Since 1986 - Covering the Fastest Computers in the World and the People Who Run Them

Language Flags
May 30, 2012

Entry-Level HPC: Proven at a Petaflop, Affordably Priced!

Nicole Hemsoth

Computing needs at many commercial enterprises, research universities, and government labs continue to grow as more complex problems are explored using ever-more sophisticated modeling and analysis programs.

A new class of Cray XE6 and Cray XK6 high performance computing (HPC) systems, based on AMD Opteron™ processors, now offer teraFLOPS of processing power, reliability, utilization rates, and other advantages of high-end supercomputers, but with a great  low purchase price. Entry-level supercomputing systems in this model line target midrange HPC applications, have an expected performance in the 6.5 teraflop to 200 teraFLOPS range, and scale in price from $200,000 to $3 million.

These systems can give organizations an alternative to high-end HPC clusters. One potential advantage of these entry-level systems is that they are designed to deliver supercomputing reliability and sustained performance. Users can be confident their jobs will run to completion. And the systems also offer predictability. “There is reduced OS noise, so you get similar run times every time,” said Margaret Williams, senior vice president of HPC Systems at Cray Inc.

These attributes are essential today in competitive industries and research fields where results and progress are based on high-throughput computational workflows. Typically, the output from one job, such as a modeling program or a simulation, is used by additional programs queued up to run in a particular sequence. Any disruption of the workflow can delay decisions on next steps to take. While a small delay can seem minor in the scheme of things, in competitive fields it can make a significant difference.

The new emphasis on sustained performance and computational workflow throughput is beginning to change the way some organizations evaluate solutions. In the past, raw benchmarks, such as the LINPACK or LINPACK per-watt benchmarks would be used to gauge a system’s potential. Now, the attention appears to be moving more towards overall productivity.

“What’s really important is how much real work gets done, not how fast each compute [node] runs,” said Williams. “It’s the science done per-watt that matters.” So rather than looking at how fast a single job can run the more important question is: ‘how many jobs can a system run in a given time?’

For an organization, the reliability and sustainability attributes can offer other advantages. With clusters, additional nodes are often added to a system to be used as spares when one node fails to meet workload spikes. With predictable performance systems, less extra capacity is needed.  Fewer nodes means low operating costs: for example, less electricity should be needed, there are fewer elements to manage, and less rack and floor space is required.

Furthermore, entry-level supercomputers offer organizations a complete system that includes the compute engine, high-performance memory and I/O, a file system and storage. All of the elements are tightly integrated and managed as a single system. Compared to a typical HPC cluster, this can help reduce system administration costs since each element does not have to be managed independently using different management consoles and tools.

And for organizations with higher-end Cray supercomputers, the entry-level Cray systems can be used to help develop and test new programs. This is possible because the entry-level systems use the same hardware and software as their higher end counterparts. The potential benefit here is that development and testing can be kept off of production systems, thus enabling a saving of time on those systems for real work.

Directly compatible with the high-end machines, these entry-level configurations not only help minimize total system costs, but also support Cray’s Cluster Compatibility Mode (CCM), providing users with out-of-the-box installation and running of diverse Independent Software Vendor (ISV) applications across numerous technology segments.

Cray and AMD as Your Technology Partners

Cray and AMD have a long history of collaboration. Over the years, the relationship has produced some of the world’s most productive supercomputers for scientific and commercial research. In fact, in the November 2011 release of the Top 500 supercomputers, three of the top ten and seven of the top twenty supercomputers in the world were Cray systems powered by AMD Opteron™ processors.

Throughout the relationship, AMD has made several major technological leaps in processor architecture and design. Processors have gone from dual-core to quad-core to six-core over the last several years. The recent launch of the AMD Opteron™ 6200 Series processor, which had gone by the code name “Interlagos,” offers the world’s first 16-core x86 processor. The processor’s architecture is very flexible and can be applied effectively to a variety of workloads and problems. Additionally, it supports AVX instructions and offers a performance boost provided by AMD Turbo CORE technology.

For existing Cray customers, the entry-level supercomputers powered by AMD Opteron processors provide investment protection. Over the years, Cray users have been able to upgrade as new technologies become available. This included upgrades to processors, blades, and storage, which allows an organization to leverage an initial investment in a Cray system and scale the system over time.

The entry-level HPC systems are just the latest result in this partnership between Cray and AMD.

The systems offer an affordable price, requisite performance, and the power efficiency needed in organizations that find they must apply more computing resources towards solving increasingly complex problems.

To learn more about the Cray entry-level supercomputers powered by AMD Opteron processors, visit: www.cray.com/ownacray