Computing Hardware

The Pitt Center for Research Computing provides different types of hardware for various advanced computing needs.

The CRC's primary distinction between hardware is called a "cluster". A cluster is a set of computers that implement some specialized High Performance Computing (HPC) trait like the parallelism of graphics processing units, or message passing computing architectures. The HPC clusters available to CRC users are:

  • Message Passing Interface (MPI) for highly parallel computing across many machines.
  • High Throughput Computing (HTC) for processing data in large quantities or for long periods of time.
  • Shared Memory Processing (SMP) for efficient exchange and access to data with a common memory space.
  • Graphics Processing Units (GPU) for accelerated computing with GPU applications.
  • Visualization and Interactive Desktop (VIZ) for projects requiring a graphical user interface.

These clusters can be further split into "partitions" of machines with similar hardware specifications (processors, memory, etc.). The individual machines that make up the clusters and that users can submit their jobs to are called "compute nodes".

Below, you will find the hardware specifications for each cluster and the partitions that compose it.   

 

MPI Cluster

The MPI nodes are for tightly-coupled codes that are parallelized using the Message Passing Interface (MPI) and benefit from low-latency communication through an Infiniband (HDR200) or Omni-Path (OPA) network. Your job must request a minimum of 2 nodes.

Partition Architecture Nodes Cores/Node Mem/Node Mem/Core Scratch Network
mpi Intel Xeon Gold 6342 (Ice Lake) 136 48 512 GB 10.6 GB 1.6 TB NVMe HDR200; 10GbE
opa-high-mem Intel Xeon Gold 6132 (Skylake) 36 28 192 GB 6.8 GB 500 TB SSD OPA

 

HTC Cluster

These nodes are designed for High Throughput Computing workflows such as gene sequence analysis, neuroimaging data processing, and other data-intensive analytics.

Partition Architecture Nodes Cores/Node Mem/Node Mem/Core Scratch Network
htc Intel Xeon Platinum 8352Y (Ice Lake) 18 64 512 GB 8 GB 2 TB NVMe 10GbE
Intel Xeon Platinum 8352Y (Ice Lake) 4 64 1 TB 16 GB 2 TB NVMe 10GbE
Intel Xeon Gold 6248R (Cascade Lake) 8 48 768 GB 16 GB 960 GB SSD 10GbE

 

SMP Cluster

Nodes that allow for shared memory processing. SMP nodes are appropriate for programs that are parallelized using the shared memory framework. They are also appropriate for those who want to move up to supercomputers while keeping the programming style of their laptops, such as running MATLAB.

 

Partition Architecture Nodes Cores/Node Mem/Node Mem/Core Scratch Network
smp AMD EPYC 7302 (Rome) 58 32 256 GB 8 GB 1 TB SSD 10GbE
Intel Xeon Gold 6126 (Skylake) 132 24 192 GB 8 GB 500 TB SSD 10GbE
high-mem Intel Xeon Platinum 8352Y (Ice Lake) 8 64 1 TB 16 GB 10 TB NVMe 10GbE
Intel Xeon Platinum 8352Y (Ice Lake) 2 64 2 TB 32 GB 10 TB NVMe 10GbE
AMD EPYC 7351 (Naples) 1 32 1 TB 32 GB 1 TB NVMe 10GbE
Intel Xeon E7-8870v4 (Broadwell) 4 80 3 TB 38 GB 5 TB SSD 10GbE
Intel Xeon E5-2643v4 (Broadwell) 24 12 256 GB 21 GB 1 TB SSD 10GbE
Intel Xeon E5-2643v4 (Broadwell) 2 12 256 GB 21 GB 3 TB SSD 10GbE
Intel Xeon E5-2643v4 (Broadwell) 2 12 512 GB 42 GB 3 TB SSD 10GbE
Intel Xeon E5-2643v4 (Broadwell) 1 12 256 GB 21 GB 6 TB SSD 10GbE

 

GPU Cluster

Nodes that enable accelerated computing using Graphics Processing Units. GPU nodes are targeted for applications specifically written to take advantage of the inherent parallelism in general purpose graphics processing unit architectures. For small problems, any of the GPUs below will suffice.

Partition: a100. This is the default partition in the gpu cluster and is comprised of the following hardware below. To request a particular Feature (such as an Intel host CPU), add the the following directive to your job script: #SBATCH --constraint=intel

Partition: a100_multi. This partition supports multi-node GPU workflows. Your job must request a minimum of 2 nodes and 8 GPUs.

Partition: a100_nvlink. This partition supports multi-GPU computation through 8-way A100s that are tightly coupled through an NVLink switch. The details of our Nvidia HGX platform are described below. To request a particular Feature (such as A100 with 80GB GPU memory), add the the following directive to your job script: #SBATCH --constraint=80g

Partition: gtx1080. 9 nodes with dual socket Intel Xeon Silver 4112 (Skylake, 4C, 2.60GHz base, up to 3.00GHz max boost)

Partition: v100. A single nodes with dual socket Intel Xeon Gold 6126  (Skylake, 12C, 2.60GHz base, up to 3.70GHz max boost)

Partition: power9. 4 nodes of IBM Power System AC922: dual-socket Power9 (16C, 2.7GHz base, 3.3GHz turbo). Code must be compile for the Power9 platform in order to work.

 

Partition Nodes GPU Type GPU/Node --constraint Host Architecture Core/Node Max Core/GPU Mem/Node Mem/Core Scratch Network
a100 10 A100 40GB PCIe 4 amd,40g AMD EPYC 7742 (Rome) 64 16 512 GB 8 GB 2 TB NVMe HDR200; 25GbE
2 A100 40GB PCIe 4 intel,40g Intel Xeon Gold 5220R (Cascade Lake) 48 12 384 GB 8 GB 1 TB NVMe 10GbE
a100_multi 10 A100 40GB PCIe 4 amd,40g AMD EPYC 7742 (Rome) 64 16 512 GB 8 GB 2 TB NVMe HDR200; 25GbE
a100_nvlink 2 A100 80GB SXM 8 amd,80g AMD EPYC 7742 (Rome) 128 16 1 TB 8 GB 2 TB NVMe HDR200; 25GbE
3 A100 40GB SXM 8 amd,40g AMD EPYC 7742 (Rome) 128 16 1 TB 8 GB 12 TB NVMe HDR200; 25GbE
gtx1080 9 GTX 1080 Ti 11GB 4   Intel Xeon Silver 4112 (Skylake) 8 2 96 GB 12 GB 480 GB SSD 10GbE
v100 1 V100 32GB PCIe 4   Intel Xeon Gold 6126 (Skylake) 24 6 192 GB 8 GB 6 TB HDD OPA; 10GbE
power9 4 V100 32GB SXM 4   IBM Power System AC922 128 threads 16 512 GB 4 GB 1 TB SSD HDR100; 10GbE

 

VIZ Nodes

Nodes equipped with graphical user interface (GUI), especially for visualization projects. (GUI Interface)

hostname GPU Type # GPUs Host Architecture Cores Mem Mem/Core Scratch Network
viz-n0 GTX 1080 8GB 2 Intel Xeon E5-2680v4 (Broadwell) 28 256 GB 9.1 GB 1.6 TB SSD 10GbE
viz-n1 RTX 2080 Ti 11GB 2 Intel Xeon Gold 6226 (Cascade Lake) 24 192 GB 8 GB 1.9 TB SSD 10GbE