Penn Advanced Research Computing Center (PARCC)

Penn Advanced Research Computing Center (PARCC) is a groundbreaking initiative that will serve as a key platform for computational and data-driven research across campus. With hardware, software, system support, and research consulting services, researchers across all schools will gain access to unparalleled support for impactful research outcomes.

PARCC SERVICES

Current available information on the equipment, technology, and services PARCC will offer:

248 B200 GPUs | 47TB HBM3e VRAM

2,232 FP8/Int8 Tensor PFLOPS
46.5 TB GPU Memory
3472 System CPU Cores

Learn More

AMD EPYC Genoa 9374F CPU

35TB Memory
4,736 Cores
18,234 GHz Frequency

Learn More

VAST Data Cluster

2.5PB NVMe Storage
500GBps Dataset Streaming
1000 GB/s Read Throughput
5.7M Read IOPs (4k)
Optimized for High Performance

Learn More

Ceph Data Cluster

2.5PB HDD StorageBalanced
Performance and Cost

Learn More

Ceph Data Recovery Cluster

2.5PB HDD
Offsite. Secure. Reliable.

Learn More

OUR TRUSTED PARTNERS

Brought to PARCC through a collaborative engineering process with:

AHEAD

HPC and Al Systems Design, Build, and Implementation

PARCC is pleased to be working with our trusted partner, AHEAD, who has been engineering digital platforms that power the most successful organizations in the world for over 30 years.

As an NVIDIA Elite Partner, AHEAD specializes in building and integrating NVIDIA DGX SuperPODs, leveraging its advanced engineering capabilities, full-stack integration services, and best-of-breed ecosystem to accelerate Al and HPC initiatives at scale.

Flexential

Data Center Design and Infrastructure Readiness

PARCC looks forward to working with Flexential to advance the experience of all those we serve. Flexential has worked for over two decades helping data-intensive organizations transform their IT infrastructure. Their approach to data center design integrates advanced technologies, operational efficiency, and deep expertise to provide clients with unparalleled performance and reliability.

Learn More