ClearSpeed Technology Announces Record Performance Per Watt on Industry-Standard Benchmark; Latest Performance Results Indicate ClearSpeed's Potential to Revolutionize HPC By Addressing Power Consumption and Heat Dissipation Issues.SAN JOSE, Calif. -- ClearSpeed Technology (LSE LSE - Language Sensitive Editor :CSD CSD Commission on Sustainable Development CSD Serbian Dinar (ISO currency code) CSD Christopher Street Day CSD Circuit Switched Data (Sprint) CSD Computer Science Department CSD Community School District ), a developer of high-performance, low-power, programmable co-processor solutions, today announced its first LINPACK benchmark score for its Advance(TM) accelerator board to coincide with the Supercomputing 2005 (SC05) conference in Seattle. The score of 30.2 Gigaflops (GIGA FLoating point OPerations per Second) One billion floating point operations per second. See FLOPS. (unit) gigaflops - (GFLOPS) One thousand million (10^9) floating point operations per second. (GFLOPS See gigaFLOPS. GFLOPS - gigaflops ) was achieved in a mainstream dual processor workstation with a single ClearSpeed Advance accelerator board. The Advance board operates at a very low power dissipation of under 25 watts, weighs 9 oz. and achieved the Linpack score without program changes. The score appears at http://top500.org/lists/linpack.php, in the "Performance of Various Computers Using Standard Linear Equations Software" report maintained by Professor Jack Dongarra, University of Tennessee The University of Tennessee (UT), sometimes called the University of Tennessee at Knoxville (UT Knoxville or UTK), is the flagship institution of the statewide land-grant University of Tennessee public university system in the American state of Tennessee. . E[acute accent]By itself, the workstation is capable of achieving a LINPACK score of 8 GFLOPS and has a power consumption of over 300 watts. Adding a single 25 watt Advance board yields an almost 4x improvement in performance, increasing the performance per watt ratio for the system considerably. In trials, the Advance board has also been shown to lower overall system power consumption at the same time as significantly increasing performance. As the dense linear algebra workload is undertaken on the power efficient accelerator board, the host CPU CPU in full central processing unit Principal component of a digital computer, composed of a control unit, an instruction-decoding unit, and an arithmetic-logic unit. uses much less power, resulting in a reduction in the total system power consumption E[acute accent]"The Advance board's performance results on the LINPACK benchmark represent ClearSpeed's potential in revolutionizing the HPC industry," said John Gustafson, ClearSpeed CTO of HPC. "Our performance per watt advantage allows us to exceed previous performance levels while addressing the problems of power consumption and heat dissipation." E[acute accent]LINPACK is a collection of Fortran routines that analyze systems of linear equations such as those that arise in physics simulations and financial models. One routine in the collection has become the standard benchmark for measuring the performance of scientific computers and is most notably used by the Top 500 organization in ranking the world's 500 most powerful computer systems. While it is common to optimize the benchmark program for specific architectures, ClearSpeed achieved the 30.2 GFLOP GFLOP Giga-Floating Point Operations score without changing a single line of the standard benchmark code. E[acute accent]"ClearSpeed appears to be a very interesting technology for high performance with very low power consumption," said Jack Dongarra, maintainer of the LINPACK benchmark. E[acute accent]ClearSpeed's announcement of 30.2 GFLOPS sustained on the LINPACK benchmark coincides with its presence at SC05, the premier international conference on high performance computing. In booth number 802 ClearSpeed will demonstrate its workstation LINPACK performance and power consumption. Additional demonstrations will show the Advance board in a variety of manufacturers systems, including those from Compusys, IBM and Sun Microsystems. It will also be present in a number of other booths, including those of AMD (Advanced Micro Devices, Inc., Sunnyvale, CA, www.amd.com) A major manufacturer of semiconductor devices including x86-compatible CPUs, embedded processors, flash memories, programmable logic devices and networking chips. , Cray, IBM, Intel, NCSA (1) (National Center for Supercomputing Applications, Urbana-Champaign, IL, www.ncsa.uiuc.edu) A high-performance computing facility located at the University of Illinois at Urbana-Champaign. and Sun Microsystems. E[acute accent]About ClearSpeed E[acute accent]Established in 1997, ClearSpeed Technology delivers high performance coprocessors to be used alongside general purpose processors in the world's most compute-intensive applications. Exploiting a unique data parallel architecture, ClearSpeed's CSX coprocessors deliver unmatched performance while operating at extremely low power. ClearSpeed's Advance(TM) accelerator board contains two ClearSpeed CSX600 coprocessors and sustains 50 GFLOPS (DGEMM) double precision floating point performance at less than 25 Watts. The board(s) fit in to most standard PCI-X (PCI eXtended) An enhanced PCI bus technology originally developed by IBM, HP and Compaq that is backward compatible with existing PCI cards. PCI and 32-bit PCI-X slots are physically the same, and PCI cards can plug into PCI-X slots. slots and can be used in workstations, servers or clusters. ClearSpeed has offices in San Jose, Calif. and Bristol, UK and has over 67 patents granted and pending. For more information visit www.clearspeed.com. |
|
||||||||||||||

Printer friendly
Cite/link
Email
Feedback
Reader Opinion