New Mac mini, featuring M6 and M5 Pro

Original Article ↗ Hacker News Discussion ↗

Apple’s new Mac mini lineup with M6 and M5 Pro chips is drawing mixed reactions, as many welcome the performance gains, 2.5GbE networking, and improved AI/LLM capabilities but balk at sharp price increases compared to the M4 era. Commenters note that RAM and storage costs, the broader “AI boom” and memory shortages are driving up hardware prices, eroding the long-standing trend of ever-cheaper personal computing and making older M1/M4 machines and used or non-Apple alternatives look increasingly attractive.

Pricing and Value

  • Many lament the end of “cheap” Mac minis; base prices have risen sharply (e.g., M4 launch at $599 now replaced by $899+ M6), with European configs (16 GB/256 GB) over €1,000 seen as a psychological barrier.
  • Several users feel they “caught the last train” with discounted M4 minis and M1/M4 Airs; used M1 minis have held value because equivalent new specs now cost roughly double.
  • Some attribute price hikes to RAM and SSD shortages and AI-driven demand; others argue this goes beyond normal inflation.
  • Consensus: value proposition is much weaker than the M4 era, especially for entry-level and 16/256 configs.

Performance, Chips, and Comparisons

  • People want M6 vs M5 Pro comparisons, but Apple mostly markets against older M1/M4, leading to skepticism.
  • Thread references Apple’s own numbers showing modest CPU gains vs M5, and notes that M5 Pro has much higher memory bandwidth and up to 64 GB RAM, likely better for heavier workloads and AI.
  • Debate over how much developers feel CPU gains: some say M1→M4/M5 feels minimal for coding, others with large monoliths or heavy Excel/report workloads see clear improvements.

AI, Local Models, and RAM

  • Strong interest in using minis for local LLMs and “always-on agentic computing,” but many consider Apple silicon memory bandwidth and pricing poor vs NVIDIA GPUs for serious inference.
  • Others counter that high-RAM Max/Ultra machines can run multiple substantial models and that unified memory matters once models exceed GPU VRAM.
  • Several say local inference isn’t yet cost-effective versus cloud APIs, but could become attractive if RAM prices fall and models become more efficient.

Gaming

  • Apple highlighting Cyberpunk 2077 performance is noted as unusual but welcome; some see real progress (Game Porting Toolkit, good integrated GPUs), others say Apple still doesn’t truly commit to gaming.
  • Mixed reports: some play Cyberpunk smoothly on M1/M-series via native port or Wine/CrossOver; others emphasize that a cheap Steam Deck or older PCs can already handle it.

Networking and Storage / Software Issues

  • 2.5 GbE as standard is widely praised and seen as overdue for NAS/media workflows.
  • APFS is criticized as slower than modern Linux filesystems, especially under parallel I/O, though it has underexposed advanced features.
  • Broader frustration that modern software (including on macOS) feels bloated and slow despite powerful hardware.

Macro Concerns and Alternatives

  • Long digression on capitalism, regulation, and AI data centers driving component scarcity and pushing users toward cloud/leased thin clients.
  • Some expect a classic boom–bust cycle with RAM/flash prices crashing later; others are pessimistic.
  • A few suggest Chinese mini PCs + Linux or repairable laptops (ThinkPad, Framework) as alternatives; concern that truly personal, affordable computing is under threat.