Frequently Asked Question

This Aziz Supercomuter boasts an architecture of 496 nodes and a total of 11,904 cores, designed for efficient processing of both large parallel jobs and numerous smaller tasks. It includes 380 standard compute nodes with 96 GB of memory each, ideal for a variety of mixed HPC applications, and 112 high-memory compute nodes offering 256 GB of memory for memory-intensive workloads. Additionally, the system features 20 GPGPU nodes equipped with NVIDIA A100 GPUs, optimized for AI, machine learning, and deep learning tasks, alongside 2 NVIDIA Tesla K20 GPGPU nodes for GPU-accelerated workloads. To further enhance its capabilities, 2 Intel Phi 5110P co-processor nodes are included, supporting applications that utilize MIC-based accelerators. This diverse computational infrastructure positions the system as a robust solution for addressing complex scientific and engineering challenges.

Thin and Fat Compute Node:

This system comprises a total of 496 nodes, offering 11,904 cores to efficiently handle both large parallel jobs and a substantial number of smaller parallel or serial jobs.

• Standard Compute Nodes:
  • Quantity: 380 nodes
  • Total Cores: 9,120 cores
  • Memory: 96 GB per node (4 GB per core)
  • Use Case: Ideal for the majority of mixed HPC applications.
• High Memory Compute Nodes:
  • Quantity: 112 nodes
  • Total Cores: 2,688 cores
  • Memory: 256 GB per node (10.6 GB per core)
  • Use Case: Designed for applications that require significant memory for execution.

A100 GPU Node:

• Quantity: 20 GPGPU nodes
• GPU Model: NVIDIA A100 with 40 GB memory
• CPU Configuration: 32 CPU cores
• Memory: 512 GB
• AI,ML,Deep Learning application.
  

K20 GPU Node:

• Quantity: 2 NVIDIA Tesla K20 GPGPU-equipped compute nodes
• CPU Configuration: 24 cores
• CUDA Cores: 2,496 CUDA cores
• These nodes are designed for running applications that leverage GPU-based acceleration.

MIC Node:

• Quantity: 2 Intel Phi 5110P co-processor-equipped compute nodes
• CPU Configuration: 24 cores
• MIC Cores: 60 MIC cores
• These nodes are optimized for running applications that utilize MIC-based accelerators.