Skip to main content

Huawei Ascend 310

Product Overview​

Huawei Ascend 310 is the first AI chip of Huawei's Ascend series, announced in October 2018 at the 5th World Internet Conference; together with the Ascend 910 it launched Huawei's "Da Vinci architecture" full-scenario compute system.

The Ascend 310 targets edge computing and on-device inference; its standout feature is 8W extreme low power, pioneering Huawei's unified approach of "one architecture spanning milliwatt to hundreds of watts". It is the core compute source for smart cameras, edge servers, and on-device AI inference, and the technical starting point of Huawei's smart driving chips (Ascend 610 series).

Core Specifications​

ParameterValue
Release2018-10 (World Internet Conference)
ArchitectureDa Vinci (1st generation)
Process12nm (TSMC)
AI Cores2 Da Vinci AI Cores (some sources say 8 cores)
CPUIntegrated control cores incl. ARM Cortex-A55
FP168 TFLOPS
INT816 TOPS
MemoryLPDDR4x
TDP8 W (extreme low power)
Video Decoding16-channel full HD H.264/H.265
Typical ScenariosSmart cameras, edge servers, on-device AI

310 Family Evolution Comparison​

ModelReleasePositioningINT8FP16Power
Ascend 310 (1st gen)2018On-device/edge inference16 TOPS8 TFLOPS8W
Ascend 310P2022Upgraded edge inference140 TOPS70 TFLOPS72W
Ascend 310B2021/2025Low-cost/low-power derivative~16-200 TOPS—≤5W+
Ascend 5102024Mainstream edge inference256 TOPS—25-70W

Key Features​

  • 8W ultra-low power: energy-efficiency benchmark of the Da Vinci architecture, fits battery-powered and fanless devices
  • Origin of the full-scenario architecture: the same Da Vinci architecture flexibly scaled from device to cloud
  • Video codec: integrated 16-channel full HD hardware decoding
  • Large-scale deployment: powers dozens of edge hardware products such as Atlas 200 / 500 / 300

Vendor Information​

ParameterValue
CompanyHuawei Technologies Co., Ltd. (HiSilicon)
Websitehttps://www.hiascend.com
Release2018-10
ArchitectureDa Vinci (1st generation)

Use Cases​

  • ✅ Smart cameras / object detection
  • ✅ Edge servers / on-device AI inference
  • ✅ Industrial inspection, security surveillance
  • ✅ Robots, drones, IoT terminals