Oracle Cloud Infrastructure Documentation

Compute Shapes

A shape is a template that determines the type and amount of resources that are allocated to a compute instance. Compute Cloud@Customer offers a choice between a flexible shape for generic workloads, and dedicated shapes for GPU-accelerated workloads.

You choose the shape configuration when you create an instance. See Creating an Instance.

When you create instances using any of the platform images that are provided with Compute Cloud@Customer, all common workloads can be handled with the VM.PCAStandard.E5.Flex shape. Because it's a flexible shape, it lets you customize the number of OCPUs and the amount of memory for each instance. You can optimize instance performance for a specific workload and ensure that resources are used efficiently.

The GPU VM shapes are optimized for enterprise GPU-accelerated workloads. They can only be used if the Compute Cloud@Customer deployment includes a GPU expansion rack. Instances created with a GPU shape have direct (passthrough) access to 1-4 physical GPUs. The ratio between GPUs, OCPUs and memory is fixed.

Generic Flexible Shape

The following table lists the number of OCPUs and the amount of memory you can configure using the VM.PCAStandard.E5.Flex Shape.

The number of OCPUs you select determines the maximum VNIC attachments and the network bandwidth for the instance.

VM.PCAStandard.E5.Flex
Specification Possible Values
Shape name VM.PCAStandard.E5.Flex
OCPUs, minimum - maximum 1 to 96
Memory, default 10 GB per OCPU
Memory, minimum 1 GB per OCPU
Memory, maximum 64 GB per OCPU, up to 960 GB

VNICs, maximum
  • 1 OCPU: 2 VNICs
  • 2 to 24 OCPUs: 1 VNIC per OCPU
  • 25 to 96: 24 VNICs
Bandwidth, maximum
  • 1 to 24 OCPUs: 24.6 Gbps
  • 25 to 40 OCPUs: 1 Gbps per OCPU
  • 41 to 96 OCPUs: 40 Gbps

The following table illustrates how the properties of the VM.PCAStandard.E5.Flex Shape can be optimized for each individual instance.

Shape Configuration Examples
OCPUs Selected Possible Memory Range (GB) Maximum VNICs Maximum Bandwidth (Gbps)
4 4 to 256 4 24.6
20 20 to 960 20 24.6
30 30 to 960 24 30
96 96 to 960 24 40

Dedicated GPU Shapes

For GPU-accelerated workloads, you have a choice between these shapes: VM.GPU.L40S.1, VM.GPU.L40S.2,VM.GPU.L40S.3, VM.GPU.L40S.4. To access these dedicated shapes, you must create an instance based on the Oracle Linux 8 or Oracle Linux 9 platform image.

Note

No GPU drivers are included in the current Oracle Linux platform images. The instance OS detects the allocated GPUs, but to use them, you need the CUDA Toolkit from the NVIDIA developer site to install the required drivers.

The large download and local repository installation need a large amount of disk space. The default 50GB boot volume is insufficient on Oracle Linux 9 and only just large enough on Oracle Linux 8. It is highly recommended to increase the boot volume size to at least 60GB, and extend the file system accordingly.

Installing GPU Drivers in an Oracle Linux 9 Instance

  1. From the command line of the instance, download and install the CUDA Toolkit rpm for your OS.

    $ wget https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda-repo-rhel9-12-8-local-12.8.0_570.86.10-1.x86_64.rpm
$ sudo rpm -i cuda-repo-rhel9-12-8-local-12.8.0_570.86.10-1.x86_64.rpm
$ sudo dnf clean all
$ sudo dnf install cuda-toolkit-12-8

  2. Enable the Oracle Linux 9 EPEL yum repository. Install the dkms package.

    $ sudo yum-config-manager --enable ol9_developer_EPEL
$ sudo dnf install dkms

  3. Install the GPU drivers.

    $ sudo dnf install cuda-12-8

  4. Verify the installation with the NVIDIA System Management Interface.

    $ nvidia-smi
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.86.10              Driver Version: 570.86.10      CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA L40S                    Off |   00000000:00:05.0 Off |                    0 |
| N/A   26C    P8             23W /  350W |       1MiB /  46068MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+
Installing GPU Drivers in an Oracle Linux 8 Instance

  1. From the command line of the instance, download and install the CUDA Toolkit rpm for your OS.

    $ wget https://developer.download.nvidia.com/compute/cuda/12.8.0/local_installers/cuda-repo-rhel8-12-8-local-12.8.0_570.86.10-1.x86_64.rpm
$ sudo rpm -i cuda-repo-rhel8-12-8-local-12.8.0_570.86.10-1.x86_64.rpm
$ sudo dnf clean all
$ sudo dnf install cuda-toolkit-12-8

  2. Enable the Oracle Linux 8 EPEL yum repository. Install the dkms package.

    $ sudo yum-config-manager --enable ol8_developer_EPEL
$ sudo dnf install dkms

  3. Install the GPU drivers.

    $ sudo dnf install cuda-12-8

  4. Install the NVIDIA kernel module.

    $ sudo scl enable gcc-toolset-13 bash
# dkms install nvidia-open -v 570.86.10

    If this make error appears while the kernel module is built, you can safely ignore it.

    Cleaning build area...(bad exit status: 2)
Failed command:
make -C /lib/modules/5.15.0-206.153.7.el8uek.x86_64/build M=/var/lib/dkms/nvidia-open/570.86.10/build clean

  5. Verify the installation with the NVIDIA System Management Interface.

    # nvidia-smi
+-----------------------------------------------------------------------------------------+
| NVIDIA-SMI 570.86.10              Driver Version: 570.86.10      CUDA Version: 12.8     |
|-----------------------------------------+------------------------+----------------------+
| GPU  Name                 Persistence-M | Bus-Id          Disp.A | Volatile Uncorr. ECC |
| Fan  Temp   Perf          Pwr:Usage/Cap |           Memory-Usage | GPU-Util  Compute M. |
|                                         |                        |               MIG M. |
|=========================================+========================+======================|
|   0  NVIDIA L40S                    Off |   00000000:00:05.0 Off |                    0 |
| N/A   26C    P8             23W /  350W |       1MiB /  46068MiB |      0%      Default |
|                                         |                        |                  N/A |
+-----------------------------------------+------------------------+----------------------+

+-----------------------------------------------------------------------------------------+
| Processes:                                                                              |
|  GPU   GI   CI              PID   Type   Process name                        GPU Memory |
|        ID   ID                                                               Usage      |
|=========================================================================================|
|  No running processes found                                                             |
+-----------------------------------------------------------------------------------------+
VM.GPU.L40S.x
Specification Possible Values
Shape name
  • VM.GPU.L40S.1
  • VM.GPU.L40S.2
  • VM.GPU.L40S.3
  • VM.GPU.L40S.4
GPUs 1-4 – corresponding with the shape name
OCPUs 27 per GPU
Memory 200 GB per GPU
VNICs Up to 24
Bandwidth Up to 400 Gbps