Sun Blade X6275 Server Module

Sixteen dual-node Sun Blade X6275 server modules, powered by two Intel Xeon X5570 processors formed a cluster interconnected via integrated InfiniBand QDR Host Channel Adapters (HCA) and Quad Data Rate Switched Network Express Modules (QNEM) housed in a Sun Blade 6048 chassis. Since each Sun Blade X6275 server module has two full-function compute nodes inside, the total number of processing cores in the cluster was 256. The Sun cluster outpaced competing SGI Altix ICE system on two popular MCAE applications—FLUENT from ANSYS Inc. and RADIOSS from Altair.

Sixteen dual-node Sun Blade X6275 server modules, running 64-bit SUSE Linux (SLES10 SP2) OS across 256 logical CPUs, demonstrated world record performance using the largest available model of FLUENT 12.0 benchmark. Truck_111M model has around 111 million cells and models external air flow over a truck body.

ANSYS's FLUENT software solves fluid flow problems, is based on a numerical technique called computational fluid dynamics (CFD) and is heavily used in the automotive, aerospace and consumer products industries. The FLUENT 12 benchmark test suite consists of seven models that are well suited for multi-node clustered environments and representative of modern engineering CFD clusters. Vendors benchmark their systems with the principal objective of providing comparative performance information for FLUENT software that, among other things, depends on compilers, optimization, interconnect, and the performance characteristics of the hardware.

The same cluster comprised of sixteen dual-node Sun Blade X6275 server modules and running 64-bit SUSE Linux (SLES10 SP2) OS across 256 logical CPUs, demonstrated world record performance on RADIOSS benchmark.

Altair's RADIOSS is a next-generation finite element solver for structural analysis and is often used in high-speed impact simulations in automotive and aerospace industries. To demonstrate the performance of RADIOSS, Altair has come up with a suite of benchmarks that include four automotive crash models based on use real data. For the 256 rank configuration running the largest available model Taurus_v_Taurus 1.8M) that simulates frontal crash of two identical cars, Sun's cluster outperformed the SGI Altix ICE by over 30%, while on the second largest model (Taurus_Frontal 835K), Sun outperformed SGI by over 40%.

The above tables point out that Sun's cluster maintained very good average scalability efficiency as the system grew in size and relied on a high performance Lustre file system to expedite the necessary data. Deploying shared file systems in clustered environments with large data sets often results in an additional performance boost due to lowered access times to files shared by all nodes.

These stellar results highlight the performance of uniquely-designed dual-node Sun Blade X6275 server modules outfitted with most recent Intel Xeon processors and allow customers to quickly compare the performance of multiple available solutions while making their purchasing decisions.

• Press Release

• FLUENT Benchmark Results

• RADIOSS Benchmark Results

• Sun's Solution for Altair RADIOSS

The Sun Blade 6048 chassis with 12 Sun Fire X6275 blade servers, interconnected via integrated InfiniBand QDR Host Channel Adapters (HCA) and Quad Data Rate Switched Network Express Modules (QNEM) and using the Lustre file system delivered up to 20x performance improvement over traditional Gigabit Ethernet/Network File System configurations.

Reverse Time Migration (RTM) is the most popular seismic processing algorithm used in geophysical studies to produce quality images of complex substructures. It can accurately image steep dips that can not be imaged correctly with traditional Kirchhoff 3D or frequency domain algorithms.

Sun's Constellation System offers a unique platform for customers looking to reduce their seismic processing time by a factor of two, as well as significantly cut down initialization time.

• Benchmark Results

• Additional Information

The Sun Blade 6048 chassis with 48 Sun Blade X6275 server modules (768 cores) was tested in this benchmark using the NAMD molecular dynamics applications software. Interconnected via Quad Data Rate InfiniBand (QDR IB), Sun's cluster delivered the best published result for this benchmark, with up to 95 percent better performance than double data rate (DDR) IB, a scalability efficiency of nearly 80 percent[1].

Molecular dynamics simulation is important to biological and materials science research. Molecular dynamics is used to determine the low energy conformations or shapes of a molecule. These conformations are presumed to be the biologically active conformations.

NAMD is a parallel molecular dynamics code designed for high-performance simulations of large biomolecular systems. NAMD was developed by the Theoretical and Computational Biophysics Group in the Beckman Institute for Advanced Science and Technology at the University of Illinois at Urbana-Champaign. NAMD is driven by major trends in computing and structural biology and received a 2002 Gordon Bell Award.

The Sun Blade X6275 server module, designed for serious HPC workloads outshines the competition and delivers the best-ever single task floating point and integer performance.

Each Sun Blade X6275 server module has two full-function compute nodes, each with two Intel Xeon processors, up to 96 GB of high-speed memory, integrated IB QDR Host Channel Adapter (HCA) and Gigabit Ethernet, optional Sun Flash Module, and optional PCIe ExpressModule I/O. The configurability of the system ensures that each node within the blade module can most efficiently run compute-intensive, memory-intensive, communication-intensive or I/O-intensive applications, so customers can address more diverse HPC workloads.

Using just one node within the Sun Blade X6275 server module, Sun's advanced design delivers ultimate compute performance with two Intel Xeon 5500 series processors and OpenSolaris OS by setting world speed records for both integer and floating point on the industry-standard SPEC CPU 2006 benchmark.

The SPEC CPU2006 benchmark exercises a computer's processor, memory architecture, and compilers on a variety of real-world compute intensive workloads and consists of two benchmark suites. One suite measures and compares compute-intensive integer performance and the other measures and compares floating-point performance. For each of these suites, two metrics are collected. One measures how fast the computer completes a single task (the speed) and the other measures how many tasks a computer can accomplish in a certain amount of time (the throughput).

Benchmark Outcome

• The innovative and integrated design of this system, in combination with OpenSolaris 2009.06 and Sun Studio 12 Update 1 software — Sun's freely available development tool for the Solaris OS and Linux — allowed one node within the Sun Blade X6275 server module (8 cores/2 chips/16 threads) to produce the record-breaking SPECint2006 score of 37.4 and the SPECfp2006 result of 50.8.
• The compound effect of using the most advanced Intel Xeon X5500 series processors in conjunction with the latest versions of OS and compiler software, the OpenSolaris 2009.06 and the Sun Studio 12 Update 1, respectively, propelled the Sun Blade X6275 server module to the top positions on the compute-intensive SPEC CPU 2006 benchmark.
• In addition to utilizing the freely available community-based version of the OS — OpenSolaris 2009.06, the latest version of Sun Studio compiler software was used to achieve these outstanding scores. The Sun Studio 12 Update 1 software contains new features and enhancements to boost performance and simplify the creation of high-performance parallel applications for the latest multicore x86 and SPARC-based systems running on leading Linux platforms, the Solaris(TM) Operating System (OS) or OpenSolaris™.
• The SPEC CPU2006 benchmark provides a broad variety of workloads such as protein sequencing, MPEG-4 decoding, XML processing, structural mechanics and speech recognition.
• The advanced features incorporated in the community-based OpenSolaris 2009.06 OS allowed Sun's system to surpass competitive systems that were running at a higher CPU clock frequency.
• OpenSolaris 2008.11 and the Sun Studio 12 Update 1 compiler software used to produce this result are available for download free of charge at www.sun.com/download.
• The Sun Studio 12 Update 1 software has set almost a dozen industry benchmark records to date, and in conjunction with OpenSolaris 2009.06 OS was instrumental in landing these new ground-breaking SPEC CPU2006 results.

• Press Release

Designed for ultimate throughput and density, the dual-node Sun Blade X6275 server module combines the compute capacity of four Intel Xeon 5500 series processors with OpenSolaris OS to deliver better performance than any other blade, setting world records for both integer and floating point throughput.

Each Sun Blade X6275 server module has two full-function compute nodes, each with two Intel Xeon processors, up to 96 GB of high-speed memory, integrated IB QDR Host Channel Adapter (HCA) and Gigabit Ethernet, optional Sun Flash Module, and optional PCIe ExpressModule I/O. The configurability of the system ensures that the same blade module can run compute-intensive, memory-intensive, communication-intensive or I/O-intensive applications, so customers can address more diverse HPC workloads.

The SPEC CPU2006 benchmark exercises a computer's processor, memory architecture, and compilers on a variety of real-world compute intensive workloads and consists of two benchmark suites. One suite measures and compares compute-intensive integer performance and the other measures and compares floating-point performance. For each of these suites, two metrics are collected. One measures how fast the computer completes a single task and the other measures how many tasks a computer can accomplish in a certain amount of time (the throughput).

Benchmark Outcome

• Innovative and integrated design of this system, in combination with OpenSolaris and Sun Studio 12 Update 1 software — Sun's freely available development tool for the Solaris OS and Linux — allowed the Sun Blade X6275 server module (16 cores/4 chips/32 threads) to produce the record-breaking SPECint_rate2006 score of 478 and the SPECfp_rate2006 result of 355.
• The compound effect of using the most advanced Intel Xeon X5500 series processors in conjunction with OpenSolaris and the latest version compiler software — the Sun Studio 12 Update 1 — propelled the Sun Blade X6275 server module to the top blade position on SPEC's compute-intensive throughput benchmark.
• Sun utilized the freely available community-based version of the OS — OpenSolaris. Additionally, freely available Sun Studio compiler software that includes the latest features and capabilities, was instrumental in landing this groundbreaking result.
• The SPEC CPU2006 benchmark provides a broad variety of workloads such as protein sequencing, MPEG-4 decoding, XML processing, structural mechanics and speech recognition.
• OpenSolaris 2008.11 and the Sun Studio 12 Update 1 compiler software used to produce this result are available for download free of charge.

• Press Release

The dual-node Sun Blade X6275 server module, powered powered by the Intel Xeon Processor 5500 series, is the newest open network system and was designed with ultimate density and performance in mind, the attributes desired by High Performance Computing (HPC) environments.

Each Sun Blade X6275 server module has two full-function compute nodes, each with two Intel Xeon Processor 5500 series, up to 96 GB of high-speed memory, integrated IB QDR Host Channel Adapter (HCA) and Gigabit Ethernet, optional Sun Flash Module, and optional PCIe ExpressModule I/O. The configurability of the system ensures that the same blade module can run compute-intensive, memory-intensive, communication-intensive or I/O-intensive applications, enabling customers to address more diverse HPC workloads.

The SPEC OMP benchmark includes workloads often used in high-energy physics, weather modeling, computational chemistry, and mechanical design, and consists of medium- and large-problem sets that stress the computer's processor, memory, compilers and OpenMP implementation.

On a very popular medium problem set of this HPC benchmark, the Sun Blade X6275 server module delivered a new x86 world record showing off the multi-threaded scalability of the system, stability of OpenSolaris 2008.11 OS and Studio 12 Update 1 compiler's support for OpenMP.

Benchmark Outcome

• On the medium problem set of SPECompM2001 benchmark, the versatile Sun Blade X6275 server module, equipped with two Intel Xeon X5570 processors and running OpenSolaris 2008.11, posted the best x86 result.
• Sun Blade X6275 server module (8 cores/2 chips/16 OMP threads) delivered a SPECompM2001 result of 48,097, using Sun Studio 12 Update 1 software that consists of performance tools, Integrated Development Environment (IDE) and C, C++, and Fortran compilers.
• Sun's x86 systems have been dominant on High Performance Computing (HPC) compute-intensive workloads using superior compiler software that delivers the best performance by providing optimizations tailored to a specific hardware platform or microprocessor architecture.
• Sun utilized the freely available community-based version of the OS — OpenSolaris. Additionally, freely available Sun Studio compiler software that includes the latest features and capabilities, was instrumental in landing this groundbreaking result.
• The Sun Blade X6275 server module produced a SPECompMbase2001 result of 43,777, beating Cisco B200-M1 server (8 cores/2 chips/16 OMP threads) that was running Red Hat Enterprise Linux OS, using Intel® Compiler 11.0 and posted a score of 43,593.
• Moreover, compared to the system equipped with the previous generation of Intel Xeon processors, the Sun Blade X6275 server module offers 1.9x better performance on SPECompM2001 benchmark, as evident by SPECompM2001 score of 17,195 recorded on Sun Fire X2250 server (8 cores/2 chips/8 OMP threads), equipped with two Intel Xeon X5482 CPUs.

• Press Release