AI System Requirements

GPU

A dedicated NVIDIA graphics card (GPU) is strongly recommended for training and applying deep learning models. Model training and applying performance on a GPU can be up to 20x faster than on a CPU, and can shorten training time from weeks to hours.

Since applying often takes seconds, compared to hours for training, a CPU can more reasonably be used if necessary. Please refer to the general System Requirements for more information on CPU requirements and recommendations.

*GPU training requires an NVIDIA GPU with compute capability 5.0 to 9.x (Maxwell, Pascal, Volta, Turing, Ampere, or Ada Lovelace architecture), and your driver must support CUDA Toolkit 11 or higher. We recommend you update your NVIDIA driver to the latest version here. Learn more about CUDA here.

Recommendations

Below is a table of recommended GPU options.

Specs

• Architectures: (later is better): Ada Lovelace > Ampere > Turing > Volta > Pascal
• RAM: The more the better. More important than core count. HBM2 > GDDR6X > GDDR6 > GDDR5X > GDDR5 in terms of memory bandwidth
• Cores: For a given architecture, more often leads to faster training and application times
  

Use Cases

• Deep Learning: Training and running deep learning models (requires 8+ GB RAM)
• Spotlight: Using Spotlight and Snap tools (requires 12+ GB RAM)

*Each use case shows support for 1K, 2K, 4K, and 8K resolutions

Examples: 1K = 1024×768, 2K = 2048×1536, 4K = 4096×3072, 8K = 8192×6144

✅✅✅✅ = recommended for 1K/2K/4K/8K
✅✅✅⚠️ = 1K/2K/4K only
✅✅⚠️⚠️ = 1K/2K only
✅⚠️⚠️⚠️ = 1K only

The recommendations below are general guidelines. Actual GPU performance and memory requirements will vary based on your specific settings. For Deep Learning, factors like Tiles, Size Factor, and Model Parameters significantly impact memory usage. For Spotlight, Tiles, Prompt Mode, and Prompt Settings affect GPU demands. Your results may differ from these recommendations depending on your workflow configuration.