About this collection
Apple silicon makes hardware and operating-system behavior unusually visible to ordinary application work. Unified memory joins CPU, GPU and accelerator pressure; fanless machines turn sustained performance into a thermal question; and macOS policy can matter as much as a benchmark loop. These investigations separate those layers instead of treating every slowdown as a single Apple-specific mystery.
The collection moves from core architecture and matrix engines to GPU wired memory, cooling, power management and reversible interface tuning. It favors repeatable observations, explicit machine context and rollback over universal tweak lists. Some results are tied to one SoC generation, macOS release or memory capacity, so the dates and test conditions matter. Start with the curated pieces for the architectural through-line, then use the complete list to follow narrower experiments around M-series laptops, Tahoe behavior and the boundary between a workload, a kernel policy and a damaged process state.
Start here
- Inside Apple M4’s SME engine A technical review of Apple M4’s SME implementation, its shared cluster-level throughput, ZA pipeline behavior, precision limits, and software implications.
- From A14 to M5: Apple’s wide cores and the x86 single-thread race An expanded Apple A14-M5 timeline against AMD Zen and Intel Core, covering CPU width, NEON, SME, clocks, process nodes, benchmarks, and efficiency.
- How a game crash left 8 GB wired in the Apple GPU stack How a Cronos crash wedged the Apple GPU stack, survived WindowServer recovery, and left 8 GB of unified memory beyond userspace repair.
- Apple Silicon GPU memory limits and wired RAM How iogpu.wired_limit_mb affects wired memory, local LLM workloads, and safer GPU limit testing on Apple Silicon.
- Making macOS Tahoe faster, without snake oil A practical macOS Tahoe performance guide for Apple silicon: power settings, useful checks, responsive UI tweaks, and myths to ignore.
All investigations
- Inside Apple M4’s SME engine
- From A14 to M5: Apple’s wide cores and the x86 single-thread race
- How a game crash left 8 GB wired in the Apple GPU stack
- Using caffeinate on an unplugged Mac
- Making macOS Tahoe faster, without snake oil
- Inside the Apple Neural Engine
- macOS animation tweaks: defaults, motion, and rollback
- Cooling a fanless MacBook Air: pads, shims, stands, and airflow
- Apple Silicon GPU memory limits and wired RAM
Related main-site work
- Apple Silicon GPU Wired-Memory Limit Benchmark Ruby tooling for repeatable Metal workloads, memory-pressure sampling and limit reports.
- Dot Files The current macOS-first terminal and developer-environment project referenced by the tuning notes.
- Main-site performance writing Performance articles and historical benchmarks on the main site.