NVIDIA CUDA Use Case

This guide will walk you through setting up and running CUDA applications using GPU Container

CUDA is a parallel computing platform and programming model developed by NVIDIA for general computing on graphical processing units (GPUs). With CUDA, developers can dramatically speed up computing applications by harnessing the power of GPUs. (NVIDIA). With the power of CUDA, multiple business domains are benefited:

1. AI/ML

2. Scientific Computing

3. Medical imaging

4. Video and image processing

5. Finance

6. NLP

When to use the CUDA template:

Choose the CUDA template when your workload requires GPU‑accelerated computation with deep learning frameworks, scientific simulations, high‑performance data analytics, or intensive media processing. It is ideal for developers and researchers who need an environment ready for training models, running inference, or performing parallel data processing.

Step 1: Create a CUDA container on GPU Container.

Select your desired GPU instance based on your workload demands.

Step 2: Enable SSH Access

Our CUDA template supports access container via SSH, we have already setup everything you need to connect. Please follow these steps to setup SSH connection:

1. In the Access Container section, check the box “SSH Terminal Access”.

2. Click the “+” button at the bottom of the SSH keys list.

1. A pop-up will appear asking for:

   1. SSH Key name: Choose a name to identify your key (e.g., my-ssh-key).

   2. SSH Public key: Paste your SSH public key here.

2. Click Save. Your key should now appear in the list, you can edit or delete if needed.

If you don’t have any SSH keys yet, follow our SSH key creation guide to generate a new key pair.

Our CUDA template already includes:

• CUDA Toolkit 12.9.1 & cuDNN, NVIDIA GPU drivers, CUDA compiler (nvcc), core libraries, and cuDNN for deep learning acceleration.

• Ubuntu 22.04 Development Environment, GNU build tools (gcc, make), debugging tools, and essential development utilities.

• AI/ML-Ready Base: Compatible with Python and major machine learning frameworks (TensorFlow, PyTorch, etc.) for AI/ML, scientific computing, and media processing workloads.

Example Workloads You Can Run

• AI Model Training: Train large‑scale neural networks with TensorFlow or PyTorch using GPU acceleration.

• Scientific Simulations: run computational fluid dynamics (CFD), Monte Carlo simulations, or molecular modeling.

• Medical Imaging Analysis: process MRI, CT, and X‑ray images in real time.

• Real‑time Video Analytics: detect and track objects in live video streams.

• Quantitative Finance Models: perform large‑scale portfolio simulations or risk analysis faster than on CPU.

• NLP Applications: fine‑tune large language models (LLMs), perform sentiment analysis, or deploy chatbots.