Scale Up

Medium Project

Recharge

Up to 50,000 GPU-hours

  • Extended allocation beyond Basic Access
  • Multi-node jobs up to 64 GPUs (8 nodes)
  • Dedicated RDS partnership for scaling
  • Priority scheduling within allocation
  • Billed at Stanford recharge rate
Apply for Medium Project

Requires Basic Access first. See the Project Application Guide.

Full Scale

Large Project

Recharge

50,000+ GPU-hours

  • Cluster-scale allocations for frontier research
  • Multi-node jobs up to 248 GPUs (full cluster)
  • Deep RDS engagement and code optimization
  • Scheduling coordination for large runs
  • Reviewed by allocation committee
Apply for Large Project

Requires Basic Access first. See the Project Application Guide.

How It Works

1

Apply

Faculty PIs complete the Basic Access form. Most applications are processed within a few business days.

2

Onboard

Receive your Marlowe account, SSH access, and an introduction to the platform from the RDS team.

3

Compute

Submit jobs via Slurm, access the parallel filesystem, and start training at scale.

4

Scale

When you're ready for more, apply for a Medium or Large Project with your RDS partner.

Frequently Asked Questions

Who is eligible?

Marlowe is available to all Stanford faculty PIs conducting research that benefits from GPU-accelerated computing. Graduate students and postdocs can use the system under a faculty PI's allocation.

What does "recharge" mean?

Beyond the free Basic Access tier, GPU-hours are billed at Stanford's internal recharge rate. Costs are charged to your research account (PTA). See recharge rates for details.

How long does approval take?

Basic Access applications are typically processed within a few business days. Medium and Large Projects are reviewed on a rolling basis by the allocation committee.

What if I need help getting started?

Every Marlowe PI is paired with a Research Data Scientist who can help with code optimization, multi-node scaling, and making the most of your allocation.