Processor state control for Amazon EC2 Linux instances - Amazon Elastic Compute Cloud

Processor state control for Amazon EC2 Linux instances

C-states control the sleep levels that a core can enter when it is idle. C-states are numbered starting with C0 (the shallowest state where the core is totally awake and executing instructions) and go to C6 (the deepest idle state where a core is powered off).

P-states control the desired performance (in CPU frequency) from a core. P-states are numbered starting from P0 (the highest performance setting where the core is allowed to use Intel Turbo Boost Technology to increase frequency if possible), and they go from P1 (the P-state that requests the maximum baseline frequency) to P15 (the lowest possible frequency).

Note

AWS Graviton processors have built-in power saving modes and operate at a fixed frequency. Therefore, they do not provide the ability for the operating system to control C-states and P-states.

C-states and P-states

The following instance types provide the ability for an operating system to control processor C-states and P-states:

  • General purpose: m4.10xlarge | m4.16xlarge

  • Compute optimized: c4.8xlarge

  • Memory optimized: r4.8xlarge | r4.16xlarge | x1.16xlarge | x1.32xlarge | x1e.8xlarge | x1e.16xlarge | x1e.32xlarge

  • Storage optimized: d2.8xlarge | i3.8xlarge | i3.16xlarge | i3en.24xlarge | h1.8xlarge | h1.16xlarge

  • Accelerated computing: f1.16xlarge | g3.16xlarge | | p2.16xlarge | p3.16xlarge

  • Bare metal: All bare metal instances with Intel and AMD processors

C-states only

The following instance types provide the ability for an operating system to control processor C-states:

  • General purpose: m5.12xlarge | m5.24xlarge | m5d.12xlarge | m5d.24xlarge | m5n.12xlarge | m5n.24xlarge | m5dn.12xlarge | m5dn.24xlarge | m5zn.6xlarge | m5zn.12xlarge | m6a.24xlarge | m6a.48xlarge | m6i.16xlarge | m6i.32xlarge | m6id.16xlarge | m6id.32xlarge | m6idn.16xlarge | m6in.16xlarge | m6in.32xlarge | m7a.medium | m7a.large | m7a.xlarge | m7a.2xlarge | m7a.4xlarge | m7a.8xlarge | m7a.12xlarge | m7a.16xlarge | m7a.24xlarge | m7a.32xlarge | m7a.48xlarge | m7i.large | m7i.xlarge | m7i.2xlarge | m7i.4xlarge | m7i.8xlarge | m7i.12xlarge | m7i.16xlarge | m7i.24xlarge | m7i.48xlarge

  • Compute optimized: c5.9xlarge | c5.12xlarge | c5.18xlarge | c5.24xlarge | c5a.24xlarge | c5ad.24xlarge | c5d.9xlarge | c5d.12xlarge | c5d.18xlarge | c5d.24xlarge | c5n.9xlarge | c5n.18xlarge | c6a.24xlarge | c6a.32xlarge | c6a.48xlarge | c6i.16xlarge | c6i.32xlarge | c6id.24xlarge | c6id.32xlarge | c6in.32xlarge | c7a.medium | c7a.large | c7a.xlarge | c7a.2xlarge | c7a.4xlarge | c7a.8xlarge | c7a.12xlarge | c7a.16xlarge | c7a.24xlarge | c7a.32xlarge | c7a.48xlarge | c7i.large | c7i.xlarge | c7i.2xlarge | c7i.4xlarge | c7i.8xlarge | c7i.12xlarge | c7i.16xlarge | c7i.24xlarge | c7i.48xlarge

  • Memory optimized: r5.12xlarge | r5.24xlarge | r5b.12xlarge | r5d.12xlarge | r5d.24xlarge | r5n.12xlarge | r5n.24xlarge | r5dn.12xlarge | r5dn.24xlarge | r6a.24xlarge | r6a.48xlarge | r6i.16xlarge | r6i.32xlarge | r6id.16xlarge | r6id.32xlarge | r6in.16xlarge | r6in.32xlarge | r7a.medium | r7a.large | r7a.xlarge | r7a.2xlarge | r7a.4xlarge | r7a.8xlarge | r7a.12xlarge | r7a.16xlarge | r7a.24xlarge | r7a.32xlarge | r7a.48xlarge | r7i.large | r7i.xlarge | r7i.2xlarge | r7i.4xlarge | r7i.8xlarge | r7i.12xlarge | r7i.16xlarge | r7i.24xlarge | r7i.48xlarge | r7iz.large | r7iz.xlarge | r7iz.2xlarge | r7iz.4xlarge | r7iz.8xlarge | r7iz.12xlarge | r7iz.16xlarge | r7iz.32xlarge | u-3tb1.56xlarge | u-6tb1.56xlarge | u-6tb1.112xlarge | u-9tb1.112xlarge | u-12tb1.112xlarge | u-18tb1.112xlarge | u-24tb1.112xlarge | u7i-12tb.224xlarge | u7in-16tb.224xlarge | u7in-24tb.224xlarge | u7in-32tb.224xlarge | x2idn.32xlarge | x2iedn.16xlarge | x2iezn.12xlarge | z1d.6xlarge | z1d.12xlarge

  • Storage optimized: d3en.12xlarge | dl1.24xlarge | i3en.12xlarge | i3en.24xlarge | i4i.16xlarge | r5b.12xlarge | r5b.24xlarge

  • Accelerated computing: dl1.24xlarge | g5.24xlarge | g5.48xlarge | g6.24xlarge | g6.48xlarge | g6e.12xlarge | g6e.24xlarge | g6e.48xlarge | inf1.24xlarge | p3dn.24xlarge | p4d.24xlarge | p4de.24xlarge | p5.24xlarge | p5e.24xlarge | trn1.32xlarge | vt1.24xlarge

You might want to change the C-state or P-state settings to increase processor performance consistency, reduce latency, or tune your instance for a specific workload. The default C-state and P-state settings provide maximum performance, which is optimal for most workloads. However, if your application would benefit from reduced latency at the cost of higher single- or dual-core frequencies, or from consistent performance at lower frequencies as opposed to bursty Turbo Boost frequencies, consider experimenting with the C-state or P-state settings that are available to these instances.

For information about different processor configurations and how to monitor the effects of your configuration for Amazon Linux, see Processor state control for Amazon EC2 Amazon Linux instance in the Amazon Linux 2 User Guide. These procedures were written for, and apply to Amazon Linux; however, they might also work for other Linux distributions with a Linux kernel of 3.9 or newer. For more information about other Linux distributions and processor state control, see your system-specific documentation.