Basic Software and Hardware Platform,Building the cornerstone of the intelligent world with ease of use and ultimate performance,Industry application:Smart city, manufacturing, energy, transportation, finance,Powerful AI computing with better energy efficiency,The Compute Architecture for Neural Networks (CANN) is a heterogeneous compute architecture designed for deep learning. It adopts core components such as AscendCL, DVPP, and HCCL that fully release the computing power of Ascend processors, allowing users to build AI applications and services on the Ascend platform. Ascend Computing Language (AscendCL) is a unified programming interface that decouples software from hardware; Huawei Collective Communication Library (HCCL) provides efficient data transmission between Ascend AI processors in distributed training; and the digital vision pre-processing (DVPP) engine implements hardware acceleration to improve parallelism for image preprocessing.
AI Perf | 8 TOPS INT8 |
Memory | LPDDR4X,4GB,support ECC |
CPU | 4 core * 1.0 GHz |
Encoding/Decoding | H.264 / H.265 hardware decoding,20-channel 1080p 30 FPS, 2-channel 4K 75FPS JPEG decoding: 1080p 768 FPS; encoding: 1080P 256FPS; Max. resolution: 16384 x 16384 |
Expansion Interface | 2x MIPI-CSI 51-pin connector, it can connect the camera module 1x MIPI-DSI 51-pin connector 2x RJ45 network ports, supporting adaptive 100/1000M 2x HDMI ports 2x USB 3.0 Type-A ports, compatible with USB 2.0 1x USB 3.0 Type-C port 1x 40-pin expansion connectors |
Memory Interface | 1x Micro SD Card Port 1x M.2 Key M port,support 1x NVMe SSD |
Power | 24W |
Dimentions | 135mm*120mm*44mm |