Xiaomi unveils three in-house Xuanjie chips spanning AI smartphones, high-bandwidth computing, and intelligent driving

Deep News
Aug 24

Xiaomi held its Xuanjie Chip Technology Communication Conference on the afternoon of August 24, officially releasing three Xuanjie chips: the AI flagship SoC Xuanjie O3, the 1.22TB/s high-bandwidth AI acceleration chip Xuanjie O100, and the high-compute intelligent driving AI chip Xuanjie D100. This triple launch marks a breakthrough for Xuanjie beyond the boundaries of single mobile chip design, establishing it as the AI computing foundation supporting Xiaomi's full ecosystem of people, cars, and homes.

Among them, the Xuanjie O3 is the second high-end flagship SoC independently developed by Xiaomi, maintaining the 3nm process technology with transistor count rising from 19 billion in the previous generation to 24 billion. Xiaomi lab data shows the Xuanjie O3 achieves an AnTuTu comprehensive score of 5,228,014, making it the first mobile flagship platform in the industry to surpass 5 million points. On the CPU front, the Xuanjie O3 adopts a 10-core all-big-core architecture, featuring 6 super-large cores plus 4 large cores with a maximum clock speed of 4.35GHz. Its GeekBench 6 multi-core score reaches 15,221, a 60% improvement over the previous generation. Power consumption in low-to-medium load scenarios during daily use is further reduced by 25%.

The Xuanjie O3 debuts the next-generation G2-Ultra NX graphics processor and continues with a 16-core ultra-large configuration, delivering an 85% improvement in traditional graphics performance over the O1 and a 182% boost in ray tracing performance. Additionally, the GPU achieves up to 64% lower power consumption at the same performance level compared to the previous generation. The Xiaomi Xuanjie O3 incorporates a total of 60MB of on-chip cache across its main modules and is the first to support LPDDR6 memory, with memory bandwidth reaching 113.8 GB/s, a 48% increase. Xiaomi also applies the Xuanjie unified fusion bus architecture, reducing memory access latency to as low as 82ns.

The Xuanjie O3 features Xiaomi's fifth-generation ISP, supporting up to a 432-megapixel single-camera lens with a maximum internal image signal processing bit width of 24-bit. The security module has been fully restructured, earning national cryptography and CCRC EAL5+ security certifications. In terms of baseband, communication performance aligns with mainstream flagship integrated solutions, with comprehensive 5G scenario power consumption reduced by over 20% compared to the previous generation. On the NPU front, multiple deep optimizations have been made at the hardware level for large language models, with a 4-core architecture delivering 200 TOPS of computing power.

The Xuanjie O3 also adopts the mobile industry's first SIMT computing architecture Vector unit, providing 3.13 TFLOPS of vector computing power. Meanwhile, the Xuanjie O3 NPU increases investment in near-memory computing, paired with 5-value quantization jointly launched with Xiaomi MiMo and hardware-based Huffman lossless compression, enabling model inference speeds 45% faster than mainstream flagship SoCs.

The Xuanjie O100 is a 6nm on-device AI acceleration chip independently designed by Xiaomi, which when paired with the Xuanjie O3 enables ultra-fast inference output for on-device large models. This chip employs 3D Wafer On Wafer vertical stacking, using advanced Hybrid Bonding technology to bond 2 layers of DRAM wafers with 1 layer of NPU compute wafer at high temperature. It delivers 16 times the memory bandwidth of mainstream phones, with on-device large model inference speeds reaching up to 330 Token/s. Built-in 14-core dedicated NPU cores for large models, combined with Xiaomi's XRING HB-Matrix high-bandwidth matrix bus, allow for rapid completion of AI task computations.

The Xuanjie D100 is a 3nm high-compute intelligent driving AI chip independently designed by Xiaomi, also marking the first 3nm intelligent driving chip in China. It contains a 20-core high-performance CPU and a 16-core high-compute NPU, supporting up to 160GB of memory on a single chip and enabling local deployment of large models with over 200 billion parameters. According to reports, the Xuanjie O3 will launch alongside the Xiaomi 18 Fold in September this year, while the Xuanjie O100 and D100 have completed research and development and will enter commercial use next year.

Disclaimer: Investing carries risk. This is not financial advice. The above content should not be regarded as an offer, recommendation, or solicitation on acquiring or disposing of any financial products, any associated discussions, comments, or posts by author or other users should not be considered as such either. It is solely for general information purpose only, which does not consider your own investment objectives, financial situations or needs. TTM assumes no responsibility or warranty for the accuracy and completeness of the information, investors should do their own research and may seek professional advice before investing.

Most Discussed

  1. 1
     
     
     
     
  2. 2
     
     
     
     
  3. 3
     
     
     
     
  4. 4
     
     
     
     
  5. 5
     
     
     
     
  6. 6
     
     
     
     
  7. 7
     
     
     
     
  8. 8
     
     
     
     
  9. 9
     
     
     
     
  10. 10