Carnegie Mellon University


Unlocking Interactive AI Development for Rapidly Evolving Research

Read Press Release

In early summer 2020, an innovative and unprecedented NSF-funded AI supercomputer, Neocortex, was awarded for deployment at the Pittsburgh Supercomputing Center (PSC). Neocortex, which captures groundbreaking new hardware technologies, is designed to accelerate AI research in pursuit of science, discovery, and societal good.

Neocortex will be a highly innovative resource that will accelerate AI-powered scientific discovery by vastly shortening the time required for deep learning training, foster greater integration of artificial deep learning with scientific workflows, and provide revolutionary new hardware for the development of more efficient algorithms for artificial intelligence and graph analytics.

Neocortex will democratize access to game-changing compute power otherwise only available to tech giants for students, postdocs, faculty, and others, who require faster turnaround on training to analyze data and integrate AI with simulation. It will also inspire the research community to scale their AI-based research and integrate AI advances into their research workflows.

With Neocortex, users will be able to apply more accurate models and larger training data, scale model parallelism to unprecedented levels and avoid the need for expensive and time-consuming hyperparameter optimization. The development of new algorithms in machine learning and graph analytics will be enabled through this innovative AI platform.

Neocortex System Specifications

Neocortex will feature two Cerebras CS-1 systems and an HPE Superdome Flex HPC server robustly provisioned to drive the CS-1 systems simultaneously at maximum speed and support the complementary requirements of AI and HPDA workflows.

Neocortex will be federated with Bridges-2 to yield great benefits including:
  • Access to the Bridges-2 filesystem for management of persistent data
  • General-purpose computing for complementary data wrangling and preprocessing
  • High-bandwidth connectivity to other XSEDE sites, campus, labs, and clouds
The configuration of each specialized system is described below:

Cerebras CS-1

Each CS-1 features a Cerebras WSE (Wafer Scale Engine), the largest chip ever built.

AI Processor

Cerebras Wafer Scale Engine (WSE)
  • 400,000 Sparse Linear Algebra Compute (SLAC) Cores
  • 1.2 trillion transistors
  • 46,225 mm²
  • 18 GB SRAM on-chip memory
  • 9.6 PB/s memory bandwidth
  • 100 Pb/s interconnect bandwidth

System I/O

1.2 Tb/s (12 × 100 GbE ports)

HPE Superdome Flex


32 x Intel Xeon Platinum 8280L, 28 cores, 56 threads each, 2.70-4.0 GHz, 38.5 MB cache (more info).


24 TiB RAM, aggregate memory bandwidth of 4.5 TB/s

Local Disk

32 x 6.4 TB NVMe SSDs 
  •  204.6 TB aggregate
  • 150 GB/s read bandwidth

Network to CS-1 systems

24 x 100 GbE interfaces
  • 1.2 Tb/s (150 GB/s) to each Cerebras CS-1 system 
  • 2.4 Tb/s aggregate

Interconnect to Bridges-2

16 Mellanox HDR-100 InfiniBand adapters 
  • 1.6 Tb/s aggregate


Red Hat Enterprise Linux

This material is based upon work supported by the National Science Foundation under Grant Number 2005597.

Any opinions, findings, and conclusions or recommendations expressed in this material are those of the author(s) and do not necessarily reflect the views of the National Science Foundation.