Huawei 256 TFLOPS@FP16 and 512 TOPS@INT8 AI CHIP
This AI processor delivers 256 TFLOPS@FP16 and 512 TOPS@INT8 of compute performance with just 350 W of max power consumption. The massive boost in power efficiency is thanks to Huawei's own Da Vinci architecture.
Ascend 910 is a high-integration SoC processor. In addition to the Da Vinci AI cores, it integrates CPUs, DVPP, and Task Scheduler. It self-manages to make full use of its high computing power.
HCCS, PCIe 4.0, and RoCE v2 build scale-up and scale-out systems both flexibly and efficiently. HCCS is Huawei's in-house high-speed interface interconnecting Ascend 910s. On-chip RoCE interconnects nodes directly. PCIe 4.0 doubles the throughput of the previous generation.