Instance Types
Instance types define the compute resources available for your workloads on Phala Cloud. Each instance type provides a specific combination of vCPU, memory, and GPU resources optimized for different use cases.
We recommend specifying your desired instance type directly when creating VMs through the UI or CLI for predictable resource allocation and pricing.
CPU Instance Types
General-purpose compute instances optimized for CPU-intensive workloads, web applications, and backend services.
| ID | Name | vCPU | Memory | Hourly Rate |
|---|---|---|---|---|
| tdx.small | Small TDX Instance | 1 | 2 GB | $0.058000 |
| tdx.medium | Medium TDX Instance | 2 | 4 GB | $0.116000 |
| tdx.large | Large TDX Instance | 4 | 8 GB | $0.232000 |
| tdx.xlarge | XLarge TDX Instance | 8 | 16 GB | $0.464000 |
| tdx.2xlarge | 2XLarge TDX Instance | 16 | 32 GB | $0.928000 |
| tdx.4xlarge | 4XLarge TDX Instance | 32 | 64 GB | $1.856000 |
| tdx.8xlarge | 8XLarge TDX Instance | 64 | 128 GB | $3.712000 |
GPU Instance Types
GPU-accelerated instances for machine learning, AI inference, graphics rendering, and other GPU-intensive workloads.
| ID | Name | vCPU | Memory | GPU |
|---|---|---|---|---|
| h200.small | H200 SXM 141GB | 24 | 192 GB | - |
| h200.16xlarge | H200 SXM 141GB x 8 | 64 | 256 GB | - |
| h200.8x.large | H200 SXM 141GB x 8 | 192 | 1536 GB | - |
Choosing the Right Instance Type
When selecting an instance type for your workload, consider these factors:
- Workload Type - Choose CPU instances for general-purpose computing, GPU instances for ML/AI workloads
- Resource Requirements - Estimate the vCPU and memory needs based on your application's requirements
- Cost Optimization - Choose the smallest instance type that meets your requirements
- Scalability - Start with smaller instances and scale up as needed
Legacy Resource Matching
For backward compatibility, if you specify custom vCPU and memory values instead of an instance type, Phala Cloud automatically matches your request to the nearest instance type that can satisfy your needs (nearest-upper-neighbor matching).
For example, if you request 3 vCPUs and 6GB of memory, the system will select the smallest instance type that provides at least 3 vCPUs and 6GB of memory.
Pricing Formula
For custom resource specifications, pricing is calculated based on Compute Units (CU):
- CU = max(vCPU, memory_GB / 2)
- Compute Rate = CU × $0.058/hr
- Storage Rate = disk_GB × $0.000139/hr
For example, a 4 vCPU, 8GB memory instance has CU = max(4, 8/2) = 4, resulting in a compute rate of 4 × $0.058 = $0.232/hr (plus storage costs).
