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.
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 |
|
Bandwidth, maximum |
|
The following table illustrates how the properties of the VM.PCAStandard.E5.Flex Shape can be optimized for each individual instance.
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.
Specification | Possible Values |
---|---|
Shape name |
|
GPUs | 1-4 – corresponding with the shape name |
OCPUs | 27 per GPU |
Memory | 240 GB per GPU |
VNICs | Up to 24 |
Bandwidth | Up to 400 Gbps |