Cineca Drives Toward Exascale hpc with 250 Petaflops leonardo Supercomputer


th Gen Intel® Xeon® Scalable Processors


Download 1.43 Mb.
Pdf ko'rish
bet5/6
Sana18.06.2023
Hajmi1.43 Mb.
#1579477
1   2   3   4   5   6
Bog'liq
cineca-leonardo-case-study-032023y

4th Gen Intel® Xeon® Scalable Processors
4th Gen Intel Xeon Scalable processors in the compute 
partition integrate built-in accelerators that are optimized 
for specific workloads. They deliver increased performance 
at improved efficiency for optimal total cost of ownership.
1
These accelerators include Intel® Advanced Matrix Extensions 
(Intel® AMX) with support for BFloat16 and int8 to accelerate 
neural network computations, Intel® QuickAssist Technology 
(Intel® QAT) to accelerate cryptography and data compression 
and Intel® Advanced Vector Extensions (AVX-512) to speed up 
AI processing, among others.
Power is a key metric in supercomputing centers today. The 
4th Gen Intel Xeon Scalable processors are Intel’s most 
sustainable data center processors, with many features for 
managing power and performance, making improved use of 
CPU resources to achieve key sustainability goals. 
LISA Expansion
Additionally, Cineca has been approved for a major expansion 
called LISA. Cineca expects the expansion will increase 
Leonardo’s already formidable computing power by around 
100 petaFLOPS and broaden the use cases the system can 
support. LISA will add two new modules. The first is a module 
with conventional nodes that utilize high-bandwidth memory, 
aimed at improving the performance of tasks that require 
fast data transfers between the memory and the CPU. The 
second module for high-end acceleration will be powered by 
next-generation GPU server nodes with the goal of providing 
significant efficiency in terms of performance per watt.
Result
Leonardo will deliver advanced HPC capabilities to Cineca 
and its members, enabling new discoveries and innovation.
The pre-production phase of the Leonardo supercomputer 
has begun. The Leonardo Early Access Program (LEAP) aims 
to support projects with high scientific impact and can take 
advantage of Leonardo’s many computational resources.
Researchers across science and industry and public sectors 
can submit proposals, regardless of their nationality. 

Download 1.43 Mb.

Do'stlaringiz bilan baham:
1   2   3   4   5   6




Ma'lumotlar bazasi mualliflik huquqi bilan himoyalangan ©fayllar.org 2024
ma'muriyatiga murojaat qiling