# All Capacity mode overview

> [!CAUTION]
> **Caution:** The [Compute Engine Service Level Agreement
> (SLA)](https://cloud.google.com/compute/sla) doesn't apply to All Capacity mode TPUs.

TPU capacity is available in two modes: managed capacity mode and All Capacity
mode. For a detailed comparison between these modes, see [TPU capacity
modes](https://docs.cloud.google.com/tpu/docs/capacity-modes).

All Capacity mode provides physically co-located TPU capacity and direct,
reservation-based control over your hardware for optimal performance and
workload scheduling.

Key features of All Capacity mode include:

- **Full capacity access**: Access your entire reserved capacity with zero holdbacks.
- **Full topology visibility**: View the host topology, health, and utilization status of your reservation, blocks, sub-blocks, and hosts.
- **Granular maintenance control**: Target specific VMs, sub-blocks, blocks, or entire reservations to schedule and start maintenance.
- **Direct failure management**: Identify and report faulty hosts directly to initiate hardware repair workflows.

## Supported TPU versions

All Capacity mode is available on TPU v6e (Trillium) and TPU7x (Ironwood). All
Capacity mode is not supported for older TPU versions.

## All Capacity mode terminology

All Capacity mode topology consists of four levels: *cluster* ,
*block* , *sub-block* , and *host*. A cluster is a deployment unit of physical TPU
capacity within one zone. A TPU reservation in All Capacity mode is always
within one cluster.

The following tables show how TPU topology concepts map to physical components:

**TPU v6e**

| Topology concept | Physical component | Cores | Chips | Hosts |
|---|---|---|---|---|
| N/A | Chip | 1 | 1 | N/A |
| Host | Host | 8 | 8 | 1 |
| Sub-block | TPU v6e Pod | 256 | 256 | 32 |
| Block | Multiple TPU v6e Pods (up to 16) in a reservation | Up to 4096 | Up to 4096 | Up to 512 |

A reservation can have one or more blocks, and each block can have 1 to 16 TPU v6e
Pods.

For more information about supported slice sizes, see [TPU v6e supported
configurations](https://docs.cloud.google.com/tpu/docs/v6e#configurations).

**TPU7x**

| Topology concept | Physical component | Cores | Chips | Hosts |
|---|---|---|---|---|
| N/A | Chip | 2 | 1 | N/A |
| Host | Host | 8 | 4 | 1 |
| Sub-block | Cube | 128 | 64 | 16 |
| Block | Multiple TPU7x cubes up to a full Pod | Up to 18,432 | Up to 9,216 (144 cubes) | Up to 2,304 |

A reservation can have one or more TPU7x cubes, up to a full Pod.

For more information about supported slice sizes, see [TPU7x supported
configurations](https://docs.cloud.google.com/tpu/docs/tpu7x#configurations).