TPU v5e

TPU v5e is a cost-efficient generation built for training and serving transformer, text-to-image and convolutional models. Each chip has 1 TensorCore and a pod holds 256 chips in a 2D torus.

Use the specification topic to size a slice, the configurations topic to choose a slice shape, and the VM types topic to see what each VM provides.

• TPU v5e specifications
  Per-chip and per-pod specifications for TPU v5e.

• TPU v5e supported configurations
  The slice shapes you can request for TPU v5e, and how many chips and hosts each uses.

• TPU v5e VM types
  The VM types available for TPU v5e and the host resources each one provides.

• Choose a TPU v5e slice shape
  Choose the slice shape for a TPU v5e TPU VM from the number of chips your workload needs, then check your quota before you create the VM.

Related information

TPU v5e specifications

TPU v5e supported configurations

TPU v5e VM types

Create a TPU VM