MacBook Neo Deep Dive: Benchmarks, Wafer Economics, and the 8GB Gamble

· hardware · Source ↗

TLDR

  • A18 Pro in the $599 MacBook Neo posts Geekbench 6 single-core of 3,569, between M3 and M4, but throttles 87% after 60 seconds of sustained load in its fanless chassis.

Key Takeaways

  • A18 Pro (TSMC N3E, 2P+4E cores, 16-core Neural Engine) beats Intel Lunar Lake Ultra 5 226V by 38% and Snapdragon X Plus by 43% on single-core at this price tier.
  • Thermal cliff is steep: all-core stress drops CPU utilization from 570% to 207% between T+60s and T+75s; case surface stays at 97.6F while the SoC hits 105C internally.
  • Multi-core at 8,668 is M1-class and trails the Snapdragon X Plus (11,345) and Intel Ultra 5 226V (9,702); the M4 Air is 70% higher.
  • Apple’s $599 price floor depends on A18 Pro wafer cost amortization across ~230M iPhones per year – no other OEM has comparable supply chain leverage.
  • 8GB unified LPDDR5x is soldered with no upgrade path; after macOS overhead, available application memory is roughly 1.5-2GB.

Hacker News Comment Review

  • The 8GB ceiling is the dominant concern: commenters note macOS already consumes over 75% of RAM on a fresh boot, leaving almost no headroom if OS memory footprint grows over the product’s lifetime.
  • Longevity splits opinion: one commenter has run a 16GB M1 Air daily for six years with Docker, VS Code, Electron, and a macOS VM without major issues, but the Neo’s 8GB floor makes that comparison unreliable.
  • The USB 2.0 second port and absent Thunderbolt are flagged as real workflow blockers for anyone needing fast external storage; anecdotal USB 3 reliability issues on recent MacBooks compound the concern.

Notable Comments

  • @beloch: argues that OS memory growth even a few percent could trigger planned obsolescence legislation by eating into the already-thin application memory budget.
  • @darkteflon: bought an 8GB M1 Air in 2020 as a throwaway travel machine, used it as primary for two years, then passed it to his son – still running fine, suggesting 8GB is survivable for light workloads long-term.

Original | Discuss on HN