P3DN Accelerated Computing
⚡ GPU AcceleratedHardware accelerators (GPUs, FPGAs) for specialized compute.
Name:
pGPU Deep Learning & HPCp = GPU Deep Learning & HPCHigh-end NVIDIA GPUs (Blackwell, Hopper, Ampere, Volta) built for frontier AI/LLM training, HPC simulations, and generative AI.33rd Generation3 = 3rd GenerationAWS 3rd generation hardware platform with updated CPU/system architecture.dDirect-Attached NVMe SSDd = Direct-Attached NVMe SSDEquipped with physical high-speed NVMe SSD instance store volumes attached locally to the host server for temporary scratch and caching.nEnhanced Networkingn = Enhanced NetworkingEnhanced networking capabilities delivering up to 100/200/400 Gbps network bandwidth and Elastic Fabric Adapter (EFA) support.Common use cases
- Machine learning training/inference
- High-performance computing
- Graphics rendering
Instance types (1)
With a 3-year Compute Savings Plan (No Upfront), p3dn sizes on Linux in US East (N. Virginia) (us-east-1) cost 47% less than the on-demand rate. Each size's page has the rates in every region. How Savings Plans are applied.
| Instance | vCPU | Memory | GPU Count | GPU Memory | Memory Type | Architecture | From (Linux, $/hr) |
|---|---|---|---|---|---|---|---|
| p3dn.24xlarge | 96 | 768 GB | 8x | 256.0 GB | Standard | x86_64 | $31.2120 |