High Performance Computing (HPC) has long been the domain of engineers, designers, data miners and 3-D renderers. Most of these high performance boxes cluster a large number of processors, often x86 (and GPUs for the graphics-inclined), and aim the overall machine at specialized applications. It is all pretty sweet, but, unfortunately, expensive and a bit of a niche.
Fortunately, there’s a battle raging amongst HPC vendors, and as this technology advances and lowers in price, HPC will find its way in more and more IT shops.
Super Micro Computer is one such vendor, having been part of this battle for some two decades. Super Micro Computers today delivering servers, storage and networking to support HPC and green computing. The company is making its case at the Supercomputing 2013 conference in Denver where it showed off its latest wares.
The company is touting its Twin (News - Alert) architecture and a new 4U FatTwin line that has two nodes, each equipped with dual Xeon E5-2600 processors. These multicore processors each act as eight individual processors. Since there are dual processors supported, this totals 16 cores.
The system also supports Intel (News - Alert) Xeon Phi. The Phi coprocessor line is not a CPU replacement, but rather a highly efficient compliment. Designed for truly parallel applications, the Xeon Phi is really a cluster on a chip offering symmetric multiprocessing and over 50 cores.
In the whitepaper “Test-driving Intel Xeon Phi coprocessors with a basic N-body
Simulation,” the authors tested an N-body simulation against Xeon system equipped with the Xeon Phi coprocessor.
“Using common code in the C language for the host processor and for the coprocessor, we benchmark the N-body simulation. The simulation runs 2.3x to 5.4x times faster on a single Intel Xeon Phi coprocessor than on two Intel Xeon E5 series processors,” wrote authors Andrey Vladimirov and Vadim Karpusenko in January. And this was on a pre-production Xeon Phi.
The combination of a 16-core computing engine and support for Phi make these new Super Micro computers super-fast. And because they are built on commodity hardware, they don’t have the price penalties that come with older, proprietary supercomputers.
“Supermicro’s Twin architecture delivers maximum performance per watt, per dollar, per square foot for many supercomputing deployments with its unique combination of high performance compute density coupled with energy saving technology and highest reliability,” said Charles Liang, President and CEO of Supermicro. “Indeed, we have invested a great amount of engineering effort to perfect our Twin server technology and now offer an unrivaled range of server solutions optimized for practically any scale application. With our new 4U 2-node FatTwin featuring dual Xeon CPUs and six Xeon Phi coprocessors per node, science, research and engineering programs can increase and accelerate project deliverables with maximized utilization of budget, resources and space.”
Intel sees partners such as Super Micro as indicative of a relatively new approach to HPC architecture. “The industry is moving from experimentation with heterogeneous computing to more efficient neo-heterogeneity which combines the benefits of heterogeneous hardware while still using the same, common and standard programming models for both the CPU and co-processor,” said Rajeeb Hazra, vice president and general manager of Technical Computing Group at Intel. “With solution providers such as Supermicro combining the high-performance Intel Xeon processors E5-2600 v2 with Intel Xeon Phi coprocessors in high-density, scalable server solutions, industry has the ideal pairing of technology to enable a neo-heterogeneous era. With a common underlying Intel architecture we provide developers with a rapid deployment environment for their programs along with enterprise class stability and reliability for the most demanding mission critical, compute intensive applications.”
Many of today’s HPC systems are commodity supercomputers, made all the more affordable by Linux. Linux is free and scalable across clusters, multicores and multiprocessors. Windows Server has been growing as an HPC OS, and the availability of either Linux or Windows means you can apply this muscle to common data processing tasks including analyzing big data.