Oracle Cloud Infrastructure Documentation

Compute Shapes

A shape is a template that determines the number of CPUs, amount of memory, and other resources allocated to a newly created instance.

The following tables provide basic information about the shapes that are available for bare metal instances, virtual machines (VMs), and dedicated virtual machine hosts.

Note

AMD shapes, X5-based and X6-based shapes, and high performance computing (HPC) shapes are not available in Government Cloud realms.

Bare Metal Shapes

The following shapes are available for bare metal instances:

Network bandwidth is based on expected bandwidth for traffic within a VCN.

Standard Shapes

Designed for general purpose workloads and suitable for a wide range of applications and use cases. Standard shapes provide a balance of cores, memory, and network resources. Standard shapes are available with Intel or AMD processors.

These are the bare metal standard series:

  • BM.Standard1: X5-based standard compute. Processor: 2.3 GHz Intel Xeon E5-2699 v3.

    X5-based shapes availability is limited to monthly universal credit customers existing on or before November 9, 2018, in the US West (Phoenix), US East (Ashburn), and Germany Central (Frankfurt) regions.

  • BM.Standard.B1: X6-based standard compute. Processor: 2.2 GHz Intel Xeon E5-2699 v4.
  • BM.Standard2: X7-based standard compute. Processor: 2.0 GHz Intel Xeon Platinum 8167M.
  • BM.Standard.E2: E2-based standard compute. Processor: 2.0 GHz AMD EPYC 7551.
ShapeOCPUMemory (GB)Local Disk

Max Network Bandwidth

Max VNICs Total: LinuxMax VNICs Total: Windows
BM.Standard1.3636256Block storage only10 Gbps361
BM.Standard.B1.4444512Block storage only25 Gbps44None
BM.Standard2.5252768Block storage only2 x 25 Gbps

52 total (26 per physical NIC)

27 total (1 on the first physical NIC, 26 on the second)
BM.Standard.E2.6464512Block storage only2 x 25 Gbps7576 (1 on the first physical NIC, 75 on the second)

Dense I/O Shapes

Designed for large databases, big data workloads, and applications that require high-performance local storage. DenseIO shapes include locally-attached NVMe-based SSDs.

These are the bare metal dense I/O series:

  • BM.DenseIO1: X5-based dense I/O compute. Processor: 2.3 GHz Intel Xeon E5-2699 v3.

    X5-based shapes availability is limited to monthly universal credit customers existing on or before November 9, 2018, in the US West (Phoenix), US East (Ashburn), and Germany Central (Frankfurt) regions.

  • BM.DenseIO2: X7-based dense I/O compute. Processor: 2.0 GHz Intel Xeon Platinum 8167M.
ShapeOCPUMemory (GB)Local Disk Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows
BM.DenseIO1.363651228.8 TB NVMe SSD (9 drives)10 Gbps361
BM.DenseIO2.5252768

51.2 TB NVMe SSD (8 drives)

2 x 25 Gbps52 total (26 per physical NIC)27 total (1 on the first physical NIC, 26 on the second)

GPU Shapes

Designed for hardware-accelerated workloads. GPU shapes include Intel CPUs and NVIDIA graphics processors.

These are the bare metal GPU series:

  • BM.GPU2: X7-based GPU compute.

    • GPU: NVIDIA Tesla P100
    • CPU: 2.0 GHz Intel Xeon Platinum 8167M
  • BM.GPU3: X7-based GPU compute.

    • GPU: NVIDIA Tesla V100
    • CPU: 2.0 GHz Intel Xeon Platinum 8167M
ShapeOCPUMemory (GB)Local Disk Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows

BM.GPU2.2

(GPU: 2xP100)

28192Block storage only2 x 25 Gbps2815 (1 on the first physical NIC, 14 on the second)

BM.GPU3.8

(GPU: 8xV100)

52768Block storage only2 x 25 Gbps5227 (1 on the first physical NIC, 26 on the second)

HPC Shapes

Designed for high-performance computing workloads that require high frequency processor cores and cluster networking for massively parallel HPC workloads.

This is the bare metal HPC series:

  • BM.HPC2: X7-based high frequency compute. Processor: 3.0 GHz Intel Xeon Gold 6154.
ShapeOCPUMemory (GB)Local Disk Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows
BM.HPC2.3636384

6.7 TB NVMe SSD (1 drive)

1 x 25 Gbps

1 x 100 Gbps RDMA

501

VM Shapes

The following shapes are available for VMs:

Network bandwidth is based on expected bandwidth for traffic within a VCN.

Standard Shapes

Designed for general purpose workloads and suitable for a wide range of applications and use cases. Standard shapes provide a balance of cores, memory, and network resources. Standard shapes are available with Intel or AMD processors.

These are the VM standard series:

  • VM.Standard1: X5-based standard compute. Processor: 2.3 GHz Intel Xeon E5-2699 v3.

    X5-based shapes availability is limited to monthly universal credit customers existing on or before November 9, 2018, in the US West (Phoenix), US East (Ashburn), and Germany Central (Frankfurt) regions.

  • VM.Standard.B1: X6-based standard compute. Processor: 2.2 GHz Intel Xeon E5-2699 v4.
  • VM.Standard2: X7-based standard compute. Processor: 2.0 GHz Intel Xeon Platinum 8167M.
  • VM.Standard.E2.1.Micro: E2-based standard compute with AMD CPUs. Processor: 2.0 GHz AMD EPYC 7551.
  • VM.Standard.E2: E2-based standard compute. Processor: 2.0 GHz AMD EPYC 7551.
ShapeOCPUMemory (GB)Local Disk (TB)Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows
VM.Standard1.117Block Storage only600 Mbps21
VM.Standard1.2214Block Storage only1.2 Gbps21
VM.Standard1.4428Block Storage only1.2 Gbps41
VM.Standard1.8856Block Storage only2.4 Gbps81
VM.Standard1.1616112Block Storage only4.8 Gbps161
VM.Standard.B1.1112Block Storage only 600 Mbps22
VM.Standard.B1.2224Block Storage only1.2 Gbps22
VM.Standard.B1.4448Block Storage only 2.4 Gbps44
VM.Standard.B1.8896Block Storage only4.8 Gbps88
VM.Standard.B1.1616192Block Storage only 9.6 Gbps1616
VM.Standard2.1115Block Storage only1 Gbps22
VM.Standard2.2230Block Storage only2 Gbps22
VM.Standard2.4460Block Storage only4.1 Gbps44
VM.Standard2.88120Block Storage only8.2 Gbps88
VM.Standard2.1616240Block Storage only16.4 Gbps1616
VM.Standard2.2424320Block Storage only24.6 Gbps2424
VM.Standard.E2.1.Micro

1

See Details of the Always Free Resources.

1Block Storage only480 Mbps1-
VM.Standard.E2.118Block Storage only700 Mbps22
VM.Standard.E2.2216Block Storage only1.4 Gbps22
VM.Standard.E2.4432Block Storage only2.8 Gbps44
VM.Standard.E2.8864Block Storage only5.6 Gbps44

Dense I/O Shapes

Designed for large databases, big data workloads, and applications that require high-performance local storage. DenseIO shapes include locally-attached NVMe-based SSDs.

These are the VM dense I/O series:

  • VM.DenseIO1: X5-based dense I/O compute. Processor: 2.3 GHz Intel Xeon E5-2699 v3.

    X5-based shapes availability is limited to monthly universal credit customers existing on or before November 9, 2018, in the US West (Phoenix), US East (Ashburn), and Germany Central (Frankfurt) regions.

  • VM.DenseIO2: X7-based dense I/O compute. Processor: 2.0 GHz Intel Xeon Platinum 8167M.
ShapeOCPUMemory (GB)Local Disk (TB)Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows
VM.DenseIO1.44603.2 TB NVMe SSD1.2 Gbps41
VM.DenseIO1.881206.4 TB NVMe SSD2.4 Gbps81
VM.DenseIO1.161624012.8 TB NVMe SSD4.8 Gbps161
VM.DenseIO2.881206.4 TB NVMe SSD8.2 Gbps88
VM.DenseIO2.161624012.8 TB NVMe SSD16.4 Gbps1616
VM.DenseIO2.242432025.6 TB NVMe SSD24.6 Gbps2424

GPU Shapes

Designed for hardware-accelerated workloads. GPU shapes include Intel CPUs and NVIDIA graphics processors.

These are the VM GPU series:

  • VM.GPU2: X7-based GPU compute.

    • GPU: NVIDIA Tesla P100
    • CPU: 2.0 GHz Intel Xeon Platinum 8167M
  • VM.GPU3: X7-based GPU compute.

    • GPU: NVIDIA Tesla V100
    • CPU: 2.0 GHz Intel Xeon Platinum 8167M
ShapeOCPUMemory (GB)Local Disk (TB)Max Network BandwidthMax VNICs Total: LinuxMax VNICs Total: Windows

VM.GPU2.1

(GPU: 1xP100)

1272Block Storage only8 Gbps1212

VM.GPU3.1

(GPU: 1xV100)

690Block Storage only4 Gbps66

VM.GPU3.2

(GPU: 2xV100)

12180Block Storage only8 Gbps1212

VM.GPU3.4

(GPU: 4xV100)

24360Block Storage only24.6 Gbps2424

Dedicated Virtual Machine Host Shapes

ShapeInstance TypeBilled OCPUUsable OCPU1Supported Shapes for Hosted VMs
DVH.Standard2.52X7-based VM host5248VM.Standard2

1: The difference between billed OCPUs and usable OCPUs is due to OCPUs reserved for hypervisor use.