Specifications
AI Processor
1x Huawei Ascend 310 (Da Vinci)
AI Performance
16 TOPS INT8 / 8 TFLOPS FP16
CPU
8-core Arm Cortex-A55 @ 1.6 GHz
Memory
8GB / 4GB LPDDR4X, 128-bit, 3200 Mbps
Form Factor
Half the size of a credit card
Host Interface
PCIe 3.0 x4
I/O
RGMII, USB 2.0/3.0, I2C, SPI, UART
Video Decode
H.264/H.265 up to 16-ch 1080p30 (2x 3840x2160@60)
Video Encode
H.264/H.265 1-ch 1080p30
Image Codec
JPEG decode 1080p 256 FPS, encode 1080p 64 FPS, up to 8192x8192
Power Consumption
~9.5W typical
Operating Temperature
Wide-range, edge/industrial grade
Software Stack
Huawei CANN, MindSpore, AscendCL
Typical Devices
Cameras, drones, robots, video terminals
Certifications
CE, FCC, RoHS
Overview
The Huawei Atlas 200 is the device-side entry point to the Ascend AI accelerator family. Packaged at roughly half the size of a credit card, it pairs the Ascend 310 processor with a general-purpose Arm CPU and hardware codecs to deliver real-time inference and multi-channel video analysis inside space- and power-constrained edge devices.
It provides 16 TOPS INT8 and 8 TFLOPS FP16 with 8GB of LPDDR4X memory, a PCIe 3.0 x4 host link and a rich set of embedded interfaces (RGMII, USB, I2C, SPI, UART). Typical power is around 9.5W, and the module decodes up to 16 channels of 1080p30 video, making it well suited to intelligent cameras, drones, robots and video terminals.
The module runs Huawei's CANN and MindSpore software stack, and is also available on the Atlas 200 DK developer kit for prototyping. QS Compute supplies the Atlas 200 for OEM edge-AI hardware programmes, camera and robot makers and intelligent video device builders.
Key Benefits
16 TOPS INT8 in a half-credit-card module. ~9.5W typical power for embedded edge devices. 16-channel 1080p30 decode for multi-stream video. Rich embedded I/O with PCIe, RGMII, USB, I2C, SPI and UART.
Applications
Intelligent cameras and video terminals, drones and UAVs, mobile robots, industrial edge inspection, and multi-channel video analytics at the device edge.
Request a Quote — HUAWEI ATLAS 200 — ASCEND 310 AI ACCELERATOR MODULE
QS Compute — global B2B supply of AI computing hardware, edge AI systems and accelerators. Volume pricing, 15-day sample lead time.
Get Your Quote →