Training time in Two notes Capture Studio depends primarily on the hardware used to process the model. Understanding how CPU and GPU performance affects training will help you optimise your system and reduce overall processing time.

CPU vs GPU — What’s the Difference?

Training can be performed on either the CPU (Central Processing Unit) or a GPU (Graphics Processing Unit).

• CPU - Handles tasks sequentially and is available on all systems. Training on CPU is reliable but significantly slower, especially for larger datasets.
• GPU - Designed for parallel processing and capable of handling many calculations simultaneously. When supported, GPU acceleration can reduce training time dramatically.

Why GPU Is Faster

Model training involves large numbers of repeated mathematical operations. GPUs are specifically designed to process these operations in parallel, allowing:

• Faster iteration speeds (higher it/s)
• Reduced overall training time
• Improved efficiency for larger capture datasets

For parametric captures, where datasets are larger, the difference between CPU and GPU can be substantial.

Expected Performance Differences

While performance varies by system, typical behaviour is:

• CPU — Suitable for snapshot captures or smaller datasets; longer training times
• GPU — Recommended for both snapshot and parametric captures; significantly faster processing

In many cases, GPU training can be several times faster than CPU.

How to Optimise Training Performance

While performance varies by system, typical behaviour is:

• Use GPU Acceleration (If Available) - Select GPU in the Training tab of Settings. This is the most effective way to reduce training time.
• Select the Correct GPU Device - If multiple GPUs are available, ensure the correct device is selected. On:
• Apple Silicon systems → MPS (Apple GPU)
• Windows systems → your dedicated graphics card
• Use the Test Function - Run the Test benchmark in the Training tab to measure performance before starting a full training run.
• Results are displayed in iterations per second (it/s)
• Higher values indicate faster training

This allows you to estimate training duration and confirm your system is configured correctly.

• Close Background Applications - Training is resource-intensive. Running other demanding applications can reduce performance.
• Close unnecessary software
• Avoid running other CPU/GPU-heavy processes during training
• Ensure Proper System Configuration
• Keep drivers up to date (especially GPU drivers on Windows)
• Ensure sufficient system memory is available
• Confirm that hardware acceleration is enabled and functioning

When to Use CPU

CPU training may still be appropriate when:

• No compatible GPU is available
• Running smaller snapshot captures if GPU options are unavailable
• GPU resources are unavailable or in use

Having issues?

Our support team is here to assist you! For any enquiries, please head over to our Help Desk and submit a ticket to speak directly to one of our in-house specialists.