HiSilicon Products

HiSilicon offers four main product categories: 1) Mobile Processors (Kirin series) - high-performance SoCs for smartphones and tablets with integrated 5G, AI, and multimedia capabilities; 2) AI Accelerators (Ascend series) - dedicated AI chips for edge and data center applications delivering industry-leading TOPS performance; 3) Server Processors (Kunpeng series) - ARM-based server CPUs optimized for cloud computing and enterprise workloads; 4) Connectivity Solutions (Balong series) - 5G modems and connectivity chips for cellular and IoT applications. Each category is designed to deliver best-in-class performance for its target applications.

Browse our product categories to find the right HiSilicon solution for your application. Contact our FAE team for product selection guidance.

Selecting the right HiSilicon processor depends on your application requirements: For mobile devices, consider Kirin processors based on performance tier (flagship, mid-range, entry-level) and 5G requirements. For AI applications, evaluate Ascend chips based on required TOPS performance, power budget, and deployment location (edge vs. data center). For server applications, choose Kunpeng CPUs based on core count, memory bandwidth, and workload characteristics. For connectivity, select Balong modems based on cellular standard (5G NR, LTE) and regional band support. BeiLuo's FAE team can provide detailed product comparisons and recommendations based on your specific requirements.

Review our product selection guides or contact our FAE team for personalized recommendations based on your application requirements.

Kirin, Ascend, and Kunpeng serve different application domains: Kirin is HiSilicon's mobile processor family designed for smartphones and tablets, featuring integrated CPU, GPU, NPU, ISP, and 5G modem with emphasis on power efficiency and multimedia performance. Ascend is the AI accelerator family optimized for artificial intelligence workloads, offering high TOPS performance for inference and training with specialized AI cores. Kunpeng is the server processor family based on ARM architecture, designed for data center and cloud computing with high core counts, large memory capacity, and enterprise-grade reliability. Each family is optimized for its specific use case and cannot be directly substituted.

Identify your primary application (mobile/AI/server) to determine which processor family is appropriate. Contact us for cross-platform migration guidance if needed.

Yes, HiSilicon provides comprehensive software support including: 1) Operating system support - Linux, Android, and HarmonyOS; 2) Development toolkits - HiAI SDK for AI development, Kunpeng DevKit for server optimization; 3) Compiler and optimization tools - GCC, LLVM with architecture-specific optimizations; 4) Debugging and profiling tools - performance analyzers, debuggers, and emulators; 5) Reference implementations - sample code, demo applications, and best practices. BeiLuo provides access to these tools along with technical documentation and training resources to accelerate your development.

Request access to HiSilicon development tools through BeiLuo. Our technical team can assist with tool setup and optimization.

HiSilicon processors are designed with power efficiency as a key priority: Kirin mobile processors feature advanced power management with big. LITTLE architecture, achieving 5-10W typical power consumption under load with sub-1W idle states. Ascend AI accelerators offer industry-leading performance-per-watt, with edge devices consuming 5-20W and data center cards ranging 50-300W depending on TOPS requirements. Kunpeng server CPUs deliver competitive performance-per-watt compared to x86 alternatives, with power envelopes from 100W to 200W depending on core count and frequency. All families support advanced power management features including dynamic voltage/frequency scaling and sleep states.

Consider your power budget and thermal constraints when selecting HiSilicon products. Our FAE team can provide detailed power analysis for your specific use case.