Compute

KIAC computing facility

Mahamathi is the high-performance computing (HPC) cluster of the Kotak IISc AI-ML Centre (KIAC), designed to support cutting-edge AI/ML research and large-scale scientific computing. The KIAC cluster features a heterogeneous architecture comprising 10 compute nodes powered by Intel Xeon Platinum processors and a diverse set of NVIDIA GPUs, including A5000, A6000, ADA6000, A100, and H200.

The cluster also includes dedicated GPU nodes, and all components are interconnected via a high-speed 10G network in a FAT-tree topology with a 1:1 subscription ratio for optimal bandwidth and scalability. The system is managed through the SLURM workload manager, and users access the cluster via login nodes to submit jobs in batch mode using SLURM scripts. A parallel file system is available for use during job execution.

Users can access the system directly over the SSH terminal. Access to the cluster requires prior approval through a formal application process.

For queries and support, users may contact the helpdesk at server.kiac@iisc.ac.in.

UG computing facility

Our high-performance desktop workstations (52) are built to deliver exceptional speed and reliability for professional and technical workloads.

Powered by the Intel Core i9 processor and a durable Gigabyte motherboard, it ensures smooth multitasking and fast computation.

Graphics performance is supported by the NVIDIA GTX 1660 6GB card, making it suitable for visualisation, design, and GPU-accelerated tasks. 

The workstations run on the latest Ubuntu OS.  

This high-end server is engineered for extreme performance and scalability, powered by Intel Xeon Platinum processors offering a combined 64 cores of compute power.

The system is equipped with high-speed storage, managed through onboard Intel RAID for data protection and flexibility. It is designed for advanced AI, HPC, and visualisation workloads, with powerful NVIDIA RTX A6000 and RTX A5000 GPUs.

The cluster’s master nodes are built with AMD EPYC processors (16-core and 96-core configurations). Mahamathi runs on a Linux environment and supports a rich development stack that includes GNU and Intel compilers, CUDA and a wide range of scientific libraries such as Intel MKL, and Python-based data analysis tools.

Fully automatic installation (FAI) is integrated into the system to enable consistent and rapid OS provisioning across all nodes, enhancing the system’s maintainability and scalability. 

Our advanced 16-bay NAS storage system is designed to provide users with a secure, scalable, and high-capacity data storage solution.

With a total storage capacity of 140TB, it offers ample space for large datasets, project archives, backups, and collaborative file sharing across teams.

This server is powered by dual 4th Gen Intel Xeon Scalable processors, delivering a minimum of 64 cores and 128 threads with a base frequency of 2.8 GHz and secure enclave technologies.

RAID controller ensures data protection and flexibility. With NVIDIA A100 GPUs, this system is ideal for AI, HPC, and large-scale data processing.

This advanced server platform is powered by dual Intel Xeon Scalable 4th Gen processors, offering a minimum of 64 cores and 128 threads with a base frequency of 2.1 GHz, along with support for secure enclave technologies like Intel SGX.

With eight NVIDIA RTX 6000 Ada, the server delivers extraordinary acceleration for AI, deep learning, and high-performance computing tasks.

The DGX H200 140 GB is an engineered powerhouse server. It combines massive memory capacity and high-throughput storage to handle large datasets and intensive workflows with ease.

Built for reliability and performance, the system is ideal for deep learning, data science, and enterprise-grade compute workloads, delivering robust processing power and scalability for demanding environments.

The NVIDIA A4000 is ideal for servers that need a balance of performance, efficiency, and scalability.

The server with NVIDIA RTX A5000 GPU is built to handle intensive professional workloads across AI, 3D design, simulation, and data analytics.