Billing

The Billing page on Control Plane shows your balance and the costs of your experiments, as well as means for setting spending limits for your organisations, projects, and organisation members.

Prepayment of credit for use on Strong Compute can be made by visiting the Profile menu on Control Plane and clicking Buy Credits.

Users are billed for consumption of the following product types.

  • Compute (chips under management, CPU or GPU)

  • Data transfer (cluster egress, cluster to cloud bucket, and cloud bucket to cluster)

  • Data storage (in cloud)

Products of these types may be associated with Strong Compute hardware or commercial cloud hardware procured wholesale (e.g. a "burst" workstation or "burst" training cluster). The following rates apply for consumption of cloud wholesale products.

Product
A: Wholesale Cost*
B: Strong Compute Fee
Price You Pay (A+B)

Chip per hour (GPUs under management)

$0.10-15.00

$0.50

$0.60-$15.50

Chip per hour (CPUs under management)**

$0.10-15.00

$0.50

$0.60-$15.50

1 trillion bytes (decimal TB) transferred

$9-110

$2

$11-112

1 trillion bytes (decimal TB) stored per month

$25.50

$2

$27.50

*Range based on and subject to wholesale vendor pricing **Only applies for non-GPU workloads, charged per instance, minimum 1 hr

The following rates also apply specifically when running your container and launching experiments in compute_mode="cycle" or compute_mode="interruptible" on the Sydney Strong Compute cluster.

Product
Amount (USD)
Charged by

Container runtime (CPU)

$1

per hour

Container runtime (GPU)

$1.25

per attached GPU per hour

Experiment compute_mode="cycle"

$1

per experiment cycle

Experiment compute_mode="interruptible"

$1.25

per GPU per hour

Experiment initialisation compute_mode="interruptible"

$18

per experiment

Typical usage charges

Charges associated with typical usage patterns are as follows.

Working in container

  • Data transfer - bucket to local charged at 9¢ per GB for shipping container data to the workstation machine.

  • Container runtime charged at US$1 per hour.

  • Container GPU usage charged at US$1.25 per GPU hour.

  • Data transfer - local to bucket charged at 9¢ per GB for shipping container data changes to cloud storage.

  • Data bucket storage charged at 2.47¢ per GB per month for container data.

Cycle mode experiment

  • Cycle compute charged at US$1 per cycle.

  • Data transfer - local to bucket charged at 9¢ per GB for shipping data to Artifacts such as checkpoints.

  • Data bucket storage charged at 2.47¢ per GB per month for Artifacts such as checkpoints.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading data such as checkpoint Artifacts to the local cluster.

Interruptible mode experiment

  • Interruptible initialization charged at US$18 per experiment.

  • Interruptible compute charged at US$1.25 per GPU per hour.

  • Data transfer - local to bucket charged at 9¢ per GB for shipping data to Artifacts such as checkpoints.

  • Data bucket storage charged at 2.47¢ per GB per month for Artifacts such as checkpoints.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading data such as checkpoint Artifacts to the local cluster.

Burst mode experiment

  • Burst initialization charged at US1¢ per experiment.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading Datasets to the burst cluster for training.

  • Burst compute charged the cloud provider rate per GPU per hour plus Strong Compute overhead.

  • Data transfer - local to bucket charged at 9¢ per GB for shipping data to Artifacts such as checkpoints.

  • Data bucket storage charged at 2.47¢ per GB per month for Artifacts such as checkpoints.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading data such as checkpoint Artifacts to the local cluster.

Working in burst container

  • Data transfer - local to bucket charged at 9¢ per GB for shipping the user container from the container home cluster to a bucket.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading the user container to the burst workstation.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading a requested dataset for mounting to the container.

  • Burst compute charged the cloud provider rate per GPU per hour plus Strong Compute overhead for the burst workstation instance.

  • Data transfer - local to bucket charged at 9¢ per GB for data egress from the container.

  • Data transfer - local to bucket charged at 9¢ per GB for shipping the user container from the burst workstation to a bucket.

  • Data transfer - bucket to local charged at 9¢ per GB for downloading the user container to the container home cluster.

Note: charges for items that have not yet been finalized - such as Data bucket storage for Artifacts that are still being stored - appear on the Current Balance tab. When finalized - such as when an Artifact is deleted - these items will appear on the Billing History tab.

Current balance

The Current Balance page displays a summary of your Organisation's credit balance.

  • Current Balance: Your total credits minus your finalized charges.

  • Total Pending charges: Accrued charges for unfinalized line items.

  • Projected charges: Current balance plus total pending charges. This represents what your current balance would be if all pending charges were finalized at their current accrued amount.

Aggregate charges

Aggregate charges for your Organisation are show on the Billing History tab by product with Daily, Monthly, and Total usage.

Spending limits

This is where you can set limits on an organisation, user, or project level which cannot be exceeded even by an administrator of the organisation. When a limit is reached, all current and future experiments created that would exceed this limit will be cancelled, and a notice is displayed on the Experiments page advising users that a spending limit has been reached. Only an organisation’s Admin and Finance users can set and change spending limits for their organisation.

Statements

Coming soon.

Last updated