Gráfica AMD CDNA GPU Architecture: Dedicated GPU for Data Centers

Eni boosts supercomputing infrastructure of the Green Data Center​


The architecture of HPC6 is based on the same technology that powers the leading systems of its kind globally. The HPC6 system and its storage facility will be provided by Hewlett Packard Enterprise (HPE), which won the tender in a competitive international process, and will use HPE Cray EX4000 systems and HPE Cray ClusterStor E1000 technologies for the HPC6 system and storage, respectively.
The key features of HPC6 are detailed below:
  1. Computing power: the system can reach an impressive peak computing power of over 600 PFlop/s (Rpeak) and 400 PFlop/s “sustained” (Rmax), making it one of the most advanced supercomputing infrastructures in the world.
  2. Node Composition: each node consists of a 64-core AMD EPYC™ CPU and four high-performance AMD Instinct™ MI250X GPUs for high computational efficiency and unmatched versatility, which makes it suitable to a vast array of applications.
  3. Size: the system includes 3472 computing nodes housing a total of 13,888 GPUs, organised into 28 racks for optimal space utilisation and maximum performance.
https://www.eni.com/en-IT/media/pre...mputing-infrastructure-green-data-center.html

Acho que é o primeiro sistema fora do "tradicional HPC" - cloud e Universidades/Centros de I&D, a ENI é uma empresa petrolífera.
 
A AMD tinha dito que a MI300 seria o produto da AMD que mais rapidamente chegaria a um revenue de mil milhões de $. Agora passou a previsão do revenue das MI300, em 2024, de 2 mil milhões para 3,5 mil milhões. 3,5 mil milhões de $ é um pouco mais de metade do revenue que a AMD teve no ultimo trimestre.
Acho que dá para perceber a importância deste produto.
 
Back
Topo