MyGPU v1.3.0

Changelog

• Rebranded to MyGPU: A GPU Management Utility (updated CLI, web UI).
• Made Notification System consistent through winrt.
• Final release for v1.x.x

Generated 2025-12-18

Using the MyGPU

Run .\setup.ps1 after extracting the zip, it will automatically set the environment and give further instructions.
For details, see recent commits.

LocalGPUMonitor-v1.2.3

Changelog

• Feature: Per-GPU VRAM caps UI
• Feature: Persisted watchlist and VRAM cap settings across restarts.
• Feature: Admin-only enforcement (prototype enforcer to reserve VRAM).
• Feature: Client-side red toasts for VRAM cap exceed alerts; 10s cooldown to avoid spam.
• Improvement: Stronger termination flow for watched processes (terminate → wait → kill → children, re-check after 5s).
• UI: Replace per-row warning with a single large red-dot badge for watched processes on exceeded GPUs; checkboxes disabled for non-admin users.
• Fix: Windows notifications: prefer winrt when available; safe fallback to win10toast without triggering WNDPROC TypeError.
• Fix: Simulation mode TFLOPS reporting - estimate and accumulate FLOPs per particle step so Avg/Peak TFLOPS are displayed in simulation.
• Decided to brand the project "MyGPU"

For details, see recent commits.

LocalGPUMonitor-v1.2.2

Usage

-Run setup.ps1 after extracting the zip file

Changelog

• Rename: rebranded to "Local GPU Monitor" since I am decreasing the scope
• CUDA: require CUDA 12.x (setup and runtime checks)
• UI: unified hover tooltips; admin and start buttons use data-hover
• Bugfix: Start Simulation/Benchmark button and progress reset on stop

For full details see repository commit history.

v1.2.0 Release

GPU Health Monitor v1.2.0

Changelog

• Improve CLI
• Inline simulation
• Web UI: Start Simulation button disabled for stress-test/custom modes to avoid unsupported states.
• minor robustness fixes

v1.1.0: Merge pull request #3 from DataBoySu/3d-sim

first proper release
added pygame simulator for stress testing while enjoying the visuals
had to remove opengl due to added complexity
future improvement:
give custom sliders
shape change
better working particle physics
limitation:
bugs in the 3d engine
error handling is poor
program may hand, simply ctrl+c the console

v1.0.0

What It Does

Real-time GPU monitoring (utilization, memory, temp, power) + intelligent benchmarking with auto-scaling workloads and performance baselines. Web dashboard + CLI interface.

Use Cases

• Power Users: Wanting to test their new GPUs, workflows etc. AIO monitoring and benchmarking tool.

Known Issues

• Per-process GPU %: Only CUDA workloads (graphics processes show 0%)
• Benchmark tab: Requires PyTorch/CuPy + CUDA Toolkit
• Setup: Windows PowerShell only (manual venv for Linux/Mac)
• Scope: Single-machine monitoring (multi-node in v2.0)

Installation

Extract ZIP → .\setup.ps1 → python health_monitor.py → http://localhost:8090

Requires: Python 3.8+, NVIDIA GPU with drivers

What's Changed

• Modulize the files by @DataBoySu in #1
• First version by @DataBoySu in #2

New Contributors

• @DataBoySu made their first contribution in #1

Full Changelog: https://github.com/DataBoySu/cluster-monitor/commits/v1.0.0