====== Cluster Hardware Overview ======

This page provides an overview of the hardware and partitions available on the cluster. The cluster is equipped with modern Intel CPUs and NVIDIA GPUs to support a wide range of high-performance and GPU-accelerated workloads.

===== Partitions =====

The cluster is divided into several partitions, each with different CPU and GPU resources. All partitions currently have **no time limit** for running jobs.

^ Partition ^ CPU Cores (Threads) ^ GPU(s)                  ^ Notes ^
| **unite**   | Up to 192 (384)         | 5 x NVIDIA Tesla T4      | General-purpose, high-throughput jobs |
| **gpu_nm**  | Up to 32 (64)           | 1 x NVIDIA Tesla T4      | GPU-accelerated jobs, smaller scale  |
| **gpu_ai**  | Up to 192 (384)         | 5 x NVIDIA Tesla T4      | AI/ML workloads,  high-performance     |
| **a40**     | Up to 64 (128)          | 2 x NVIDIA A40            | AI/visualization,  GPU-intensive   |

===== CPU Specifications =====

The cluster features several generations of high-performance Intel Xeon processors.

  * **8 x Intel(R) Xeon(R) Gold 6148 @ 2.40GHz**
    - 20 cores / 40 threads per CPU
    - Suitable for parallel and memory-intensive tasks

  * **2 x Intel(R) Xeon(R) Gold 6142 @ 2.60GHz**
    - 16 cores / 32 threads per CPU
    - Higher base frequency for faster single-thread performance

  * **2 x Intel(R) Xeon(R) Platinum 8358 @ 2.60GHz**
    - 32 cores / 64 threads per CPU
    - High core density for large parallel workloads

===== GPU Specifications =====

The cluster supports GPU-accelerated workloads using NVIDIA Tesla and A40 GPUs.

  * **5 x NVIDIA Tesla T4**
    - 16 GB GDDR6 memory
    - Optimized for inference and lightweight training
    - Supported in `unite`, `gpu_nm`, and `gpu_ai` partitions

  * **2 x NVIDIA A40**
    - 48 GB GDDR6 memory each
    - Excellent for large model training, rendering, and visualization
    - Available in the `a40` partition

===== Summary =====

This cluster provides a flexible and powerful environment for both CPU and GPU workloads. The modular partition setup allows users to select resources best suited for their tasks, whether that’s compute-heavy simulations, AI training, or smaller-scale GPU processing.