Intel tops high-performance computing list with Xeon Phi

Intel has jumped to the top of the Top 500 SuperComputer list, with its Xeon Phi processors powering the Milky Way-2 high-performance computer in China.

Located at the National University of Defense Technology in Changsha, the supercomputer registered performance levels of 33.86 petaflop/s on the Linpack benchmark. That is twice the throughput of the previous number one machine, the Cray XK7 system, known as Titan.

The Milky Way-2 is made up of 32,000 Intel Xeon processors together with 48,000 Intel Xeon Phi co-processors, the latter of which were introduced six months ago.

Intel said the machine is expected to be used to process data relating to life sciences, geophysical, weather and big data analytics.

Rajeeb Hazra, VP of the Intel architecture group, claims that high-performance computing is now a necessity and businesses that want to "compete must compute".

During a pre-brief with journalists, Hazra announced that Intel would also be adding to the Xeon Phi lineup.

"We will have a lineup of products to meet a spectrum of end user needs from the [7100 family] high performance, biggest memory footprint co-processors to the 3100 family [offering] performance and value," he said.

"We believe this will be an accelerator of HPC going forward."

Meanwhile, Hazra also drip fed information about the second generation Intel Xeon Phi range. Codenamed Knights Landing, Intel will manufacture the chipsets using the 14nm process.

"[Knights Landing] is no longer a PCIe co-processor. It is also a standalone CPU. Why is this important? Offload acceleration today is a burden for programmers. They simply want the ability to take what they do on a single CPU and have much more pervasive threading, parallism and vectorisation going on. And that is what we will be [offering]," he added.