Download a PDF of this rate sheet.
Paid allocations are available for researchers who require guaranteed access to compute and storage resources. With a minimum 36 month term length, purchase 40 cores of Basic Memory, 20 cores (or equivalent) of Standard Memory, High Memory, or GPUs, and receive 5TiB of group storage, free. Want to try out an allocation before you buy? Request a free 30-day trial allocation.
Estimate your monthly SLA fees
Establishing a new SLA or considering changes to an existing account? Use the SLA Rate Calculation Worksheet to quickly generate an estimate. You can submit your worksheet to an ICDS representative if you have questions or if you’re ready to start your SLA.
Note: All prices reflect discounts made possible through substantial funding from the Provost to encourage the use of ICDS services for University-sponsored research.
Note: Storage allocations are subject to inode restrictions: 1,000,000/TiB for active and 200,000/TiB for archive. Each file you store uses an inode; if your workflow generates an extremely large number of very small files, it’s possible to run out of inodes before running out of storage space.
Roar Collab Resources
The Roar Collab cluster is not to be used with restrict (Level-3 or Level- 4) data. Researchers with restricted data should utilize the Roar cluster instead.
| Type | Description | Cost/Month |
|---|---|---|
| Basic Memory | For minimal memory jobs. • 5.33/4 GB/core • 24-64 cores and 128-256 GB RAM per node | $80 per 20 core block |
| Standard Memory | For moderate memory jobs and multicore processing • 8-10 GB/core, Infiniband • 20/48 cores and 256/512 GB RAM per node | $140 per 20 core block |
| High Memory | For large memory jobs and multicore processing • 25 GB/core, Infiniband • 40 cores and 1 TB RAM per node | $420 per 20 core block |
| Single GPU (A100) | For GPU enabled compute. • NVIDIA A100 GPU, 24 cores and 192 GB RAM, Infiniband • 2 GPU, 48 cores, and 384 GB RAM per node | $400 per card |
| Single GPU (P100) | For GPU enabled compute. • 1 NVIDIA P100 GPU, Infiniband, 24 cores and 256/512 GB RAM | $300 per card |
| Multi-Instance GPU (1.2 A100 Slice) | A 1/2 "slice" of an NVIDIA A100 card; allows for GPU enabled compute without purchasing a whole card. • ~ 3600 CUDA cores and 20 GB VRAM per slice • 12 cores and 96 GB RAM per slice | $200 per partial |
| Multi-Instance GPU (1/8th A100 Slice) | A 1/8th “slice” of an NVIDIA A100 card; allows for GPU enabled compute without purchasing a whole card. • ≈900 CUDA core and 5GB VRAM • 3 CPU cores and 24 GB RAM | $57 per partial |
| Active Storage | Storage mounted to compute resources - designed for actively used files and datasets. | $33.35 per 5TiiB |
| Archive Storage | Archival storage separate from compute resources - designed for long term storage and infrequently used files and data sets. | $6.25 per 5TiB |
Purchasers of cores will access servers with at least these specifications.
Roar Resources
The Roar cluster is maintained for researchers who utilize restricted data.
| Type | Description | Cost/Month |
|---|---|---|
| Standard Memory | For moderate memory jobs and multicore processing. • 10 GB/core, Infiniband • 20 cores and 256 GB RAM per node | $140 per 20 core block |
| Single GPU (P100) | For GPU enabled compute. • 4 NVIDIA P100 GPUs, 24 cores and 256 GB RAM, Infiniband • 1 GPU, 6 cores, and 64 GB RAM per card | $300 per card |
| Active Storage | Storage mounted to compute resources - designed for actively used files and datasets. | $33.35 per 5TiB |
Special Services
Your service level agreement may include special services such as RISE consultations, science gateways, or hosted compute. Contact us at icds@psu.edu to discuss your needs and pricing.
Payment for ICDS Services
ICDS utilizes iLab for monthly invoicing of our Service Level Agreements.
Contact ICDS to Get Started
Call or email to discuss your research goals so we can create a package that meets your needs and your budget.
icds@psu.edu 814-867-1467