Windows Kernel Monitor (Native) — Ultimate Robust Monitor

A single-file, dependency-light Windows monitoring tool (Python) focused on real-time CPU + memory + kernel + paging visibility, with an optional GPU block (via Windows Performance Counters).

This is designed to be run in a terminal and refreshed continuously (like a lightweight top for Windows, but with paging + kernel memory focus).

---

Features

CPU

• Overall CPU utilization (rolling)
• Per-core utilization
• Per-core effective clock (GHz) and normalized speed (% of max observed)
• Detects and displays core counts and hybrid/core labeling when available

Memory (Deep)

• Physical Total / Available
• Commit Total / Commit Limit
• Kernel Paged and Non-Paged pool
• Swap/Pagefile total/used/free (via psutil)

Paging / Memory Pressure (PDH)

• Page Faults/sec (PF/s) + EMA smoothing
• Pages/sec + EMA smoothing
• Page Reads/sec + EMA smoothing
• PagingFile(*) % Usage
• Memory\Available Bytes
• Derived HardFaultPressure index (0–100) and qualitative label (OK/WARN/HOT)

Memory Compression

• “Compressed” memory estimate using the MemCompression process RSS (working set)

Processes

• Top processes by CPU%
• Top processes by RSS (working set)
• Handles / processes / threads totals

GPU (Windows PDH — NEW)

• Overall GPU utilization (aggregated from GPU engine counters)
• Per-engine breakdown (top engine categories), typically:
  • 3D, Copy, Compute, VideoDecode, VideoEncode (others may appear depending on driver)
• GPU adapter memory usage:
  • Dedicated usage
  • Shared usage

Example GPU line:

GPU: Util: 2.6% | Engines: 3D:2.6% | Copy:0.2% | ... | Mem: Dedicated 853.36 MB | Shared 661.43 MB

---

Requirements

• Windows 10/11
• Python 3.9+ recommended (3.8+ typically OK)
• No Visual Studio / compilation required

Python dependencies

Install:

python -m pip install --upgrade pip
python -m pip install psutil

(Everything else is standard library: ctypes, time, datetime, etc.)

---

Run

Basic

python .\native_win_monitor_v29.py

Run as Administrator (recommended / some environments)

Some counters or process visibility can be restricted on certain systems. If you see Access is denied or missing counters, run PowerShell or Windows Terminal as Administrator:

python .\native_win_monitor_v29.py

Stop anytime with Ctrl+C.

---

Output Notes / Interpretation

Page Faults/sec vs Pages/sec vs Page Reads/sec

• Page Faults/sec includes soft faults and hard faults (not all are disk I/O).
• Pages/sec is a stronger indicator of paging activity (potentially hard faults).
• Page Reads/sec indicates paging reads from disk and is often the most “painful” when elevated.

HardFaultPressure Index

A simple heuristic derived from Page Reads/sec and Pages/sec:

• OK: low paging pressure
• WARN: noticeable paging activity
• HOT: sustained hard faults / disk reads likely impacting responsiveness

Memory Compression

Windows memory compression isn’t always exposed as a clean global counter. This tool estimates it using the MemCompression process working set.

GPU Block

GPU values come from Windows Performance Counters:

• \GPU Engine(*)\Utilization Percentage
• \GPU Adapter Memory(*)\Dedicated Usage
• \GPU Adapter Memory(*)\Shared Usage

Some drivers expose extra engine categories (e.g., Security, High). That’s normal.

---

Troubleshooting

“n/a” for some PDH counters

Some counters may be missing or named differently on certain Windows builds/locales. The tool already tries multiple counter path variants.

Verify counters exist with typeperf

Examples:

typeperf "\Memory\Page Faults/sec" -sc 3
typeperf "\Memory\Pages/sec" -sc 3
typeperf "\Memory\Page Reads/sec" -sc 3
typeperf "\Paging File(*)\% Usage" -sc 3
typeperf "\Memory\Available Bytes" -sc 3
typeperf "\GPU Engine(*)\Utilization Percentage" -sc 3
typeperf "\GPU Adapter Memory(*)\Shared Usage" -sc 3

Missing GPU counters

Some systems (or drivers) may not expose GPU Engine counters. Task Manager may still show GPU usage even if counters are missing.

Permission issues

Run PowerShell/Terminal as Administrator (recommended / some environments).

---

Repository Layout (suggested)

windows-kernel-monitor/
  native_win_monitor_v29.py
  README.md
  LICENSE

If you prefer the canonical filename to stay native_win_monitor.py, rename:

Rename-Item .\native_win_monitor_v29.py native_win_monitor.py

---

License

MIT License. See LICENSE.

---

Contributing

PRs welcome. If you’re adding new counters:

• Prefer PDH counters when available (portable and scriptable)
• Keep the terminal output stable and readable (avoid overly wide lines)
• Add a typeperf verification line to this README when introducing new counters