Xiaomi: New CPU matches Apple cores single threaded, much faster multithreaded

Original Article ↗ Hacker News Discussion ↗

Xiaomi’s new Xring O3 system-on-chip, built on TSMC’s 3nm N3P process and using ARM’s latest C1-Ultra cores, is claimed to match or beat recent Apple silicon in single‑threaded and multithreaded benchmarks, raising questions about how it will perform in real phones under power and thermal limits. Commenters highlight that much of the achievement rests on standard ARM IP and Taiwanese fabrication rather than fully custom Chinese CPU design, but still see it as a major competitive threat to Qualcomm and MediaTek. The announcement also reignites broader debates over China’s state-backed industrial strategy, supply-chain dependence, security concerns, and whether Western export controls are accelerating rather than slowing Chinese advances in chips and EVs.

Chip design & benchmark results

  • SoC uses ARM’s off‑the‑shelf C1‑Ultra / C1‑Premium / C1‑Pro cores on TSMC’s 3 nm N3P, with a 6+4 core layout and large caches (e.g., 16 MB SLC, ~44 MB total reported).
  • Xiaomi’s role is integrating ARM IP blocks: interconnect, cache hierarchy, NPU, LPDDR6 controller, packaging, etc., rather than a fully custom CPU microarchitecture.
  • Vendor benchmarks (Geekbench, AnTuTu, SPEC 2026, 3DMark) suggest:
    • Single‑core roughly in the Apple M5 / A19 class but slightly lower.
    • Multi‑core matching or exceeding some Apple chips helped by more cores.
    • GPU performance reportedly near M5 at around 12–20 W on a dev board.
  • Comparisons to MediaTek Dimensity 9500: both use C1‑Ultra on N3P; prior lab scores around 4000 Geekbench ST dropped significantly in shipping phones due to thermals, so similar skepticism applies here.

Power, thermals, and real‑world behavior

  • Many comments say raw performance is meaningless without power and sustained‑load data.
  • Concerns:
    • Lab boards can be in cold rooms with oversized cooling; phones are thermally constrained.
    • Mobile SoCs often deliver “headline” performance for seconds, then throttle.
  • Some reports claim O3 uses ~25% less power than Xiaomi’s earlier O1 and that O1 tablet SoCs didn’t throttle, but independent measurements and battery‑life data are still missing.
  • For typical “burst” workloads (UI, browsing), even short high‑perf bursts may be enough; heavy gaming and compute are the real test.

Manufacturing, supply chain & sanctions

  • This chip is fabbed by TSMC in Taiwan; Xiaomi isn’t under Huawei‑style US sanctions, so it can use cutting‑edge N3P.
  • Separate discussion: Chinese domestic 5 nm via multi‑patterning DUV has poor yields (~20–40%) and isn’t yet economically competitive with EUV.
  • US pressure to tighten ASML exports (including DUV service) is noted; unclear how far such restrictions will go.
  • Several comments stress that this particular design is not “made in China from scratch” but built on UK‑origin ARM IP and Taiwan fabs, though Xiaomi is still a Chinese OEM and potential policy lever.

Competitive and ecosystem implications

  • Many see this as bad news for Qualcomm and MediaTek, good for OEM competition, and a sign that ARM’s stock cores are finally close to Apple in mobile.
  • Some argue Apple’s real moat is ecosystem (macOS/iOS, tooling, UX, efficiency) rather than sheer speed; others say Apple’s perf‑per‑watt hardware is still unmatched in laptops.
  • Questions about whether Xiaomi will scale this SoC into laptops/desktops; current Xiaomi laptops use Intel.

Security & national‑security debates

  • Large subthread on whether Chinese hardware is an unacceptable security risk:
    • One side: you can’t reliably audit adversary‑supplied hardware; China coerces firms, engages in IP theft, subsidies, political influence, censorship abroad, and should be treated as an adversary supplier.
    • Other side: US and allies also spy, backdoor, and coerce (e.g., cloud laws, historical Crypto AG, management engines); for many individuals, domestic corporations/governments feel like the bigger threat.
  • Some argue it’s rational for a country to avoid foreign network gear (Cisco vs Huawei, etc.) but note that this mostly reallocates “who gets to spy” rather than eliminating spying.
  • Hardware backdoors in Chinese devices are debated: a few recent CVEs in minor vendors are cited; others say clear proof in major CPUs is lacking and that layered defenses are still essential.

China’s industrial strategy & global manufacturing

  • Multiple comments link this SoC to a broader, long‑running Chinese industrial policy:
    • Massive state investment in semiconductors, EVs, green tech, batteries, robotics.
    • Use of subsidies and “dumping” to undercut foreign firms and then dominate sectors (examples: solar, EVs, RAM, TVs).
  • Some view this as smart long‑term strategy correcting market failures; others call it unfair, de‑industrializing the West and creating dangerous single‑country dependence.
  • EU “China Shock 2.0” concerns and efforts to reshore or diversify manufacturing are mentioned; opinions split on whether this is too late or now accelerating.

Android, software, and UX notes

  • Several point out that impressive silicon does not guarantee great UX:
    • Android’s ART (AOT/JIT) and JNI layers introduce overhead, though native NDK code is available.
    • iOS makes more pervasive use of ahead‑of‑time native code and can exploit Apple’s microarchitecture more directly.
  • Xiaomi phones get praise for price‑performance and some system features, but ads, inconsistent extension support, and friction in unlocking bootloaders are recurring complaints.
  • Running full Linux on phones remains hard due to closed drivers and missing kernel sources, despite some bootloader unlocking support.

Skepticism and enthusiasm

  • Enthusiastic commenters see this as:
    • Proof China (and ARM licensees generally) can now meet or beat Apple‑class cores.
    • A healthy shock to incumbents, potentially lowering prices and spurring innovation.
  • Skeptical voices emphasize:
    • These are vendor‑supplied dev‑board numbers; phone‑level, long‑duration, and perf‑per‑watt results are still unknown.
    • Xiaomi’s earlier marketing (e.g., EV pre‑orders) and general Chinese corporate hype culture warrant caution until independent reviews confirm claims.