GPU Benchmarking Project

This project is designed to automatically run HPCDL and Gaming benchmarks on your NVIDIA RTX 4070, collect performance metrics, and generate a CSV report for analysis.

Features

• GPU Verification: Checks whether the GPU in the system is an NVIDIA RTX 4070.
• HPCDL Benchmark: Runs a deep learning style benchmark using PyTorch (a simple network forward pass on dummy data).
• Gaming Benchmark (Simulated): Simulates a graphics-intensive workload and collects GPU utilization data.
• Automated Reporting: Aggregates performance metrics and outputs a CSV report.

Setup

1. Clone the repository:

   git clone https://your-repo-link.git
   cd gpu_benchmark_project

2. Install the Python dependencies:

   pip install -r requirements.txt

3. (Recommended) Install PyTorch with CUDA support: If you're using an NVIDIA GPU (like the RTX 4070), install PyTorch with CUDA to run benchmarks on GPU:

   pip install torch==2.0.1+cu118 torchvision==0.15.2+cu118 torchaudio==2.0.2 --extra-index-url https://download.pytorch.org/whl/cu118

4. Run the benchmarking script:

   python main.py

Configuration

Adjust the benchmark parameters (e.g., number of iterations, descriptions) in the config.yaml file:

gpu_model: "NVIDIA GeForce RTX 4070"
benchmarks:
  hpcdl:
    enabled: true
    iterations: 3
    description: "Training a dummy neural network to benchmark GPU performance."
  gaming:
    enabled: true
    iterations: 3
    description: "Simulated gaming benchmark workload."

Requirements

• Python 3.8+
• PyTorch
• PyYAML
• pandas

Project Structure

gpu_benchmark_project/
├── README.md
├── requirements.txt
├── config.yaml
├── main.py
├── my_benchmarks/
│   ├── __init__.py
│   ├── hpc_benchmark.py
│   ├── gaming_benchmark.py
│   └── stress_benchmark.py       # New stress benchmark module
└── utils/
    ├── __init__.py
    ├── metrics.py
    ├── report.py
    └── plotting.py               # New dynamic plotting module

How It Works

1. Configuration and GPU Check:
   The script loads benchmark configuration from config.yaml and uses nvidia-smi to verify that the installed GPU is the NVIDIA RTX 4070. A warning is issued if the GPU does not match.

2. Running Benchmarks:

   • HPCDL Benchmark: Uses PyTorch to create a simple neural network that performs a forward pass on dummy data. Execution time and throughput are collected.
   • Gaming Benchmark: Simulates a gaming workload by performing heavy matrix multiplications and collects GPU utilization data through nvidia-smi.

3. Reporting:
   Benchmark results from each iteration (including type, execution time, and utilization) are aggregated and exported to a CSV file using Pandas.

4. Extensibility:
   This project structure makes it easy to add additional benchmarks or customize the tests for other GPU models.

Running the Project

1. Install the dependencies:

   pip install -r requirements.txt

2. Execute the main script:

   python main.py

3. After completion, review the benchmark_report.csv file that contains all the aggregated metrics.

Simply copy the above content into your README.md file in your GitHub repository, and you’re all set!