Compute Optimized
Compute optimized instances are suitable for compute-intensive applications. These instances can be used for batch processing workloads, media transcoding, high-performance web servers, high-performance computing (HPC), scientific modeling, dedicated game servers and ad server engines, machine learning inference, and other compute-intensive applications.
- Applicable types:
c1ix,c2a
c1ix
The c1ix instance uses 2nd-generation Intel Xeon Scalable processors and supports a high turbo clock speed of up to 4.0GHz. It is cost-optimized for workloads that require high CPU performance and delivers strong performance per vCPU.
Hardware specifications
- Up to 4.0GHz 2nd-generation Intel Xeon Scalable processor (Gold 5220R Cascade Lake)
- Up to 10Gbps network bandwidth
- Instance sizes supporting up to 32 vCPUs and 64GiB of memory
- A 2:1 memory-to-vCPU ratio across all sizes
Details
| Instance size | vCPU | Memory (GiB) | Network bandwidth (Gbps) |
|---|---|---|---|
c1ix.large | 2 | 4 | Max 1 |
c1ix.xlarge | 4 | 8 | Max 1 |
c1ix.2xlarge | 8 | 16 | Max 2 |
c1ix.4xlarge | 16 | 32 | Max 5 |
c1ix.8xlarge | 32 | 64 | Max 10 |
c2a
The c2a instance is powered by 3rd-generation AMD EPYC 7003 series processors and is designed for compute-intensive workloads.
Hardware specifications
- Up to 3.6GHz 3rd generation AMD EPYC processor (AMD EPYC 7643)
- Up to 25Gbps network bandwidth
- Instance sizes supporting up to 96 vCPUs and 192GiB memory
- A 2:1 Memory to vCPU ratio across all sizes
- Support for AMD instruction sets (AVX, AVX2)
Details
| Instance size | vCPU | Memory (GiB) | Network bandwidth (Gbps) |
|---|---|---|---|
c2a.large | 2 | 4 | Max 10 |
c2a.xlarge | 4 | 8 | Max 10 |
c2a.2xlarge | 8 | 16 | Max 10 |
c2a.4xlarge | 16 | 32 | Max 10 |
c2a.8xlarge | 32 | 64 | Max 10 |
c2a.12xlarge | 48 | 96 | Max 12.5 |
c2a.16xlarge | 64 | 128 | Max 12.5 |
c2a.24xlarge | 96 | 192 | Max 25 |
NUMA topology
Non-Uniform Memory Access (NUMA) architecture allows each CPU to access its own allocated memory (local memory). Using NUMA architecture can enable high scalability as it allows multiple processors to share memory.
c2a instance has the following NUMA topology depending on the instance size.
| Instance size | Number of NUMA domains | Cores per NUMA domain |
|---|---|---|
c2a.large | 1 | 1 |
c2a.xlarge | 1 | 2 |
c2a.2xlarge | 1 | 4 |
c2a.4xlarge | 1 | 8 |
c2a.8xlarge | 2 | 8 |
c2a.12xlarge | 2 | 12 |
c2a.16xlarge | 2 | 16 |
c2a.24xlarge | 2 | 24 |
NUMA topology architecture
c2a.large
c2a.xlarge
c2a.2xlarge
c2a.4xlarge
c2a.8xlarge
c2a.12xlarge
c2a.16xlarge
c2a.24xlarge
