonomous driving