Fully support L2/L2+ autonomous driving
Verified by
both ISO26262 and AEC-Q100 automotive standard
High performance
8 cores ARM Cortex A55 Processor @1.2GHz
Automotive grade dual channels
32bit LPDDR4
Multi-dimensional heterogeneous architecture DynamAI NN engine
Perform up to 16 TOPS(INT8)，
Computation-energy ratio > 5 TOPS/W
Support access of multiple sensors
such as camera, millimeter wave radar
High performance NeurallQ ISP technology,
supporting 8 channels processing of high-res dynamic images in real time
Embedded high performance
CV accelerator and 4*720p video encoder-decoder engine
• 8 channels HD camera input
• Support 8MP
• 3-exposure HDR, up to 140dB dynamic range
• Offline low-light denoising and LED flash suppression
• Mix of inline and offline processing mode
• 3 cores high performance vision DSP
• CV Hardware acceleration
• H.264 video encoder and decoder for 4*720p video
Typical power consumption 15W
FCBGA 25mm*25mm, 0.8mm pitch
Environment-40°C ~ 105 °C with passive heat removal
• Hybrid precision 4-bit/8-bit MAC Array
• Perform up to 16 TOPS(INT8)
• Energy efficiency>5TOPS/Watt (INT8)
• Overall 80% utilization of convolution layers MAC array
• Sparse support for storage and acceleration
• Open-source DynamAI NN compiler tools
• Verified by ISO 26262 ASIL-B and AEC-Q100 standard
• Security MCU with dual-cores lockstep
• Real time security monitoring and inspection
• Embedded safety guard mechanism involving ECC、Parity
• Secure booting
• OTP on chip for private-key storage and life-cycle management
• 1x PCIe Gen3 4L or 2x PCIe Gen3 2L, and can be managed as Root Complex or Endpoint port
• USB3.0 DRD and USB2.0 device
• 10/100/1000Mbps automotive grade Ethernet
• eMMC/SD
• CAN_FD
• UART、SPI、I2C、I2S、GPIO
