Computational Resources

The Computational Physics Group maintains several local servers and access to supercomputers and quantum computers around the country. This page gives a brief overview of the computing resources we investigate and marshal for our work.

In house

Group-owned

COC/PACE ICE

GPU nodes

  • 10x, Dual Xeon Gold 6226, 4x NVIDIA Quadro Pro RTX6000 24GB
  • 11x, Dual Xeon Gold 6226, 2x NVIDIA V100 16GB
  • 4x, Dual Xeon Gold 6248, 1x NVIDIA V100 32GB
  • 4x, Xeon Gold 6248, 4x NVIDIA V100 32GB
  • 2x, Dual AMD EPYC 7452, 2x NVIDIA A40 48GB
  • 2x, Dual AMD EPYC 7513, 2x NVIDIA A100 40GB
  • 2x, Dual AMD EPYC 7452, 2x NVIDIA A100 80GB
  • 3x, “Newell”, Power9 O2CY417, 2x NVIDIA V100 32GB

CPU nodes

  • 60x, Dual Xeon Gold 6226
  • 4x, Dual AMD EPYC 7513

Find more information here.

The Rogues

We work with CRNCH, the Center for Research into Novel Computing Hierarchies. CRNCH maintains the Rogues Gallery, a collection of (mostly) non-von Neumann systems. Here, we work with

  • ARM servers (NVIDIA ARM HPC Dev Kits)
  • FPGAs (Xilinx Alveo U50, U250, U280, AC-510)
  • RISC-V boards (MangoPi)
  • The first RISC-V GPGPU (called Vortex)
  • Neuromorphic nodes
  • Near-memory servers (Pathfinder),
  • Reconfigurable networked machines, including
    • FPGA-based network devices
    • CPU-based smart network devices
    • 5G networking testbed equipment

Check out the links above to learn more about this ever-evolving list of systems.

Phoenix

We maintain allocations on Georgia Tech’s Phoenix, a Top500 supercomputer managed by PACE. Phoenix houses a variety of CPU and GPU nodes. You can see the available resources here, though they are not always up-to-date. In December 2022, Phoenix housed just over 1000 CPU nodes and 100 GPU nodes.

Outside resources

We maintain allocations on clusters and leadership class supercomputers around the country. This includes leadership-class DOE systems like

We use university supercomputers like

Quantum computers

We also work on quantum algorithms for solving scientific problems, often based on continuum physics. For this, we use quantum simulation (on the classical hardware above) and quantum computers from