According to a tweet by the user lemire, Xiaomi’s new CPU roughly matches Apple cores on single‑threaded tasks while being much faster in multithreaded execution. The tweet notes that Apple may soon announce its next processor, which could diminish this advantage, and that it may be difficult to find a phone equipped with the upcoming Xring O3 chip. Nevertheless, the author argues that the Xiaomi processor is worth discussing because it reveals an important trend in chip design. The source text states that the chip has a total of 44 MB of cache, which is described as more than most laptop CPUs. It adds that if a user has an Intel processor in their laptop, they likely have less cache than this Xiaomi chip. The biggest cores on the Xring O3 are identified as the C1‑Ultra cores. These cores are described as really powerful and support SME2 (Scalable Matrix Extension 2) for matrix/AI acceleration and SVE2 for data parallelism (SIMD). The chip is said to be astonishingly wide, with 21 execution ports, six of which support SIMD operations of 128 bits. The tweet compares this to desktop processors, noting that this is more execution ports than found on typical Intel/AMD chips, while acknowledging that AMD Zen 5 can perform 4×512‑bit operations but that the best achievable on an ARM chip is 6×128‑bit. The overall trend highlighted is a move toward cores that are massively parallel in terms of the number of execution units, providing better SIMD capability and more units able to perform arithmetic operations. This allows many independent additions or multiplications per cycle, coupled with the large cache size. The tweet concludes by stating that this is where the transistor budget is being allocated. Finally, the source provides specific benchmark figures: the Xiaomi Xring O3 scores 3,945 in Geekbench single‑core and an unprecedented 15,221 in multi‑core.
Key facts
- Xiaomi’s new CPU matches Apple cores on single‑threaded tasks
- The CPU is much faster than Apple’s in multithreaded execution
- Total cache size is 44 MB, exceeding most laptop CPUs
- C1‑Ultra cores support SME2 and SVE2
- Chip has 21 execution ports, six supporting 128‑bit SIMD
- Geekbench scores: 3,945 single‑core, 15,221 multi‑core