Gitorian/sysmon: Lightweight Windows system monitor for GPU/CPU/RAM/VRAM · GitHub

███████╗██╗   ██╗███████╗███╗   ███╗ ██████╗ ███╗   ██╗
██╔════╝╚██╗ ██╔╝██╔════╝████╗ ████║██╔═══██╗████╗  ██║
███████╗ ╚████╔╝ ███████╗██╔████╔██║██║   ██║██╔██╗ ██║
╚════██║  ╚██╔╝  ╚════██║██║╚██╔╝██║██║   ██║██║╚██╗██║
███████║   ██║   ███████║██║ ╚═╝ ██║╚██████╔╝██║ ╚████║
╚══════╝   ╚═╝   ╚══════╝╚═╝     ╚═╝ ╚═════╝ ╚═╝  ╚═══╝

Lightweight Windows system monitor designed for minimal latency impact on games and applications.

• Low-latency design: below-normal priority, 1ms timer resolution, zero-copy buffers
• Real-time monitoring: GPU, CPU, RAM, VRAM, GPU temperature
• Auto-detects NVIDIA or AMD GPUs, falls back to CPU/RAM only
• Three display modes: Minimal, Graph, Headless
• Automatic CSV logging with timestamps
• Visual warnings when GPU hits 95%+ utilization

• Windows
• NVIDIA or AMD GPU (optional)
• Rust toolchain (for building from source)

Download binary:

1. Go to Releases
2. Download the version for your GPU:
   • sysmon-nvidia.exe — NVIDIA only
   • sysmon-amd.exe — AMD only
   • sysmon-dual.exe — both (auto-detects)
3. Run it

Build from source:

git clone https://github.com/Gitorian/sysmon.git
cd sysmon
cargo build --release

Run the executable and select a display mode:

Minimal - Compact text display

┌─ Current ──────────────────────────────────────────────────────┐
│ GPU: 87% │ CPU: 45% │ VRAM:6543MB (68%) │ Temp:72°C │ RAM: 54% │
├─ Maximum ──────────────────────────────────────────────────────┤
│ GPU: 95% │ CPU: 78% │ VRAM:7890MB (82%) │ Temp:75°C │ RAM: 67% │
└────────────────────────────────────────────────────────────────┘

Graph - Full-screen ASCII graph with 300 data points of history

Headless - Background logging only, no UI overhead

All modes create timestamped CSV logs in the executable directory.

Sampling intervals:

• Interactive modes: 1s, 2s, 5s
• Headless mode: 1s, 2s, 5s, 10s, 30s

Architecture:

• CPU/RAM: Windows API (GetSystemTimes, GlobalMemoryStatusEx)
• GPU: NVML (NVIDIA) or ADLX (AMD) with automatic detection
• Terminal: crossterm with raw mode
• Logging: 16KB buffered writes with batched flushes
• Modular design: separate modules for metrics, GPU backends, display modes

Low-latency optimizations:

• Below-normal process and thread priority
• 1ms Windows timer resolution (timeBeginPeriod)
• Zero-allocation hot paths (reused buffers)
• GPU handle caching
• Fat LTO and aggressive compiler optimizations
• CPU-native instruction targeting

MIT License - See LICENSE

• nvml-wrapper
• adlx
• crossterm
• chrono
• anyhow
• Windows API