AI-X™

 

AI-X™

Turnkey Solution for Edge-AI Vision SMEs

Optimized AI-Chip Solution

 
AI-X turnkey solution, powered by GTI high-performing accelerator chips, provides chip-level optimization solutions for realizing deep neural network development on edge devices. AI-X provides an efficient SW-HW design framework for developing CNN-based models on chips. It enables industries and SMEs to accelerate their AI product turnaround time, maximize on-chip resource utilization, minimize CPU runtime, boost performance efficiency and reduce power consumption by leveraging GTI Networks (GNet) HW-SW solution. AI-X combines end-to-end HW and SW solutions for quickly deploying your data, tasks, AI custom model, algorithm design, and chip implementation challenges on the AI devices equipped with the GTI high-performing chips.

At-a-Glance:

 
 
  • Reduced CPU Utilization
  • Holistic HW-SW Optimization
  • Unmatched TOPS
  • Ultra-Low Power
  • Model Pruning
  • Quantization-Aware Training
  • Scalable Architecture
 

Applications:

 
 
  • Edge-AI Vision Devices
  • Consumer Electronics
  • AI-Camera Sensors
  • Automotive

Turnkey Solution Offering:

 
  • Vision-based Solution for AI Chip Design Challenges
  • GTI Chip-Level Optimization Solution  
  • Seamless Integration between GNet and GTI Chips
  • Holistic HW-SW Optimization
  • AI Model on-Chip Realization
  • Customized Model Conversion (MDK)
  • Model Pruning
  • Joint Chip-Level and Neural Network Optimization
  • Reduced Latency During Inferencing
  • Multi-Chip Solution Capabilities
  • Quantization-Aware Training
 

Key Benfits:

 
  • Optimized Precision and Processing Efficiency
  • Boosting Performance Efficiency
  • Reduce Power Consumption
  • Minimize AI CPU Utilization / Runtime
  • Over 3x CPU Saving (compared to other solutions)
  • Increased Performance by 5x
  • Maximizes AI On-Chip Utilization (>90%)
  • Model Conversion, Training, and E2E Validation

AI-X: Highly Efficient Software-Hardware Design Workflow

Powered by GTI Accelerator Chips

 
  • 2803: 24 TOPS/W or 556 FPS/W
  • 5801: 12.6 TOPS/W or 468 FPS, 250mW
  • 2801: 9.3 TOPS/W
  • Unmatched TOPS & Power Efficiency
  • Integrated On-Chip Memory
  • Scalable Multi-Chip Architecture
  • Executes a Large Number of Calculations in Tensor Operations
 

Full-Stack Solution for AI Development