The 2018-2019 Buy-in Program configurations are available with upgraded processors (from Broadwell to Skylake). GPU configuration options have been expanded to offer both NVIDIA P100 (now with 16 GB memory) and the new NVIDIA V100 GPU cards, which provide more CUDA cores. Additionally, GPU offerings are now available in 1x, 2x, and 4x GPU configurations. The 1 Terabyte (TB) Memory node has been upgraded to include an AMD EPYC processor with more memory bandwidth. The VirtualGL offering is available for complex interactive visualization applications.
The following configurations are available for purchase under the 2018-2019 Buy-in Program. Configurations include 5 year next business day hardware maintenance.
-
Standard Compute Nodes
- The Value and Large Memory Nodes are the most popular nodes in the Shared Computing Cluster and provide the best platform for general purpose computing.
-
Base Node Value Node Large Memory Node - Dell PowerEdge C6420 Server
- Two Intel Xeon Gold 6132 processors
2.6GHz, 14 core, Skylake (28 total cores) - 10 Gigabit Ethernet networking
- 1TB 7.2K SATA Hard Drive
Base Node +
- 192GB memory
Base Node +
- 384GB memory
See Price See Price
HPC Compute Nodes
The High Performance Compute (HPC) nodes are identical to the “Value” and “Large Memory” nodes with the addition of EDR Infiniband interconnects. The Infiniband fabric provide high bandwidth connection between compute nodes and are intended for MPI based workloads that scale beyond the 10-Gigabit Ethernet connections provided in the standard configuration. These nodes cannot be purchased individually and interested groups must consult with Research Computing Services before purchase.
| Base HPC Node | HPC Node | HPC Large Memory Node |
|
HPC Base Node +
|
HPC Base Node +
|
| See Price | See Price |
-
GPU Nodes
- The GPU offerings have expanded and now come in 1x, 2x, and 4x GPU configurations. GPU options have also expanded to include both the NVIDIA P100, now with 16 GB memory, and now offer the new NVIDIA V100 GPU cards. Please see our GPU comparison chart for details on the upgraded GPU options. Additional consultation with RCS on GPU configuration and application testing is highly recommended.
-
Single GPU Systems
-
GPU Base Node 1xGPU P100 Node 1xGPU V100 Node - Dell PowerEdge R740 Server
- Two Intel Xeon Gold 6132 processors 2.6GHz, 14 core, Skylake (28 total cores)
- 192GB memory
- 10 Gigabit Ethernet networking
- 1TB 7.2K SATA Hard Drive
GPU Base Node + - One NVIDIA Pascal P100 16GB GPU cards
GPU Base Node + - One NVIDIA Volta V100 16 GB GPU cards
See Price See Price
-
Dual GPU Systems
-
GPU Base Node 2xGPU P100 Node 2xGPU V100 Node - Dell PowerEdge R740 Server
- Two Intel Xeon Gold 6132 processors 2.6GHz, 14 core, Skylake (28 total cores)
- 192GB memory
- 10 Gigabit Ethernet networking
- 1TB 7.2K SATA Hard Drive
GPU Base Node + - Two NVIDIA Pascal P100 16GB GPU cards
GPU Base Node + - Two NVIDIA Volta V100 16 GB GPU cards
See Price See Price
Quad GPU Systems
-
* The Quad GPU platform is limited to 240GB local disk. Consult with RCS staff for details. Quad GPU Base Node 4xGPU P100 Node
4xGPU V100 Node - Dell PowerEdge C4140 Server
- Two Intel Xeon Gold 6132 processors 2.6GHz, 14 core, Skylake (28 total cores)
- 192GB memory
- 10 Gigabit Ethernet networking
- 240GB Solid State Disk*
Quad GPU Base Node + - Four NVIDIA Pascal P100 16GB GPU cards
Quad GPU Base Node + - Four NVIDIA Volta V100 16GB GPU cards
See Price See Price
-
1 TB Memory Node
- The 1TB Memory node provides a resource for computing with very large memory requirements using the AMD EPYC processor with 1 TB of memory and increased memory bandwidth.
-
1 TB Memory Node - Dell PowerEdge R7425 Server
- Two AMD EPYC 7351 2.4 GHz, 16 core processors (32 cores total)
- 1024 GB memory
- 10 Gigabit Ethernet networking
- 1TB 7.2K SATA Hard Drive
See Price
-
VirtualGL Node
- The VirtualGL node is a scaled down system intended for complex interactive visualization applications that require OpenGL full 3D hardware accelerations. Most SCC graphical applications such as MATLAB, RStudio, and QGIS do not require this and perform well using a standard VNC session on standard node configurations. Additional consultation with RCS is required before purchasing. Since this node is priced less than $5,000, it does not qualify as capital equipment and may incur indirect costs (IDC) if purchased using grant funds. Please consult with RCS for other options concerning singe node grant funded purchases.
| VirtualGL Node |
|
| See Price |
Additional Options
Other customized options may be available for the above base systems. We are happy to discuss your research needs and specific hardware requirements.
Ready to Buy or Have More Questions?
For additional technical details and limitations of the program please refer to the Buy-in Program details page. Detailed information for the order process is available on the Buy-in Program Procedures page. If you have additional questions about the Buy-in Program or are interested in purchasing buy-in equipment, please email buyin@rcs.bu.edu.
Comparison of NVIDIA GPU Cards
| Feature Comparison | P100 (2017) | P100 (2018) | V100 |
| Peak double precision floating point performance (board) | 4.7 TFlops | 4.7 TFlops | 7 TFlops |
| Peak single precision floating point performance (board) | 9.3 TFlops | 9.3 TFlops | 14 TFlops |
| Peak half precision floating point performance (board) | 18.7 TFlops | 18.7 TFlops | 112 TFlops* |
| CUDA cores | 3584 | 3584 | 5120 |
| Memory size | 12 GB | 16 GB | 16 GB |
| Memory bandwidth (ECC off) | 549 GB/s | 732 GB/s | 900 GB/s |
* V100 measures Tensor Performance instead of half-precision operations.