Support BGU Support BGU
HPC
HPC
Ofakim1

HPC Ofakim Cluster

​​​​​
section-bottom

Back to main page >> HPC High-Performance Computing

Ofakim Cluster

Ofakim Cluster Features

  • Run:ai: Delivers intelligent orchestration that maximizes compute efficiency and dynamically scales AI training and inference by allowing fractioning of GPUs to maximize utilization.
  • GPUs: Extensive GPU resources significantly accelerate AI/ML, and other GPU-accelerated workloads, enabling researchers to tackle complex computational challenges with greater speed and efficiency.
  • High-Performance I/O: Using IBM's ESS 3500 GPFS (General Parallel File System) and InfiniBand fabric, Ofakim delivers exceptionally fast I/O performance. This ensures rapid data access and transfer, crucial for data-intensive applications and large-scale simulations.

Purpose:

Optimized for AI/ML and GPU-intensive workloads while also providing robust support for a variety of additional applications.

Specs:

Uses state-of-the-art enterprise technology designed for parallel computing:

IBM ESS 3500 parallel storage

NVIDIA ConnectX InfiniBand fabric

General Information

Site: Ben-Gurion University of the Negev
Installation Year: May 2023
Status: Operational
Manufacturer: Gigabyte, Lenovo
Cluster CPU Cores: 2130
GPU’s: 48
Coda Cores: 93,440
RAM: 22.66 TB
VRAM: 2.817 TB
Processor: AMD EPYC 9654, AMD EPYC 7763, Intel Xeon Gold 5520, etc..
System Type: Dual HPC Cluster (CPU + GPU)
Scheduler: Run:AI (Kubernetes-based)
Storage: IBM ESS 3500 (parallel file system, GPFS)
Interconnect: InfiniBand (IB)
Operating System: Oracle Linux Server 9.4

Performance

Linpack Performance (Rmax): ~321 TFLOPS
Theoretical Peak (Rpeak): ~240 TFLOPS

Power Consumption

Power: ~32.5 kW
Power Measurement Level: 1

Nodes

  • 2 Nvidia A100 80GB Nodes

o   8x NVIDIA-A100-SXM4-80GB GPUs

o   AMD EPYC 7763 64-Core Processor

o   2 TB RAM

  • 1 Nvidia A100 40GB Nodes

o   4x NVIDIA-A100-SXM4-40GB GPUs

o   Intel(R) Xeon(R) Gold 6242R CPU @ 3.10GHz

o   1 TB RAM

  • 1 Nvidia RTX 6000 Ada Generation Nodes

o   8x NVIDIA-RTX-6000-Ada-Generation

o   Intel(R) Xeon(R) Gold 6448Y

o   1 TB RAM

  • 3 Nvidia L40S Nodes

o   8x NVIDIA-L40S

o   AMD EPYC 9454 48-Core Processor

o   1 TB RAM

  • 5 AMD EPYC 9654 96 Nodes

o   AMD EPYC 9654 96-Core Processor

o   2 TB RAM

  • 2 AMD EPYC 9554 64 Nodes

o   AMD EPYC 9654 96-Core Processor

o   1 TB RAM

  • 2 INTEL(R) XEON(R) GOLD 5520+

o  Intel Xeon Gold 5520+

o   1 TB RAM

Get Started with the Cluster

Learn how to access, submit jobs, and make the most of our computing resources.
Our documentation provides step-by-step guides and best practices.

View Documentation

Contact Our Staff

Contact support: hpc_admins@bgu.ac.il

  • Ran Dahan – Head of HPC Section, +972-8-6472-128
  • Saar Skittel – DevOps Engineer, +972-8-6461-185
  • Orion Peeters – DevOps Engineer, +972-8-6461-185
  • Sean Simoni – Student DevOps Engineer, +972-8-6461-185