Home News

Sunday16 December 2018

Intel Xeon Phi™ is in Singapore!


Get an instant boost in performance for your highly-parallel applications by simply recompiling them


Xeon Phi™ usage models benefits:

Xeon-Phi-5110P 

Multicore Hosted Model: in multicore computing, general-purpose serial and parallel computing are supported main(), and a work function foo() run on the multicore.
Offload Model: for codes with highly-parallel phases, main(), with the program’s serial code fragments, may run on the multicore, whereas a parallelizable work function foo() may execute on many-core. This is called the offload model foo() may still execute on the multicore, for cases where   the ratio of communication to computation is low, such that offload would be inefficient. The multicore cores are still the ranks for MPI.
Symmetric Model: the same program may run on either multicore or many-core. This is suitable for codes with balanced needs.
Many-core Hosted Model: at the other end of the spectrum, highly- parallel codes can run main() and foo() entirely on many-core. Each many- core may be treated as an MPI rank in that model. This execution model is sometimes referred to as native.
Reverse Offload Model: this model supports the case where there are highly-parallel codes with serial phases. The highly-parallel codes are run entirely on the many-core and the serial phases on multicore.

Order your NOVATTE M40 Phi system with Xeon Phi™ 5110P coprocessor by January 28, 2013 and get 10% discount on Intel® Parallel Studio XE 2013 or Intel® Cluster Studio XE 2013!

Subcategories

JA Minisite

2857.orig.q75.o0 - Copy IntelTPP amd    qctlogo e   Mellanox APAC Partner   1