Lonestar6
Supporting Texas researchers with a balanced set of resources to support simulation, data analysis, visualization, and machine learning
Lonestar6 is the newest system in TACC's Lonestar series of high performance computing systems that are deployed specifically to support Texas researchers. Lonestar6 is funded through the University of Texas Research Cyberinfrastructure, a collaboration between TACC and the University of Texas System, and is supported by partners at Texas A&M University, Texas Tech University, and the University of North Texas, as well as a number of research centers and faculty at UT-Austin, including the Oden Institute for Computational Engineering & Sciences and the Center for Space Research. Lonestar6 is available to researchers at all of these institutions.
System Specifications
Lonestar6 is composed of 560 compute nodes and 16 GPU nodes. The system employs Dell Servers with AMD's EPYC Milan processor, Mellanox's HDR Infiniband technology, and 8 PB of BeeGFS based storage on Dell storage hardware. Additionally, Lonestar6 supports GPU nodes utilizing NVIDIA's Ampere A100 GPUs to support machine learning workflows and other GPU-enabled applications.
The compute nodes are housed in four dielectric liquid coolant cabinets and 10 air-cooled racks. The air cooled racks also contain the 16 GPU nodes. Each compute node has two AMD EPYC 7763 64-core processors (Milan) and 256 GB of DDR4 memory. Each GPU node also contains two AMD EPYC processes and two NVIDIA A100 GPUs each with 40 GB of high bandwidth memory (HBM2).
Dielectric Liquid Cooled Compute
| Processors | 2x AMD EPYC 7763 |
|---|---|
| Cores/Node | 128 (64 per socket) |
| Clock Rate | 2.45Ghz (Base Frequency) |
| Peak Node Performance | 5TF |
| Memory/Node | 256GB DDR-4 |
| Local Disk | 144GB/tmp partition on 288GB SSD |
| Network | Mellanox Infiniband, HDR-100 |
| Storage | Mellanox Infiniband, HDR-100 |
Air Cooled GPU Racks
| Processors | 2x AMD EPYC 7763 |
|---|---|
| Cores/Node | 128 (64 per socket) |
| Clock Rate | 2.45Ghz (Base Frequency) |
| Peak Node Performance | 4.8TF, Double Precision |
| GPU | 2x NVIDIA A100 GPUs with with 40GB HBM2 |
User Guide
Access full documentation on system architecture, software, new & advanced user information, best practices, and troubleshooting.
Lonestar6 User Guide