Skip to main content

Compute Resources

Selecting the right compute tier keeps latency low while controlling costs. This guide explains the available options and strategies for different workloads.

Available Tiers​

Refer to the targon.Resources reference for the full list of identifiers. They fall into:

  • CPU tiers (cpu-small → cpu-xlarge) for lightweight services, background jobs, and orchestration.
  • GPU tiers by accelerator family — H200, H100, B200, B300 (each small → xlarge), plus RTX6000B (small) and RTX4090 (small → large).
  • VM tiers (type=vm in inventory) for confidential GPU virtual machines. See the Virtual Machines guide.

Check live availability and pricing:

targon inventory --gpu
targon inventory --type vm --gpu

Matching Workloads to Tiers​

  • API backends / web hooks: Start with cpu-small or cpu-medium. Increase tiers only if you see sustained CPU saturation.
  • Batch jobs / ETL: Use cpu-large or cpu-xlarge for parallel processing.
  • LLM inference: Pick a tier that matches model size and throughput — start with h200-small and scale up (h200-large / h200-xlarge) as needed.
  • Interactive development: A single dedicated rental with SSH access is often the most flexible starting point.

Cost Optimization Tips​

  • Right-size rentals instead of over-provisioning GPU memory you do not use.
  • Use Volumes to persist model weights so restarts do not require re-downloading.
  • Cache model weights in your container image or on a volume to shorten startup time.
  • Shut down or delete rentals when they are not in use.