New clusters in MetaCentrum

New clusters in MetaCentrum

Dear users,

Masaryk University (CERIT-SC) has become a pioneer in the field of artificial Intelligence (AI) and powerful computing technology by installing latest and most advanced NVIDIA DGX H100 system. This is the first facility of its kind in the entire country that delivers extreme computing power and innovative research capabilities.

Thanks to the latest NVIDIA Hopper DGX H100 GPU architecture, it features eight advanced NVIDIA H100 Tensor Core GPUs, each with a GPU 80GB of memory with a total computing power of 32 TeraFLOPS. This enables parallel processing of huge data volumes and significantly accelerates computing tasks. Thanks to the high-performance memory subsystems in the graphics  accelerators, it provides fast data access and optimizes performance when working with large data sets. Users can achieve unparalleled efficiency and responsiveness in their AI tasks.

The DGX H100 server comes with a pre-installed software package NVIDIA DGX, which includes a comprehensive set of software tools for deep learning tools, including pre-configured environments.

The machine is available on-demand in a dedicated queue at gpu_dgx@meta-pbs.metacentrum.cz.
To request access, contact meta@cesnet.cz. In your request, describe the reasons for allocating this resource (need and ability to use it effectively). At the same time, briefly describe the expected results, the expected volume of resources and the time scale of the approach needed.

 

 

NVIDIA DGX H100 configuration (capy.cerit-sc.cz)

GPUs:

8× NVIDIA H100 SXM5 80 GB

GPU memory

640 GB total

CPU

Dual 56-core 4th Gen Intel Xeon

Scalable CPU

Výkon (FP8 tensor operace)

32 TeraFLOPS

# CUDA jader

135 168

# Tensor jader

4 224

Multi-instantce GPU

56 instancí

RAM

2 TB

HDD

OS: 2× 1.92 TB NVMe

data: 30 TB (8× 3.84 TB) NVMe

Network

8x single-port ConnectX-7 VPI 400 Gb/s InfiniBand/ 200Gb/s Ethernet

2x dual-port ConnectX-7 VPI 400 Gb/s InfiniBand/ 200Gb/s Ethernet

Max. spotřeba

~10.2kW max

 

 

 Kompletní seznam aktuálně dostupných výpočetních serverů je na http://metavo.metacentrum.cz/pbsmon2/hardware.


S přáním příjemného počítání,

MetaCentrum

 

 


Ivana Křenková, Thu Jun 01 23:40:00 CEST 2023