+65 6591 8608

Our blog

Home Our Blog Xeon Phi based HPC

Sunday24 October 2021

Xeon Phi based HPC

Introducing Xeon PhiTM coprocessors Intel is starting a new era in Supercomputing. Working closely with Intel for several months prior to Xeon Phi announcement we were happy to launch P10K Xeon Phi based Cluster and Xeon Phi servers the same day Intel announcement was made. 

Xeon Phi coprocessors are build on established CPU architecture and programming concept to provide the benefits of code re-use to developers of highly parallel applications. Common programming models for Intel Xeon processors extend to Intel Xeon Phi coprocessors, so as developers embrace high degrees of parallelism, they don’t need to rethink the entire problem.


The same techniques that deliver optimal performance on Intel Xeon processors—such as scaling applications to cores and threads, blocking data for hierarchical memory and caches, and effective use of SIMD—also apply to maximizing performance on Intel Xeon Phi coprocessors.

In other words if your code is well parallelized on multiple CPU cores, you just recompile it using Intel corresponding compiler and get an instant speedup of up to 70% per each Xeon Phi card.

As Xeon Phi is not just the add-on card but instead “a standalone computer” by itself with its own embedded Linux OS, Intel® MIC Platform Software Stack and Software Configuration Tools, a specialized skills are needed to build and support Xeon Phi based HPC systems.

The first time we put our hands on the Many Integrated Core (MIC) coprocessor was during our Intel Xeon Phi training at Intel in August 2012. Back then we were building, servicing and launching early sample codes (the only few that were available that time) on a simple Xeon Phi server based on early (alpha) but already real (not emulators) and working (pre-production quality) samples of Xeon Phi coprocessors.

MICtraining1   MICtraining2

Today, from the first minute of official announcement of Xeon Phi coprocessors by Intel, we are proud to be the first in Asia to provide tested and certified RapidCores Xeon Phi based HPC Cluster available immediately as HPC as a Service, private full-size P10K Xeon Phi based HPC clusters which can scale up to hundreds of TFlops and beyond and build with the latest 56Gb/s bandwidth interconnect and Parallel File System storage, and Xeon Phi software development tools.

The whole Xeon Phi ecosystem in one place. We happy to share it with you!

JA Minisite

2857.orig.q75.o0 - Copy IntelTPP amd  tesla preferred partner  quadro partner  qctlogo e   Mellanox APAC Partner   1