SiliconScope

Sudoless Apple Silicon monitor: SwiftUI dashboard plus menu-bar suite tracking ANE, Media Engine and memory bandwidth Activity Monitor won't show — with DVR-style record & replay.

771 48 Swift MITupdated today

local

View on GitHubDispute this mapping →

Curator's take

The monitor to run when local AI is your workload: it surfaces what actually saturates an Apple Silicon box — GPU vs Neural Engine vs memory bandwidth — and shows per-process ANE memory, a signal no other monitor exposes. Sudoless, no kexts, and Record/Replay lets you scrub a session like a DVR to compare before/after tuning. Credible iStat Menus stand-in too. NOT for non-Apple hardware (macOS 14+, Apple Silicon only), and accelerator metrics beyond CPU are system-wide — macOS won't attribute GPU/Media per process, and it honestly labels that instead of faking numbers. A monitor, not a profiler: Instruments still owns trace-level work.

Mapped by ShipWithAI editors · links verified

README.md

SiliconScope

English · Deutsch · 简体中文 · 繁體中文 · 日本語 · 한국어

A sudoless Apple Silicon system monitor — a native SwiftUI dashboard and a full menu-bar suite — with first-class ANE (Neural Engine), Media Engine, and memory-bandwidth tracking that Activity Monitor and terminal monitors don't surface.

Born from wanting to see how on-device AI and media workloads drive the Apple Silicon accelerators — and grown into a daily-driver monitor that can stand in for iStat Menus.

Featured on AAPL Ch. (JP) and ifun.de (DE).

SiliconScope dashboard with the Replay scrubber

The whole machine at a glance — an AI-workload bottleneck classifier, E/P-core overlaid trends, GPU / GPU-memory / ANE / Media, memory measured against the M1 Max's 400 GB/s ceiling, per-core temperatures, power, and live processes. The bar along the bottom is Replay (new in 3.0): every metric is recorded, so you can scrub back through a session like a DVR.

Menu bar — every metric, iStat-style

Pin any card to its own menu-bar item — CPU · GPU · Memory · Network · SSD · Sensors · Battery — each with a live glyph and a rich dropdown. All sudoless.

The per-metric menu-bar suite

GPU / Media / Neural dropdown Per-core temperatures Combined SS cockpit — workload, all engines, trends, top processes

The richest dropdowns. GPU / Media / Neural — GPU, GPU memory, ANE and Media as live meters plus a four-line 60-second trend. Sensors — per-unit temperatures from real E-Core / P-Core / GPU / Memory sensors (curated SMC keys per chip generation, M1–M5; HID fallback elsewhere). SS cockpit — the whole machine in one dropdown: workload verdict, every engine, 60-second trends, and the top processes.

Measuring a local model's speed and efficiency

On-demand benchmark: "Measure tok/s" runs one short generation and reports the model's decode speed and energy efficiency — tokens/sec · tokens/Wh — stored per model.

📊 Measured tok/s on your Mac? Post it in Discussions — a crowd-sourced per-chip table helps others pick the right hardware.

New in 3.0

🧠 Process Inspector — per-process metrics, sudoless

Click any process to open the Inspector. It shows what Activity Monitor can't: CPU (P/E split) · IPC · per-process power (W) · memory · disk — each with a live sparkline — and the one signal nobody else surfaces per process: Neural-Engine memory. See exactly which app is on the ANE, and how much it's holding.

Process Inspector — per-process CPU, IPC, power, and Neural-Engine memory

*An on-device transcription app running live (rig

Continue your stack

What teams reach for next — and why each earns a place beside SiliconScope. Ranked by curator confidence.

Pairs well with

MLX-LoRA-Studio

Native Mac app for on-device LLM fine-tuning via mlx-lm-lora: pick a model, choose SFT/LoRA/DPO-family algorithms, watch loss fall live, push to Hugging Face. No cloud, no code.

Why it fitsFine-tuning on Apple Silicon is a unified-memory balancing act — SiliconScope shows the GPU/ANE load, memory pressure and bandwidth of a Studio training run live, so you can size batch and quantization to the machine instead of guessing.

Ollama

Run Llama, Mistral and other open models locally with a single command and a clean API.

Why it fitsThe question SiliconScope was built to answer — what is my on-device model actually doing to the silicon — comes up the moment Ollama is serving: watch GPU, ANE and bandwidth load per model and quantization instead of guessing.

ai-token-monitor

Menu-bar/tray app for macOS and Windows that reads Claude Code, Codex and OpenCode session logs and shows live token spend — per-model pricing with cache reads, plan-limit bars, webhook alerts.

Why it fitsTwo menu-bar monitors covering the two halves of local AI work: ai-token-monitor watches what agents spend in tokens, SiliconScope watches what the hardware pays in compute, power and bandwidth.