• <Intel.com 還有更多

Intel® Xeon Phi™ Product Family

Highly parallel processing to power your breakthrough innovations

Intel® Xeon Phi™ Coprocessor

Power your breakthrough innovations with the highly parallel processing of the Intel® Xeon Phi™ coprocessor. We have packed over a teraFLOPS of double-precision peak performance into every chip.


The weather vertical needs an tremendous amount of compute capabilities to accurately predict weather all over the world, including the formation of tornados, the direction of a hurricane, and rain fall predictions. Weather Research and Forecasting (WRF) is one of the more frequently used weather simulation packages.

Intel Measured as of May 2014


WRF v3.5.1:

Hardware: Two-socket server with Intel® Xeon® processor E5-2697 v2 (12C, 2.7 GHz, 130W), each node equipped with one Intel® Xeon Phi™ coprocessor (7120 c0, 61 core, 1.238 GHz, 16 GB at 5.5 GT/s) in a four-node Fourteen Data Rate (FDR) cluster.

WRF is available from the US National Center for Atmospheric Research in Boulder, Colorado; it is available from http://www.wrf-model.org/.

Intel® Manycore Platform Software Stack (Intel® MPSS) 6720-16, Compiler rev 192, Intel® MPI Library rev 3.6

WRF CONUS2.5km workload available from www.mmm.ucar.edu/wrf/WG2/bench/.

Performance comparison is based upon average timestep. We ignore initialization and post simulation file operations.

Conus2.5km run ATS KNC "Profit" % 4x12x2-20+0x0x0-0 2.224642 4x12x2-20+4x8x30-30 1.425497 56.06079844

Modifications: -DINTEL_ALIGN64 - removed from AVX version; -fp-model fast=1 -> fast=2         


1    2   3   4   5   6   7    8

For the equation above, 1 corresponds to the first number in the equation, 2 corresponds to the second, etc. Below is the legend for what each number represents.

1 number of nodes

2 number of ranks on the host

3 number of omp threads/rank

4 number of tiles

5 number of knc cards

6 number of mpi ranks/card

7 number of omp threads/rank

8 number of tiles

Source:  Intel Internal Testing TR2054

Additional information: 1 2 3 4 5



1. Software and workloads used in performance tests may have been optimized for performance only on Intel® microprocessors. Performance tests, such as SYSmark and MobileMark, are measured using specific computer systems, components, software, operations, and functions. Any change to any of those factors may cause the results to vary. You should consult other information and performance tests to assist you in fully evaluating your contemplated purchases, including the performance of that product when combined with other products. For more information, go to www.intel.com/performance.

2. 對於本文件提及的各項第三方效能標竿或網站,Intel 並不會控制或稽核其設計或執行。Intel 鼓勵所有的客戶參閱提及的網站,或是其他發佈類似評量效能標竿報告的網站,確認提及的效能標竿是否正確,並且反映出可購買系統的效能表現。

3. Intel 處理器編號並不代表效能。處理器編號乃是用來區別屬於同一處理器系列中的不同特色,並非用來區別不同的處理器系列。如需詳細說明,請參閱 www.intel.com/content/www/us/en/processors/processor-numbers.html。

4. 對於不是 Intel® 微處理器特有的最佳化,用於非 Intel 微處理器時,Intel 的編譯器可能會,也可能不會最佳化到同樣程度。這些最佳化包括 SSE2 與 SSE3 指令集,也包括其他最佳化。對於任何最佳化,用於並非由 Intel 製造的微處理器,Intel 不保證最佳化的可用性、功能性或效力。 在本產品中取決於微處理器的最佳化,乃是為了用於 Intel 微處理器而設計。某些最佳化並非專門針對 Intel® 微架構,而是保留給 Intel 微處理器。關於本公告所涉及的具體指令集,如需詳細資訊,請參閱適用產品的使用指南與參考指南。 公告修訂版編號 20110804

5. 不同的硬體架構,可能需要不同的原始程式碼。其結果乃是根據 Intel 盡最大努力,使用經過最佳化,可在所有架構執行,並且執行相同工作的程式碼。日後的各種程式碼最佳化,可能會導致不同的結果。 在本產品中,取決於微處理器的最佳化,乃是為了用於 Intel® 微處理器而設計。某些最佳化並非專門針對 Intel® 微架構,而是保留給 Intel 微處理器。關於本公告所涉及的具體指令集,如需詳細資訊,請參閱適用產品的使用指南與參考指南。 公告修訂版編號 20110804