Peak Performance
C100: 1.2TOPS @ INT16
T100: 0.5TOPs~2TOPs @ INT8
E100: 10TOPs~32TOPs @ INT8
DNN Architecture
ODR, MC, SHR
Framework
Caffe, ONNX, PyTorch,
TensorFlow, Keras, MXNet
Processor
ARM, x86, RISC-V
Cutting-edge AI Chip Stacking Technology
Performance Boosts to 10x or above
ATOM Series
C100/T100
-
Self-developed NPU Engine
-
Self-developed Image Processing Engine
-
Rich Interfaces
High Integration
-
APIs for Mainstream Software Platforms
-
Complete Algorithm Toolchain, Supports Mainstream Algorithm Frameworks
-
Provides Image Processing Tools
-
Provides Cost-effective Hardware Reference Designs
Portability
-
Supports Multiple Perception Sensors
-
Supports Deep Learning
-
Supports AI-ISP
Advanced Technology
-
Built-in Security Engine fully protects Data Security, Algorithm Model Security and Device Security
Safety and Reliability
Visual Intelligence
ASIC/eASIC/FPGA/GPU Design
Re-configurable CNN and FC layers
Parallel Computing
Unified Platform
Customized Hardware-compatible Runtime Libraries
Revolutionary Edge AI Chip Design Technology
Offering a Wide Variety of Cutting-edge Design Solutions
Offering a Wide Variety of Computing Platforms based on Real-life Applications
Powerful Multi-function
Visual Intelligence Technology
Integrate Multiple AI Functions into a Single Processor
Integrated Sensor Fusion Technology
Enhance Advance Driver Assistance System (ADAS) Capabilities for Driving Safety and Efficiency with a Single Processor
ATOM Series
E100
-
Supports 32 TOPs @ INT8
-
Supports 16 TOPs @ FP16
-
Memory Bandwidth of 68.3GB/s
High Computing Power
High Bandwidth
-
Supports up to 32 Channels of 1080p Decoding
-
Supports 12 Channels of Video Encoding
-
Video Pre- and Post-Processing Capabilities boosted up to 4X
Powerful Video Encoding
and Decoding Capabilities
-
Supports Dual Gigabit Ethernet Ports
-
Power Consumption reduced by > 50%
Rich Interfaces
Ultra-Low Power Consumption
-
Built-In Security Engine fully protects Data Security, Algorithm Model Security and Device Security
Safety and Reliability