Endeavour Resource Overview

The Center for Advanced Research Computing's high-performance computing condo cluster, Endeavour, comprises compute nodes that are dedicated for use by researchers that lease/subscribe to them. For information on the CARC's Condo Cluster Program (CCP), see the Condo Cluster Program pages.

For general CARC system specifications, including information on our file systems and operating system, see our System Information page.

Compute nodes

The CCP offers both an annual subscription-based pricing model and a traditional system purchase pricing model, and the Endeavour cluster includes nodes for both pricing models. Below are the specifications for each model.

Subscription model:

Model# of cores per nodeGHzMemory (GB)Price per year
Lenovo NX360162.664 GB$800
Lenovo NX360202.664 GB$1,000

We recognize that these are not the latest and greatest compute nodes currently on the market, and the low subscription price reflects this. We are currently in the process of purchasing a new batch of compute nodes for the Condo Cluster Program. The COVID-19 situation has delayed this process, but the new compute nodes will be available for subscription on the Endeavour cluster in early 2021.

More information here.

Purchase model:

CPU nodes

CPUCPU quantityMemory (GB)Estimated price
Dual AMD EPYC 7542 32-core 2.90GHz2256 GB$12K
Dual AMD EPYC 7542 32-core 2.90GHz2512 GB$16K
---
Dual Intel Xeon Gold 6248R Processor 24-core 3.0GHz2192 GB$11K
Dual Intel Xeon Gold 6248R Processor 24-core 3.0GHz2384 GB$14K

GPU nodes

CPUCPU quantityMemory (GB)GPUGPU quantityEstimated price
Dual AMD EPYC 7282 Processor 16-core 2.80GHz2256 GBRTX 50002$12.5K
---
Dual AMD EPYC 7532 Processor 32-core 2.90GHz2256 GBV100s2$33K
Dual AMD EPYC 7532 Processor 32-core 2.90GHz2256 GBA1002$40K
---
Dual Intel Xeon Gold 4216 Processor 16-core 2.1GHz2192 GBRTX 50002$11K
Quad Intel Xeon Gold 4216 Processor 16-core 2.1GHz2192 GBRTX 50004$17K
---
Dual Intel Xeon Gold 6248R Processor 24-core 3.0GHz2192 GBV100s2$31K

Example system specifications (Dell and HPE/Exxact) for each model of node can be downloaded from GitHub here. Final configurations of system specs will depend on system availability from the vendor.

More information here.

Job limits

Job limits on Endeavour are more flexible than on Discovery and can be configured on a partition basis, but we still must enforce time limits for maintainence.

Default run time if no run time is specified: 1 hour
Maximum run time: 14 days
Maximum number of jobs or job steps (running or pending): 3,000

More information on Endeavour jobs can be found in the Running Jobs user guide.

Back to top