Cryo

The Alienware command center Dell never shipped for Linux, for Alienware and Dell G-Series machines on the mainline alienware-wmi kernel driver. A root daemon + QML GUI + CLI — no acpi_call, no out-of-tree modules — that does things stock AWCC (and the Windows original) don't:

• Custom fan curves — temp→boost curves per fan group with hysteresis, driven through the kernel's custom platform profile, editable in the GUI or with cryoctl curve.
• Thermal Guard — an emergency failsafe in any profile: trip temperatures force 100% fans, hold, then restore your previous state. Thresholds are editable in the GUI or with cryoctl guard.
• CPU power cap — a per-power-mode ceiling on the CPU's performance state (intel_pstate max_perf_pct), re-applied on every mode switch and after resume. On the m18 R2, 90% ≈ 5.2 GHz single-core: about 12 W and 9 °C off the in-game CPU peak while the GPU gets more headroom.
• Game detection — sustained dGPU load (NVML) flips the machine into G-Mode automatically, and restores your previous profile when you quit.
• Auto profiles — battery → quiet, AC → balanced, all configurable, fired on transitions only so manual picks stick.
• Live telemetry — CPU/GPU temps, fan RPM, dGPU load, power draw and clocks, VRAM usage with session peak and a per-process breakdown, sparkline history — in-window, in the tray tooltip, and streamable to CSV with cryoctl watch --log for long debugging sessions.
• AlienFX lighting — 4-zone keyboard effects (protocol ported from AWCC), including a quantum cyan↔violet preset, restored on boot.

Will it work on my machine?

Developed on an Alienware m18 R2 (Pop!_OS); designed for the wider alienware-wmi ecosystem — the daemon probes what your machine supports at startup (profiles, G-Mode, fan boost, lighting, turbo control, NVML) and every client renders only what exists. Reported so far:

Model	Verdict
Alienware m18 R2 (Intel + RTX 4080)	SUPPORTED — everything works (maintainer)
Alienware Aurora R4 (2011 desktop)	LIMITED — GPU telemetry, game detection, turbo and CPU cap only (#9)

No Dell G-Series report yet. See docs/COMPATIBILITY.md for the feature matrix. cryoctl doctor works before you trust Cryo with your fans; a model report is the most useful thing you can send, working or not:

cryoctl doctor   # hardware/driver report with a supported / partial /
                 # unsupported verdict, made for pasting into an issue

Architecture

┌──────────────┐   unix socket (JSON lines)   ┌─────────────────────┐
│ cryo-gui     │◄────────────────────────────►│ cryod (root,        │
│ (QML + tray) │      /run/cryo.sock          │  systemd service)   │
└──────────────┘                              │  · platform profile │
┌──────────────┐                              │  · hwmon fan boost  │
│ cryoctl (CLI)│◄────────────────────────────►│  · NVML game sense  │
└──────────────┘                              │  · AlienFX ELC USB  │
                                              └─────────────────────┘

Everything thermal goes through sysfs (alienware-wmi kernel driver): the per-handler node under /sys/class/platform-profile/ and the alienware_wmi hwmon (fan*_boost, fan*_input, temp*_input). Lighting talks to the AW-ELC USB controller (187c:0550/0551) directly.

Profile names are translated per machine from what the kernel advertises. On G-Mode-capable models:

Cryo/AWCC name	kernel platform_profile
performance	balanced-performance
gmode	performance
custom	custom (manual fans)

On models without G-Mode, kernel performance is AWCC Performance and Cryo labels it accordingly.

Install

git clone https://github.com/I4cTime/cryo.git && cd cryo
sudo packaging/install.sh   # venv sync + /etc/cryo + systemd unit + icons + .desktop

Requires: the alienware-wmi kernel driver (mainline; fan control needs its hwmon, which landed in Linux 6.16), Python 3.12+, uv on your user's PATH, and — for game detection — the NVIDIA proprietary driver. There is no distro package, Flatpak or PyPI release yet.

To update, pull and re-run the installer; it restarts cryod on the new code:

cd cryo && git pull && sudo packaging/install.sh

Use

cryoctl status | watch
cryoctl profile gmode / cryoctl gmode
cryoctl boost cpu 60
cryoctl cap 90                       # CPU cap for the current mode (cap gmode 95 for another)
cryoctl curve cpu 45:0 60:15 70:35 80:60 95:100
cryoctl guard cpu 90 / cryoctl guard show
cryoctl light quantum / cryoctl light static 00D1FF
cryoctl doctor
cryo-gui

Config lives at /etc/cryo/config.json (deep-merged over defaults in cryo/daemon/config.py) — fan curve points, guard thresholds, CPU caps per mode, auto-rule targets, game detection thresholds, socket group. The file holds only your overrides. The daemon rewrites it only when you change a setting from the GUI or cryoctl curve/guard/cap (every patch is validated first, and keys you never touched keep tracking new defaults); what the daemon changes on its own (last lighting, the curves-off latch a manual boost sets) lives in /var/lib/cryo/state.json.

After suspend, cryod-resume.service runs cryoctl reapply, which reopens the lighting controller, re-asserts the platform profile and re-initializes NVML. The daemon also retries NVML on its own every 30 s while the NVIDIA driver is missing, so starting before the driver loads is fine.

Notes

• Lighting protocol ported from tr1xem/AWCC (GPL-3.0); this repo is GPL-3.0-or-later accordingly.
• Game detection needs the NVIDIA driver's NVML (present with the proprietary driver). Without it, Cryo degrades to manual + AC/battery rules only.
• Fan boost is not implemented by every model's firmware; when the startup probe finds it missing, curves and the Thermal Guard disable themselves visibly rather than failing silently.
• On a machine the alienware-wmi driver only partly covers (no hwmon and/or no platform profiles — typically pre-2012 desktops), cryod runs in limited mode: fans, curves, the guard and power modes are out, but GPU telemetry, game detection, the turbo toggle, the CPU cap and lighting still work where present, and every client hides the rest. Only a machine with nothing at all to drive makes cryod log one plain-language reason and exit 78 without restarting. cryoctl doctor gives the same SUPPORTED / PARTIAL / LIMITED / UNSUPPORTED verdict without the daemon.

Uninstall

cryod leaves the firmware where it last put it, so hand the machine back to stock first, then remove what the installer added:

cryoctl profile balanced && cryoctl boost cpu 0 && cryoctl boost gpu 0 && cryoctl cap 100
sudo systemctl disable --now cryod.service cryod-resume.service
sudo rm /etc/systemd/system/cryod.service /etc/systemd/system/cryod-resume.service
sudo systemctl daemon-reload
sudo rm -r /etc/cryo /var/lib/cryo
rm ~/.local/bin/cryoctl ~/.local/bin/cryo-gui ~/.local/share/applications/cryo.desktop
rm ~/.local/share/icons/hicolor/*/apps/cryo.png ~/.local/share/icons/hicolor/scalable/apps/cryo.svg

Then delete the cloned cryo directory. Skip the boost or cap commands your machine doesn't support (cryoctl doctor lists them).