Hardware Resources
Ofakim HPC Cluster – Hardware Specifications
Ofakim is a high-performance Kubernetes cluster comprising 20 physical compute nodes, 5,915 CPU cores, and 48 enterprise NVIDIA GPUs managed dynamically by Run:ai for on-demand scientific and AI compute workloads.
Hardware Resource Breakdown
| Node Category | Count | Cores/Node | Total Cores | GPUs/Node | Total GPUs | GPU Model | VRAM / GPU | CUDA Cores | Tensor Cores | RAM / Node |
|---|---|---|---|---|---|---|---|---|---|---|
| GPU Node | 2 | 256 | 511 | 8 | 16 | NVIDIA A100-SXM4-80GB | 80 GB | 6,912 | 432 | ~2 TB |
| GPU Node | 1 | 128 | 128 | 8 | 8 | NVIDIA RTX 6000 Ada | 48 GB | 18,176 | 568 | ~1 TB |
| GPU Node | 2 | 96–192 | 288 | 8 | 16 | NVIDIA L40S | 48 GB | 18,176 | 568 | ~1 TB |
| GPU Node | 1 | 192 | 192 | 4 | 4 | NVIDIA L40S | 48 GB | 18,176 | 568 | ~1 TB |
| GPU Node | 1 | 80 | 80 | 4 | 4 | NVIDIA A100-PCIE-40GB | 40 GB | 6,912 | 432 | ~1 TB |
| CPU Node | 4 | 512 | 2,048 | — | — | — | — | — | — | ~1.5 TB |
| CPU Node | 4 | 384 | 1,536 | — | — | — | — | — | — | ~2.25 TB |
| CPU Node | 1 | 384 | 384 | — | — | — | — | — | — | ~1 TB |
| CPU Node | 2 | 256 | 512 | — | — | — | — | — | — | ~1 TB |
| CPU Node | 2 | 112 | 224 | — | — | — | — | — | — | ~1 TB |
Cluster Totals: 20 Compute Nodes • 5,915 CPU Cores • 48 GPUs • 2,368 GB Aggregate VRAM
