n-control // NVIDIA RTX GPU Fan Curve, Telemetry & Remote Management Suite

A modern, lightweight, and professional hardware telemetry and fan control suite designed specifically for Linux-based NVIDIA GPU workstations, AI compute servers, and main home-lab rigs (tailored for GeForce RTX 50-series such as RTX 5090 and RTX 40-series cards).

Unlike Windows, Linux workstations and AI inference servers traditionally lacked an intuitive, responsive graphical fan controller. n-control bridges this gap: it runs natively on the Linux host as a dedicated systemctl --user daemon (FastAPI + WebSockets + NVML) and serves a sleek, glassmorphic React web interface (Vite + Tailwind CSS). This allows you to monitor and control your GPU directly in a local browser or manage it remotely from your Windows workstation, laptop, or mobile device across the network.

---

📸 Dashboard Overview

---

🎯 Ideal For

• 🧠 AI & LLM Inference Servers: Running local AI workloads like Ollama, vLLM, PyTorch, or ComfyUI? Monitor exact VRAM allocations per process and prevent thermal throttling during heavy continuous generation.
• 🖥️ Linux Workstations & GPU Servers: Full zero-RPM and custom multi-fan control for your daily driver Linux machine (Pop!_OS, Ubuntu, Debian, etc.).
• 🌐 Remote GPU Management from Windows: Keep your power-hungry Linux AI/compute rig in another room or rack while monitoring thermals, switching profiles, and tweaking curves from your Windows desktop via browser or encrypted SSH tunnel.

---

✨ Features

• 📊 Real-Time Hardware Telemetry (1 Hz):

  • Physical GPU core temperature, hotspot, and GDDR7/GDDR6X VRAM junction temperatures.
  • True physical tachometer feedback (RPM & %) with hardware zero-RPM detection (no fake states).
  • Core and memory clock frequencies (MHz), power draw (Watts & % of power envelope).
  • VRAM utilization gauge and core utilization tracking.

• 📈 Interactive Fan Curve Editor with Auto-Apply:

  • Drag-and-drop curve control points directly on the chart canvas.
  • Live Temperature Tracker: A vertical real-time indicator shows the current GPU temperature and its active target fan speed.
  • Auto-Apply: Curve changes immediately take effect upon releasing points.
  • Hysteresis Drop Protection: Smooth cooling transitions prevent fan ramping up and down during transient temperature spikes.
  • Hardware Zero-RPM Support: Automatically activates BIOS-level quiet mode (GPUFanControlState=0) under 50°C.
  • Emergency Thermal Safeguard: Automated 100% emergency override if temperatures reach critical limits (e.g. 88°C).

• 🔍 Integrated nvidia-smi Suite & Process Explorer:

  • VRAM Process Monitor: Detailed breakdown of every process running on the GPU (PID, executable name, compute/graphics classification, exact VRAM footprint in MB and %).
  • Live CLI Terminal: Authentic, monospaced nvidia-smi ASCII box output in a terminal view with a 1-click copy button and auto-refresh.
  • PCIe & Power Limits: Real-time PCIe link generation and bus width (Current vs. Max, e.g. Gen 4 @ 16x) and enforced power envelopes (Min, Default, Current, Max).

• ⚙️ Profile Management & Live Timeline:

  • Built-in presets: Silent / Zero-RPM, Gaming / Balanced, Extreme Cool, and Factory Default.
  • Create and switch custom profiles stored persistently in profiles.json.
  • 60-second rolling telemetry graph for core temperature and fan speeds.

• ❄️ 1-Click Cooldown Turbo Mode:

  • Instantly ramps all fans to 100% turbo power for 60 seconds with an animated live countdown.
  • Safely cancels at any moment with a single click, returning fans smoothly to the active curve.

• 🛠️ System Diagnostics & Coolbits Validation:

  • Instant checks for NVIDIA driver readiness (580.x+), NVML handle status, Coolbits (28) activation, and nvidia-settings execution permissions.

• 🌐 Seamless Remote Access & Dual Language (EN / DE):
  • Native bilingual interface with instant [ DE | EN ] toggle in the header navbar.
  • Optimized for remote browser management from Windows, macOS, or Linux clients.

• 📋 Structured Live Log Stream:
  • Transparent in-memory and file-based execution logging (n-control.log).
  • Records every driver command, exit code, STDOUT, and STDERR with keyword filtering and line-by-line copying.

---

🚀 Quick Setup (All-In-One Automated Installer)

The installer handles the complete setup in one step: Python virtual environment, dependencies, Coolbits 28 configuration, Linux firewall opening, sudoers permissions, and systemd user service registration.

Step 1: Clone Repository

On your Linux GPU machine / server:

git clone <repository-url>
cd n-control

Step 2: Run All-In-One Installer

./install_service.sh

What the installer automatically configures:

1. Python Virtual Environment: Installs dependencies (fastapi, uvicorn, pynvml).
2. NVIDIA Coolbits 28: Writes /etc/X11/xorg.conf.d/20-nvidia-coolbits.conf to unlock manual GPU fan speed control.
3. Linux Firewall (Port 9191/tcp): Automatically permits incoming TCP traffic on port 9191 in UFW or firewalld for LAN/remote management.
4. Driver Sudoers Rule: Sets up /etc/sudoers.d/n-control-nvidia so nvidia-settings can execute uninterrupted in the background.
5. Systemd User Service: Registers and starts n-control.service with loginctl enable-linger (keeps running across logouts and headless reboots).

⚠️ One-Time Reboot Required if Coolbits was just enabled: The NVIDIA proprietary Linux driver requires a single reboot to apply Coolbits permissions:

sudo reboot

After rebooting, verify fan control status:

nvidia-settings -q [gpu:0]/GPUFanControlState

(Returns Attribute 'GPUFanControlState' ...: 0 $\rightarrow$ Hardware fan control active!)

---

🎮 Access & Daily Operations

Once installed, the web interface is immediately available at: 👉 http://localhost:9191

Common Systemd Management Commands:

Task	Command
Check service status	systemctl --user status n-control
View live log stream	journalctl --user -u n-control -f
Restart daemon	systemctl --user restart n-control
Stop daemon	systemctl --user stop n-control
Inspect raw log file	cat n-control/backend/n-control.log

---

💻 Remote Management from Windows

You can manage your Linux GPU / AI server from any client on the local network:

Option A: Direct LAN Access

Enter your Linux machine's local network IP in your Windows browser:

http://192.168.1.XX:9191

Option B: Encrypted SSH Port Forwarding (Recommended)

Forward the daemon port securely to your Windows PC via PowerShell or Command Prompt:

ssh -N -L 9191:localhost:9191 user@pop-os-ip

(Alternatively, double-click the included scripts/start_tunnel_windows.bat file!)

Then open in your Windows browser: http://localhost:9191

🛡️ Linux Firewall Configuration (Port 9191):

The installer automatically opens port 9191, but if you manage your firewall manually:

• UFW (Ubuntu / Pop!_OS / Debian):

  sudo ufw allow 9191/tcp comment "n-control Daemon"

• firewalld (Fedora / RHEL / CentOS):

  sudo firewall-cmd --add-port=9191/tcp --permanent
  sudo firewall-cmd --reload

---

🧩 Project Architecture

n-control/
├── backend/
│   ├── app/
│   │   ├── hardware/
│   │   │   ├── nvidia.py         # NVML controller, hardware sensors, nvidia-smi data
│   │   │   └── fan_engine.py     # Fan curve math, hysteresis smoothing & safe fallback
│   │   ├── models/               # Pydantic schemas (Telemetry, Profiles, SMI models)
│   │   │   ├── telemetry.py
│   │   │   ├── profile.py
│   │   │   └── smi.py
│   │   ├── logger.py             # Structured file logging & live in-memory ring buffer
│   │   └── main.py               # FastAPI REST/WebSocket server & static asset host
│   ├── install_service.sh        # Systemd user-service installer & sudoers setup
│   ├── uninstall_service.sh      # Clean uninstaller script
│   ├── profiles.json             # Persistent curve & fan configuration
│   └── requirements.txt          # Python dependencies
├── docs/
│   └── images/                   # Screenshots for documentation & README
├── frontend/
│   ├── src/
│   │   ├── components/           # UI components (CurveEditor, NvidiaSmiModal, Gauges, ...)
│   │   ├── i18n/                 # Localization context & translations (DE / EN)
│   │   ├── hooks/                # WebSocket & telemetry synchronization
│   │   └── App.tsx               # Root application component
│   └── dist/                     # Pre-compiled production bundles (committed to Git)
├── scripts/
│   ├── set_fans.sh               # Execution wrapper for nvidia-settings commands
│   ├── setup_coolbits.sh         # Automated Coolbits 28 X11 configurator
│   ├── start_tunnel_windows.bat  # 1-click Windows SSH tunnel launcher
│   └── test_fans.py              # Standalone CLI hardware diagnostic tool
├── GEMINI.md                     # Engineering rules & hardware constraints
├── IDEAS.md                      # Roadmap & planned feature ideas
└── README.md                     # Project documentation

---

❓ Troubleshooting & FAQ

1. Fans do not spin up when clicking Cooldown or setting manual speed?

• Open the Logs modal in the top header navbar.
• Check that driver commands return exit code: 0.
• If you see Operation not permitted or permission issues, re-run ./install_service.sh on the Linux host to ensure the /etc/sudoers.d/n-control-nvidia rule is in place.

2. Fans remain at 0 RPM during low load?

• This is the intentional Zero-RPM feature active in the Silent / Zero-RPM profile. The fans spin down silently below 50°C by delegating control to the GPU BIOS (GPUFanControlState=0).
• As temperature rises or when adjusting the curve upwards in the Curve Editor, fans will engage automatically at their calibrated starting speed (30%+).

3. How do I inspect what is consuming VRAM on my AI server?

• Click the [ nvidia-smi ] button in the header.
• The Processes (VRAM) tab will display each active process (e.g. ollama_llama_server, python, Xorg), its PID, whether it is a CUDA compute or graphics task, and the exact megabytes of VRAM consumed.

---

📄 License

This project is open-source and licensed under the MIT License.