Can Intel finally beat ARM on performance per Watt?

Original Article ↗ Hacker News Discussion ↗

Intel’s new low-power Core 5 laptop chip has prompted claims that it finally matches Apple Silicon on performance per watt, at least in certain FP64-heavy benchmarks like HPL. Commenters welcome signs of improved x86 efficiency but note that Apple’s ARM-based chips still lead in overall performance, idle power use, and tightly integrated hardware–software design, and they question whether a supercomputing-style test reflects real-world laptop workloads. The thread also touches on pricing disparities, YouTube clickbait and reviewer credibility, and the broader competitive landscape between Intel, Apple, AMD, and ARM for efficient consumer and server systems.

Intel vs Apple Silicon efficiency

  • Many see the new Intel Core 5–based XPS 13 as the first x86 laptop that can “punch in the same weight class” as recent Apple Silicon on performance per watt, at least in some scenarios.
  • Others argue Apple remains 1–2 generations ahead in efficiency and raw performance, especially comparing against higher-end M-series chips rather than the A-series “phone” SoC in the MacBook Neo.

Benchmarks and methodology

  • A lot of focus is on an FP64 HPL (supercomputing) benchmark, where the Dell reportedly beats most tested M1–M4 systems except the M4 mini in Gflops/W.
  • Several commenters criticize this as cherry-picked and unrepresentative: FP64 is niche for typical laptop users, and broader benchmarks (integer, GPU, browser tests, common apps) matter more.
  • Others note additional benchmarks (Geekbench, other suites) were run and show the Intel chip “holding its own” in multicore under sustained load, especially thanks to active cooling.

Performance vs real-world workloads

  • MacBook Neo is reported as ~1.4× faster single-core and ~2× faster in graphics, while Intel wins specific efficiency metrics under sustained CPU load.
  • There is debate over how much FP64 is used in practice: some say it’s everywhere (web rendering, general non-integer math), others say most consumer workloads are FP32, integer, or GPU/AI-heavy.

Process, architecture, and tuning

  • The Intel chip uses a new 18A process and a mix of P-cores and low-power E-cores, with clocks tuned for efficiency instead of max frequency.
  • Some argue Intel’s historical problem was lagging process technology and OEMs tuning for peak benchmarks, not efficiency.
  • Others insist Apple still leads in IPC, perf/area, and perf/W on comparable nodes.

Pricing and market positioning

  • In the US, the XPS 13 is priced close to the MacBook Neo, making comparisons natural; in parts of Europe it’s ~50%+ more expensive, shifting its competitive set to MacBook Air instead.
  • Some suspect regional pricing is deliberate marketing to generate favorable US reviews.

Platform & OS factors

  • Several note Apple’s tight integration of hardware, OS, and software optimizations (scheduling, display refresh, background tasks) as a major contributor to efficiency.
  • There is curiosity about how Intel/AMD would compare to Apple if both ran similarly optimized OSes (e.g., Linux vs macOS).

Low-power x86 vs ARM mini-systems

  • Intel’s N100-class chips are praised as highly efficient, fanless, and well-supported for Linux/BSD, making them strong router/homelab choices versus often-fragile ARM SBC ecosystems.

Design choices: fans and ports

  • Some value the Dell’s fan for preventing throttling; others strongly prefer fanless designs like the Neo.
  • Removal of the 3.5mm headphone jack on laptops sparks extensive debate: cost and reliability vs user convenience and classroom use; USB-C DACs and dongles seen as both adequate and annoying.

Meta: reviews, clickbait, and trust

  • A long subthread questions whether framing Intel as having “matched” Apple based heavily on one FP64 metric is misleading.
  • There is disagreement on whether YouTube-style clickbait titles are acceptable tradeoffs for reach, and how much that should affect perceived reviewer credibility.