SRE operates under a cost-recovery model for both cloud and on-premises based services that ensures cost-competitive rates, equitable access to our services, and aligns with federal agreements that govern Facilities and Administration (F&A) rates.
For assistance in determining the best option that meet the needs of your research group Contact Us
On-Premises Rates
| CPU | Per Project | Compute Usage | Storage Usage |
|---|---|---|---|
| Small | $850.00 | Up to 186,000 hours | Up to 10 TB |
| Medium | $1200.00 | Up to 620,000 hours | Up to 30 TB |
| Large | $2000.00 | Up to 930,000 hours | Up to 70 TB |
| GPU | Per Project | Compute Usage | Storage Usage |
| Small | $1000.00 | Up to 432 hours | Up to 10 TB |
| Medium | $1500.00 | Up to 1,440 hours | Up to 30 TB |
| Large | $2500.00 | Up to 2,160 hours | Up to 70 TB |
| Development Consulting | |||
| $70.00 | Per Hour |
Note – Billing rates were last updated on July 8th, 2024, and are reviewed and updated annually
Usage Charges
- Depend on the amount and type of resources requested.
- Can be paid for by grants, start-up funds, or by departmental funds.
- Require a fund code for all HPC requests.
- Does NOT include Facilities & Administrative (F&A) cost.
If unsure about your source of funding, please contact your CRU or departmental business manager. CRUs and OOs | Duke University School of Medicine
Buy Your Own Hardware (BYOH)
Principal Investigators, departments and units have the option to contribute to the HPC on-premises cluster by paying for the full cost of the hardware and then pay annual support costs. Hardware can include servers, GPUs, and/or storage.
The RCC staff will work with approved vendors to provide you options that meet your budget and requirements. Below are the standard options available as of January 2025:
| Server Type | Specs | Cost |
|---|---|---|
| CPU Only |
2 x 32 cores/64 threads 1TB RAM |
$22,280 |
| Standard Performance GPU |
2 x 32 cores/64 threads 1TB RAM 4 x NVIDIA L40S (48GB) |
$60,500 |
| High Performance GPU |
2 x 56 cores/112 threads 2TB RAM 8 x NVIDIA H200 (141GB) |
$290,000 |
| BYOH Annual Maintenance | ||
| Annual Support/Maintenance | $2,000 |
Billing and Reporting SOP
Billing for Services
Project and annual maintenance costs will be pre-paid annually to the approved fund code provided.
- Allocated size is based on estimated yearly compute hours and/or the maximum scratch space data storage that will be used on a project per year.
- Project costs are not prorated at fiscal year end and no billing adjustments will be made for unused hours.
- Extreme over usage will result in billing adjustment.
- Projects will be re-evaluated annually for any size/rate adjustment for the next 12 months of expected usage.
Usage Reporting
- Compute usage (CPU and GPU) will be measured in compute hours.
- Each project will have associated usage reports showing your current allocation and actual usage. This is to gauge your usage and will not impact the price unless you exceed the compute hours you have been allocated. If that occurs, you may be charged at the next allocation level.
- Each project is assigned a 10/30/70TB storage folder based on project size.
- Storage usage will be measured in bytes.
- Usage will be tracked by project/fund code.