Intel Offers New Cable Design for HPC

The chip maker will also provide design specifications for HPC clusters created with the Intel Architecture.

Intel, a company better known for its microprocessors, will begin offering a new connector cable for the high-performance computing market as the company looks to expand its offerings within the x86 cluster and supercomputer space.

At the International Supercomputing Conference in Dresden, Germany, June 27, executives with the Santa Clara, Calif., company will detail Intel Connects Cables, giving companies more options when building large HPC clusters or supercomputers.

These cables, according to Intel, are lighter, smaller and have less of a bend radius than the types of traditional copper cables used with high-performance clusters and supercomputers.

In terms of bandwidth, the Intel cables will allow users of either InfiniBand or 10 Gigabit Ethernet interconnect technology to produces data rates of 20G bps. The cables will also allow customers to extend the reach between individual servers in a cluster to as much as 100 meters.

By offering cables that are about 83 percent smaller and 84 percent lighter than traditional copper cables—these numbers were based on Intels own internal testing—the Connects Cables will cut down on overloaded floors and racks found in the data centers that house these massive clusters. This will increase the airflow along the floor, which will cut down on cooling costs, according to Intel.

At this years ISC, Intel will show off several systems that are already using the Connects Cables technology, including HPC products and supercomputer offerings from Bull, Ciara, Cisco, Dell, Hewlett-Packard, IBM, Mellanox, Microsoft, QLogic, Silicon Graphics, Sun Microsystems, Supermicro, Tektronix, Tyan and Voltaire.

The cables will be generally available starting in the second half of 2007, and Intel said the technology will be distributed through Bell Micro and Synnex.

The HPC and supercomputer field is continuing to grow as more and more OEMs are offering larger systems with even more performance. At this years show, IBM and Sun are detailing supercomputers that break the so-called "petaflop" barrier, or 1 quadrillion—1 thousand trillion—calculations per second.

/zimages/3/28571.gifClick here to read more about IBM and Sun breaking the "petaflop" barrier in supercomputing.

When the Top 500 Supercomputers are announced June 27 in Germany, IBM will once again hold the top spot with its Blue Gene/L system that is installed at the Department of Energys Lawrence Livermore National Laboratory, in Livermore, Calif. The systems peak performance is 280.6 teraflops, or 280.6 trillion calculations per second. IBMs petaflop system, Blue Gene/P, will eventually replace the companys L system.

At the same time, more and more systems are being built using x86 architecture, which means that Intel and its rival Advanced Micro Devices will begin to play an even more prominent role in the HPC market.

Charles King, an analyst with Pund-IT Research, said that a look at the Top 500 Supercomputer list shows that between 350 and 400 of the worlds fastest computers use x86 processors from either Intel or AMD. With the potential for growth in this area, Intel and other vendors are looking for additional areas to expand their technological offerings.

The announcement that Intel has started to develop cables for x86 clusters and supercomputers, King said, should come as no surprise.

"Its a good market right now, and it will only continue to grow and get larger and larger over time," King said. "A lot of these vendors first try to figure out the cost of developing the technology to enter the market, and if that pays out favorably, they start to move forward into other areas. The evolution of x86 clusters and supercomputing is increasingly opening up, and the market is getting more affordable for customers."

/zimages/3/28571.gifHP and Microsoft are looking to bring Windows to the HPC market. Click here to read more.

With that in mind, Intel is also launching a new HPC validation program called Intel Cluster Ready, which is designed for HPC customers as well as ISVs to help validate applications that are designed for HPC systems that are based on Intel Architecture.

The Cluster Ready program also sets standards for software and hardware that can be used with Intel technology in creating high-performance clusters.

/zimages/3/28571.gifCheck out eWEEK.coms for the latest news, views and analysis on servers, switches and networking protocols for the enterprise and small businesses.