Compute optimized instances - Amazon Elastic Compute Cloud
Hardware specificationsInstance performanceNetwork performanceSSD I/O performanceInstance featuresRelease notes

Compute optimized instances

Compute optimized instances are ideal for compute-bound applications that benefit from high-performance processors.

C5 and C5n instances

These instances are well suited for the following:

  • Batch processing workloads

  • Media transcoding

  • High-performance web servers

  • High-performance computing (HPC)

  • Scientific modeling

  • Dedicated gaming servers and ad serving engines

  • Machine learning inference and other compute-intensive applications

Bare metal instances, such as c5.metal, provide your applications with direct access to physical resources of the host server, such as processors and memory.

For more information, see Amazon EC2 C5 Instances.

C6g, C6gd, and C6gn instances

These instances are powered by AWS Graviton2 processors and are ideal for running advanced, compute-intensive workloads, such as the following:

  • High-performance computing (HPC)

  • Batch processing

  • Ad serving

  • Video encoding

  • Gaming servers

  • Scientific modeling

  • Distributed analytics

  • CPU-based machine learning inference

Bare metal instances, such as c6g.metal, provide your applications with direct access to physical resources of the host server, such as processors and memory.

For more information, see Amazon EC2 C6g Instances.

C6i and C6id instances

These instances are ideal for running advanced, compute-intensive workloads, such as the following:

  • High-performance computing (HPC)

  • Batch processing

  • Ad serving

  • Video encoding

  • Distributed analytics

  • Highly scalable multiplayer gaming

For more information, see Amazon EC2 C6i Instances.

Hpc6a instances

These instances are ideal for running high performance computing (HPC) workloads, such as the following:

  • Molecular dynamics

  • Computational chemistry

  • Computational fluid dynamics

  • Weather forecasting

  • Materials simulation

  • Crash simulations

  • Astrophysics

For more information, see Amazon EC2 Hpc6a Instances.

Hardware specifications

The following is a summary of the hardware specifications for compute optimized instances. A virtual central processing unit (vCPU) represents a portion of the physical CPU assigned to a virtual machine (VM). For x86 instances, there are two vCPUs per core. For Graviton instances, there is one vCPU per core.

Instance type Default vCPUs Memory (GiB)
c4.large 2 3.75
c4.xlarge 4 7.5
c4.2xlarge 8 15
c4.4xlarge 16 30
c4.8xlarge 36 60
c5.large 2 4
c5.xlarge 4 8
c5.2xlarge 8 16
c5.4xlarge 16 32
c5.9xlarge 36 72
c5.12xlarge 48 96
c5.18xlarge 72 144
c5.24xlarge 96 192
c5.metal 96 192
c5a.large 2 4
c5a.xlarge 4 8
c5a.2xlarge 8 16
c5a.4xlarge 16 32
c5a.8xlarge 32 64
c5a.12xlarge 48 96
c5a.16xlarge 64 128
c5a.24xlarge 96 192
c5ad.large 2 4
c5ad.xlarge 4 8
c5ad.2xlarge 8 16
c5ad.4xlarge 16 32
c5ad.8xlarge 32 64
c5ad.12xlarge 48 96
c5ad.16xlarge 64 128
c5ad.24xlarge 96 192
c5d.large 2 4
c5d.xlarge 4 8
c5d.2xlarge 8 16
c5d.4xlarge 16 32
c5d.9xlarge 36 72
c5d.12xlarge 48 96
c5d.18xlarge 72 144
c5d.24xlarge 96 192
c5d.metal 96 192
c5n.large 2 5.25
c5n.xlarge 4 10.5
c5n.2xlarge 8 21
c5n.4xlarge 16 42
c5n.9xlarge 36 96
c5n.18xlarge 72 192
c5n.metal 72 192
c6a.large 2 4
c6a.xlarge 4 8
c6a.2xlarge 8 16
c6a.4xlarge 16 32
c6a.8xlarge 32 64
c6a.12xlarge 48 96
c6a.16xlarge 64 128
c6a.24xlarge 96 192
c6a.32xlarge 128 256
c6a.48xlarge 192 384
c6a.metal 192 384
c6g.medium 1 2
c6g.large 2 4
c6g.xlarge 4 8
c6g.2xlarge 8 16
c6g.4xlarge 16 32
c6g.8xlarge 32 64
c6g.12xlarge 48 96
c6g.16xlarge 64 128
c6g.metal 64 128
c6gd.medium 1 2
c6gd.large 2 4
c6gd.xlarge 4 8
c6gd.2xlarge 8 16
c6gd.4xlarge 16 32
c6gd.8xlarge 32 64
c6gd.12xlarge 48 96
c6gd.16xlarge 64 128
c6gd.metal 64 128
c6gn.medium 1 2
c6gn.large 2 4
c6gn.xlarge 4 8
c6gn.2xlarge 8 16
c6gn.4xlarge 16 32
c6gn.8xlarge 32 64
c6gn.12xlarge 48 96
c6gn.16xlarge 64 128
c6i.large 2 4
c6i.xlarge 4 8
c6i.2xlarge 8 16
c6i.4xlarge 16 32
c6i.8xlarge 32 64
c6i.12xlarge 48 96
c6i.16xlarge 64 128
c6i.24xlarge 96 192
c6i.32xlarge 128 256
c6i.metal 128 256
c6id.large 2 4
c6id.xlarge 4 8
c6id.2xlarge 8 16
c6id.4xlarge 16 32
c6id.8xlarge 32 64
c6id.12xlarge 48 96
c6id.16xlarge 64 128
c6id.24xlarge 96 192
c6id.32xlarge 128 256
c6id.metal 128 256
c7g.medium 1 2
c7g.large 2 4
c7g.xlarge 4 8
c7g.2xlarge 8 16
c7g.4xlarge 16 32
c7g.8xlarge 32 64
c7g.12xlarge 48 96
c7g.16xlarge 64 128
hpc6a.48xlarge 96 384

The compute optimized instances use the following processors.

AWS Graviton processors

  • AWS Graviton2: C6g, C6gd, C6gn

  • AWS Graviton3: C7g

AMD processors

  • 2nd generation AMD EPYC processors (AMD EPYC 7R32): C5a, C5ad

  • 3rd generation AMD EPYC processors (AMD EPYC 7R13): C6a, Hpc6a

Intel processors

  • Intel Xeon Scalable processors (Haswell E5-2666 v3): C4

  • Intel Xeon Scalable processors (Skylake 8124): C5n

  • Intel Xeon Scalable processors (Skylake 8124M or Cascade Lake 8223CL): Smaller C5 and C5d

  • 2nd generation Intel Xeon Scalable processors (Cascade Lake 8275CL): Larger C5 and C5d

  • 3rd generation Intel Xeon Scalable processors (Ice Lake 8375C): C6i, C6id

For more information, see Amazon EC2 Instance Types.

Instance performance

EBS-optimized instances enable you to get consistently high performance for your EBS volumes by eliminating contention between Amazon EBS I/O and other network traffic from your instance. Some compute optimized instances are EBS-optimized by default at no additional cost. For more information, see Amazon EBS–optimized instances.

Some compute optimized instance types provide the ability to control processor C-states and P-states on Linux. C-states control the sleep levels that a core can enter when it is inactive, while P-states control the desired performance (in CPU frequency) from a core. For more information, see Processor state control for your EC2 instance.

Network performance

You can enable enhanced networking on supported instance types to provide lower latencies, lower network jitter, and higher packet-per-second (PPS) performance. Most applications do not consistently need a high level of network performance, but can benefit from access to increased bandwidth when they send or receive data. For more information, see Enhanced networking on Linux.

The following is a summary of network performance for compute optimized instances that support enhanced networking.

Instance type Network performance Enhanced networking
c4.large Moderate Intel 82599 VF
c4.xlarge | c4.2xlarge | c4.4xlarge High Intel 82599 VF
c5.4xlarge and smaller | c5a.4xlarge and smaller | c5ad.4xlarge and smaller | c5d.4xlarge and smaller | c6g.4xlarge and smaller | c6gd.4xlarge and smaller Up to 10 Gbps † ENA
c4.8xlarge 10 Gbps Intel 82599 VF
c5.9xlarge | c5a.8xlarge | c5ad.8xlarge | c5d.9xlarge 10 Gbps ENA
c5.12xlarge | c5a.12xlarge | c5ad.12xlarge | c5d.12xlarge | c6g.8xlarge | c6gd.8xlarge 12 Gbps ENA
c6a.4xlarge and smaller | c6i.4xlarge and smaller | c6id.4xlarge and smaller | c7g.xlarge and smaller Up to 12.5 Gbps † ENA
c6a.8xlarge | c6i.8xlarge | c6id.8xlarge 12.5 Gbps ENA
c7g.2xlarge | c7g.4xlarge Up to 15 Gbps ENA
c7g.8xlarge 15 Gbps
c6a.12xlarge | c6i.12xlarge | c6id.12xlarge 18.75 Gbps ENA
c5a.16xlarge | c5a.24xlarge | c5ad.16xlarge | c5ad.24xlarge | c6g.12xlarge | c6gd.12xlarge 20 Gbps ENA
c7g.12xlarge 22.5 Gbps ENA
c5n.4xlarge and smaller | c6gn.4xlarge and smaller Up to 25 Gbps † ENA
c5.18xlarge | c5.24xlarge | c5.metal | c5d.18xlarge | c5d.24xlarge | c5d.metal | c6a.16xlarge | c6g.16xlarge | c6g.metal | c6gd.16xlarge | c6gd.metal | c6gn.4xlarge | c6i.16xlarge | c6id.16xlarge 25 Gbps ENA
c7g.16xlarge 30 Gbps ENA
c6a.24xlarge | c6i.24xlarge | c6id.24xlarge 37.5 Gbps ENA
c5n.9xlarge | c6a.32xlarge | c6a.48xlarge | c6a.metal | c6gn.8xlarge | c6i.32xlarge | c6i.metal | c6id.32xlarge | c6id.metal 50 Gbps ENA
c6gn.12xlarge 75 Gbps ENA
c5n.18xlarge | c5n.metal | c6gn.16xlarge | hpc6a.48xlarge 100 Gbps ENA, EFA

† These instances have a baseline bandwidth and can use a network I/O credit mechanism to burst beyond their baseline bandwidth on a best effort basis. For more information, see instance network bandwidth.

Instance type Baseline bandwidth (Gbps) Burst bandwidth (Gbps)
c5.large .75 10
c5.xlarge 1.25 10
c5.2xlarge 2.5 10
c5.4xlarge 5 10
c5a.large .75 10
c5a.xlarge 1.25 10
c5a.2xlarge 2.5 10
c5a.4xlarge 5 10
c5ad.large .75 10
c5ad.xlarge 1.25 10
c5ad.2xlarge 2.5 10
c5ad.4xlarge 5 10
c5d.large .75 10
c5d.xlarge 1.25 10
c5d.2xlarge 2.5 10
c5d.4xlarge 5 10
c5n.large 3 25
c5n.xlarge 5 25
c5n.2xlarge 10 25
c5n.4xlarge 15 25
c6a.large .781 12
c6a.xlarge 1.562 12
c6a.2xlarge 3.125 12
c6a.4xlarge 6.25 12
c6g.medium .5 10
c6g.large .75 10
c6g.xlarge 1.25 10
c6g.2xlarge 2.5 10
c6g.4xlarge 5 10
c6gd.medium .5 10
c6gd.large .75 10
c6gd.xlarge 1.25 10
c6gd.2xlarge 2.5 10
c6gd.4xlarge 5 10
c6gn.medium 1.6 25
c6gn.large 3 25
c6gn.xlarge 6.3 25
c6gn.2xlarge 12.5 25
c6gn.4xlarge 15 25
c6i.large .781 12.5
c6i.xlarge 1.562 12.5
c6i.2xlarge 3.125 12.5
c6i.4xlarge 6.25 12.5
c6id.large .781 12.5
c6id.xlarge 1.562 12.5
c6id.2xlarge 3.125 12.5
c6id.4xlarge 6.25 12.5

SSD I/O performance

If you use a Linux AMI with kernel version 4.4 or later and use all the SSD-based instance store volumes available to your instance, you can get up to the IOPS (4,096 byte block size) performance listed in the following table (at queue depth saturation). Otherwise, you get lower IOPS performance.

Instance Size 100% Random Read IOPS Write IOPS
c5ad.large 16,283 7,105
c5ad.xlarge 32,566 14,211
c5ad.2xlarge 65,132 28,421
c5ad.4xlarge 130,263 56,842
c5ad.8xlarge 260,526 113,684
c5ad.12xlarge 412,500 180,000
c5ad.16xlarge 521,053 227,368
c5ad.24xlarge 825,000 360,000
c5d.large 20,000 9,000
c5d.xlarge 40,000 18,000
c5d.2xlarge 80,000 37,000
c5d.4xlarge 175,000 75,000
c5d.9xlarge 350,000 170,000
c5d.12xlarge 700,000 340,000
c5d.18xlarge 700,000 340,000
c5d.24xlarge 1,400,000 680,000
c5d.metal 1,400,000 680,000
c6gd.medium 13,438 5,625
c6gd.large 26,875 11,250
c6gd.xlarge 53,750 22,500
c6gd.2xlarge 107,500 45,000
c6gd.4xlarge 215,000 90,000
c6gd.8xlarge 430,000 180,000
c6gd.12xlarge 645,000 270,000
c6gd.16xlarge 860,000 360,000
c6gd.metal 860,000 360,000
c6id.large 33,542 16,771
c6id.xlarge 67,083 33,542
c6id.2xlarge 134,167 67,084
c6id.4xlarge 268,333 134,167
c6id.8xlarge 536,666 268,334
c6id.12xlarge 804,999 402,501
c6id.16xlarge 1,073,332 536,668
c6id.24xlarge 1,609,998 805,002
c6id.32xlarge 2,146,664 1,073,336
c6id.metal 2,146,664 1,073,336

As you fill the SSD-based instance store volumes for your instance, the number of write IOPS that you can achieve decreases. This is due to the extra work the SSD controller must do to find available space, rewrite existing data, and erase unused space so that it can be rewritten. This process of garbage collection results in internal write amplification to the SSD, expressed as the ratio of SSD write operations to user write operations. This decrease in performance is even larger if the write operations are not in multiples of 4,096 bytes or not aligned to a 4,096-byte boundary. If you write a smaller amount of bytes or bytes that are not aligned, the SSD controller must read the surrounding data and store the result in a new location. This pattern results in significantly increased write amplification, increased latency, and dramatically reduced I/O performance.

SSD controllers can use several strategies to reduce the impact of write amplification. One such strategy is to reserve space in the SSD instance storage so that the controller can more efficiently manage the space available for write operations. This is called over-provisioning. The SSD-based instance store volumes provided to an instance don't have any space reserved for over-provisioning. To reduce write amplification, we recommend that you leave 10% of the volume unpartitioned so that the SSD controller can use it for over-provisioning. This decreases the storage that you can use, but increases performance even if the disk is close to full capacity.

For instance store volumes that support TRIM, you can use the TRIM command to notify the SSD controller whenever you no longer need data that you've written. This provides the controller with more free space, which can reduce write amplification and increase performance. For more information, see Instance store volume TRIM support.

Instance features

The following is a summary of features for compute optimized instances:

EBS only NVMe EBS Instance store Placement group
C4 Yes No No Yes
C5 Yes Yes No Yes
C5a Yes Yes No Yes
C5ad No Yes NVMe * Yes
C5d No Yes NVMe * Yes
C5n Yes Yes No Yes
C6a Yes Yes No Yes
C6g Yes Yes No Yes
C6gd No Yes NVMe * Yes
C6gn Yes Yes No Yes
C6i Yes Yes No Yes
C6id No Yes NVMe * Yes
C7g Yes Yes No Yes
Hpc6a Yes Yes No Yes

* The root device volume must be an Amazon EBS volume.

For more information, see the following:

Release notes

  • C4 instances and instances built on the Nitro System require 64-bit EBS-backed HVM AMIs. They have high-memory and require a 64-bit operating system to take advantage of that capacity. HVM AMIs provide superior performance in comparison to paravirtual (PV) AMIs on high-memory instance types. In addition, you must use an HVM AMI to take advantage of enhanced networking.

  • C7g instances are powered by the latest generation AWS Graviton3 processors. They offer up to 25% better performance over the sixth generation AWS Graviton2-based C6g instances. C7g instances are the first in the cloud to feature DDR5 memory, which provides 50% higher memory bandwidth compared to DDR4 memory to enable high-speed access to data in memory.

    • C7g instances limit the aggregate execution rate of high power instructions such as floating point multiply accumulates across cores in an instance. Future instance types focused on other workload segments may not have this restriction

  • Instances built on the Nitro System have the following requirements:

    The following Linux AMIs meet these requirements:

    • Amazon Linux 2

    • Amazon Linux AMI 2018.03

    • Ubuntu 14.04 (with linux-aws kernel) or later

    • Red Hat Enterprise Linux 7.4 or later

    • SUSE Linux Enterprise Server 12 SP2 or later

    • CentOS 7.4.1708 or later

    • FreeBSD 11.1 or later

    • Debian GNU/Linux 9 or later

  • Instances with an AWS Graviton processors have the following requirements:

    • Use an AMI for the 64-bit Arm architecture.

    • Support booting through UEFI with ACPI tables and support ACPI hot-plug of PCI devices.

    The following AMIs meet these requirements:

    • Amazon Linux 2 (64-bit Arm)

    • Ubuntu 16.04 or later (64-bit Arm)

    • Red Hat Enterprise Linux 8.0 or later (64-bit Arm)

    • SUSE Linux Enterprise Server 15 or later (64-bit Arm)

    • Debian 10 or later (64-bit Arm)

  • To get the best performance from your C6i instances, ensure that they have ENA driver version 2.2.9 or later. Using an ENA driver earlier than version 1.2 with these instances causes network interface attachment failures. The following AMIs have a compatible ENA driver.

    • Amazon Linux 2 with kernel 4.14.186

    • Ubuntu 20.04 with kernel 5.4.0-1025-aws

    • Red Hat Enterprise Linux 8.3 with kernel 4.18.0-240.1.1.el8_3.ARCH

    • SUSE Linux Enterprise Server 15 SP2 with kernel 5.3.18-24.15.1

  • Instances built on the Nitro System instances support a maximum of 28 attachments, including network interfaces, EBS volumes, and NVMe instance store volumes. For more information, see Nitro System volume limits.

  • To get the best performance from your C6gn instances, ensure that they have ENA driver version 2.2.9 or later. Using an ENA driver earlier than version 1.2 with these instances causes network interface attachment failures. The following AMIs have a compatible ENA driver.

    • Amazon Linux 2 with kernel 4.14.186

    • Ubuntu 20.04 with kernel 5.4.0-1025-aws

    • Red Hat Enterprise Linux 8.3 with kernel 4.18.0-240.1.1.el8_3.ARCH

    • SUSE Linux Enterprise Server 15 SP2 with kernel 5.3.18-24.15.1

  • To launch AMIs for all Linux distributions on C6gn instances, use AMIs with the latest version and run an update for the latest driver. For earlier versions, download the latest driver from GitHub.

  • Launching a bare metal instance boots the underlying server, which includes verifying all hardware and firmware components. This means that it can take 20 minutes from the time the instance enters the running state until it becomes available over the network.

  • To attach or detach EBS volumes or secondary network interfaces from a bare metal instance requires PCIe native hotplug support. Amazon Linux 2 and the latest versions of the Amazon Linux AMI support PCIe native hotplug, but earlier versions do not. You must enable the following Linux kernel configuration options:

    CONFIG_HOTPLUG_PCI_PCIE=y
CONFIG_PCIEASPM=y

  • Bare metal instances use a PCI-based serial device rather than an I/O port-based serial device. The upstream Linux kernel and the latest Amazon Linux AMIs support this device. Bare metal instances also provide an ACPI SPCR table to enable the system to automatically use the PCI-based serial device. The latest Windows AMIs automatically use the PCI-based serial device.

  • Instances built on the Nitro System should have acpid installed to support clean shutdown through API requests.

  • There is a limit on the total number of instances that you can launch in a Region, and there are additional limits on some instance types. For more information, see How many instances can I run in Amazon EC2? in the Amazon EC2 FAQ.