This AI-based high-integration SoC processor delivers 320 TFLOPS@FP16 and 640 TOPS@INT8 of computing performance with just 310 W of max power consumption. Such a massive boost in power efficiency is achieved thanks to HUAWEI's own Da Vinci architecture. In addition to the Da Vinci AI cores, Ascend 910 integrates CPUs, DVPP, and Task Scheduler, and self-manages to maximize its high computing power. Meanwhile, HCCS, PCIe 4.0, and RoCE v2 flexibly and efficiently build scale-up and scale-out systems. HCCS is HUAWEI's in-house high-speed interface interconnecting Ascend 910s, on-chip RoCE interconnects nodes directly, and PCIe 4.0 doubles throughput compared to previous generations.