Форум » Дискуссии » Boewie wichislitel'nie kompleksi » Ответить
Boewie wichislitel'nie kompleksi
milstar: http://drops.dagstuhl.de/opus/volltexte/2006/732/pdf/06141.AthanasPeter.Paper.732.pdf Although an FPGA’s clock rate rarely exceeds one-tenth that of a PC, hardware implemented digital filters can process data at ###################################################################### many times that of software implementations [4] ################################### . Additional performance gains have been described for cryptography [5], network packet filtering [6], target recognition [7] and pattern matching [8], among other ########################################################################## applications. A. Present Day Cost-Performance Comparison Owing to the prevalence of IEEE standard floating-point in a wide range of applications, several researchers have designed IEEE 754 compliant floating-point accelerator cores constructed out of the Xilinx Virtex-II Pro FPGA’s configurable logic and dedicated integer multipliers [16-18]. Dou et al published one of the highest performance benchmarks of 15.6 GFLOPS by placing 39 floating-point processing elements on a theoretical Xilinx XC2VP125 FPGA [19]. Interpolating their results for the largest production Xilinx Virtex-II Pro device, the XC2VP100, produces 12.4 GFLOPS, compared to the peak 6.4 GFLOPS achievable for a 3.2 GHz Intel Pentium processor. Assuming that the Pentium can sustain 50% of its peak, the FPGA outperforms the processor by a factor of four for matrix multiplication. One of the earlier projects demonstrated a 23x speedup on a 2-D FFT through the use of a custom 18-bit floating-point format [26]. More recent work has focused on parameterizible libraries of floating-point units that can be tailored to the task at hand [27-29]. By using a custom floating-point format sized to match the width’s of the FPGA’s internal integer multipliers, a speedup of 44 was achieved for a hydrodynamics simulation [30] using four large FPGAs. Nakasato and Hamada’s 38 GFLOPS of performance is impressive, even from a cost-performance standpoint. For the cost of their PROGRAPE-3 board, estimated at $15,000, it is likely that a 15-node processor cluster could be constructed producing 196 single precision peak GFLOPS. Even in the unlikely scenario that this cluster could sustain the same 10% of peak performance obtained by Nakasato and Hamada’s for their software implementation, the PROGRAPE-3 design would still achieve a 2x speedup. As in many FPGA to CPU comparisons, it is likely that the analysis unfairly favors the FPGA solution. Hardware implementations require specialized skills in digital design and vendor-specific tool flows. Development time and costs are significantly higher than for software. Many comparisons in literature spend significantly more time optimizing the hardware implementations than they do optimizing their software implementations. Previous research has demonstrated significant compiler inefficiency for common HPCfunctions [31]. For the DGEMM matrix multiplication function, a hand-coded version outperformed the ############################################### compiler by greater than eight times. ############################ A to- tal of 39 PEs can be integrated into the xc2vp125-7 FPGA, reaching performance of, e.g., 15.6 GFLOPS with 1600 KB local memory and 400 MB/s external memory bandwidth 1 is s 1700 nozkami i wisokoj stoimost'ju porjadka 8000 $ segodnja http://ce.et.tudelft.nl/~george/publications/Conf/FPGA05/FPGA05Dou.pd http://www.xilinx.com/publications/matrix/virtexmatrix.pd Xilinx Vertex FPGA
Ответов - 76, стр:
1 2 3 4 All
milstar: Radiation hardened by process is still the only demonstrated method for achieving the full range of radiation hardened program requirements,” Nootbaar says. “Honeywell would anticipate hardening by process to continue to be the method at least through 90 nanometer development. Programs continue to adapting commercial technologies using redundancy at the circuit and chip level. “Beyond 90 nanometers, where the cost of capitalizing a 65 nanometer wafer fab or smaller is extremely high compared to earlier geometries, radiation hardening by design such as triple redundancy voting will most likely be the most accepted method,” Nootbaar says. “This will not provide the strategic hardness that is required by some applications, and therefore 90 nanometer and 150 nanometer radiation hardened by process will continue to be utilized for these critical applications. In addition, radiation hardened by design will lose some of the geometry benefits with respect to power and density due to the redundancy and will effectively be equivalent to one generation earlier technology.” For example, 65 nanometer radiation hardened by design would be equivalent to 90 nanometer radiation hardened by process for power and density, he continues. “Therefore, it is believed that the next generation technology beyond 90 nanometer would be radiation hardened by design at the 45 nanometer process node.
milstar: Rad-hard company information Actel, Mountain View, Calif. 650-318-4200 www.actel.com Aeroflex, Colorado Springs, Co. 719-594-8000 www.aeroflex.com/ams Aitech, Chatsworth, Calif. 818-700-2000 www.rugged.com ASIC Advantage, Sunnyvale, Calif. 408-541-8686 www.asicadvantage.com Atmel, San Jose, Calif. 408-441-0311 www.atmel.com BAE Systems, Manassas, Va. 703-361-1471 www.baesystems.com C-MAC Micro Technology, Buckinghamshire, United Kingdom 44 1628 859 680 www.cmac.com Crane Interpoint, Redmond, Wash. 425-882-3100 www.interpoint.com Harris Corp., Melbourne, Fla. 800-4-HARRIS www.govcomm.harris.com Honeywell, Plymouth, Minn. 612-951-1000 www.honeywell.com International Rectifier, El Segundo, Calif. 310-726-8000 www.irf.com Intersil Corp., Milpitas, Calif. 408-432-8888 www.intersil.com Linear Technology Corp., Milpitas, Calif. 408-432-1900 www.linear.com Maxwell Technologies, San Diego, Calif. 877-511-4324 www.maxwell.com Microelectronics Research Development Corp., Colorado Springs, Colo. 719-531-0805 www.micro-rdc.com Micropac Industries, Garland, Texas 972-272-3571 www.micropac.com Modular Devices, Shirley, N.Y. 631-345-3100 www.mdipower.com MS Kennedy, Liverpool, N.Y. 315-701-6751 www.mskennedy.com Peregrine Semiconductor Corp., San Diego, Calif. 858-731-9400 www.psemi.com Radiation Assured Devices, Colorado Springs, Colo. 731-531-0800 http://radiationassureddevices.com Semicoa, Costa Mesa, Calif. 714-979-1900 www.semicoa.com STMicroelectronics, Geneva, Switzerland 41 22 929 29 29 www.st.com Synova, Melbourne, Fla. 321-728-8889 www.synova.com Teledyne Microelectronic Technologies, Los Angeles, Calif. 310-822-8229 www.teledynemicro.com 3D-Plus, McKinney, Texas 214-733-8505 www.3d-plus.com Triad Semiconductor, Winston-Salem, N.C. 336-774-2150 www.triadsemi.com Ultra Communications, Vista, Calif. 760-652-0011 www.ultracomm-inc.com VPT, Everett, Wash. 425-353-3010 www.vpt-inc.com Xilinx, San Jose, Calif. 408-559-7778 www.xilinx.com
milstar: Председателем совета директоров ОАО «Ангстрем» Дмитрий Милованцев был избран в конце 2008 г. Своей главной задачей он называл разработку нового продуктового ряда чипов по топологии 130 нм. Напомним, что на покупку оборудования и технологий AMD для производства чипов с таким топологическим размером компания «Ангстрем-Т» взяла кредит у Внешэкономбанка на сумму 815 млн евро сроком на 9 лет. Строительство завода планировалось завершить осенью 2009 г. Реально «Ангстрем-Т» смог получить около 290 млн евро, практически все из которых были перечислены AMD. В настоящее время завод так и не построен. Более того, оборудование до сих пор не привезено в Россию. Оно находится на складе в Нидерландах. i yze ystarelo ,esche ne doexaw do Rossii http://www.rosrep.ru/news/index.php?ELEMENT_ID=3474&SECTION_ID=12
milstar: Акционеры АФК "Система" одобрили условия сделки с РОСНАНО по созданию производства интегральных схем размером 90 нм Акционеры ОАО "АФК "Система" на внеочередном собрании 8 февраля одобрили условия сделки в рамках инвестиционного договора с государственной корпорацией "Российская корпорация нанотехнологий" (РОСНАНО) по созданию серийного производства интегральных схем на основе наноэлектронной технологии с проектным нормами 90 нм на пластинах диаметром 200 мм. В рамках проекта планируется создать на территории РФ серийное производство сверхбольших интегральных схем с проектными нормами 90 нм на базе вновь создаваемой проектной компании в форме ООО "Ситроникс- Нано". Уставный капитал ООО "Ситроникс-Нано" составит 2 592 000,00 руб., с последующими его увеличениями до 6 486 485 000,00 руб. и 12 966 483 515,00 руб. Проект будет реализован на производственной базе с использованием инфраструктуры завода "Микрон", который входит в состав ОАО "Ситроникс", крупнейшим акционером которого является АФК "Система". Технологическим партнером проекта выступает компания "ST- Mикроэлектроникс". Продукция нацелена на развивающиеся массовые рынки в области цифрового телевидения, ГЛОНАСС/GPS-навигации, системы автоматизации производства, автомобильной электроники и смарт-карт с высокой степенью защиты. Общий объем финансирования проекта составит 16,5 млрд. руб. При этом объем инвестиций РОСНАНО составит 6,5 млрд. руб. "Ситроникс" вложит в проект равную долю в виде высокотехнологичного оборудования завода "Микрон". В долгосрочной перспективе предусматривается также привлечение взаимного финансирования. "Umpro.ru" 19 февраля 2010 года linija 0.09 microna ,t.e. 100 gigaflop na odnom chipe ne realizowat* Resultat NEC SX-9 na 0.065 microna -3-letnej dawnosti nojabr* 2007
milstar: powtor - http://www.nec.co.jp/press/en/0710/2501.html Tokyo, October 25, 2007 - NEC Corporation today announced the worldwide launch and availability of the SX series model "SX-9," the world's fastest vector supercomputer with a peak processing performance of 839 TFLOPS*1. The SX-9 features the world's first CPU capable of a peak vector performance of 102.4 GFLOPS*2 per single core. Klientam stran NATO mozno bilo kupit* 3 goda nazad na texnologii 0.065 microna single chip vektornij processor 102 gigaflop
milstar: 1.NEC SX-9 0ktjabr* 2007 moznobilo kupit* klientam s stran NATO ,vektornij odin processor na odnom chipe 102 gigaflop ,0.065 micron ,3.2 gigagerz 2. Fujitsu The SPARC64 VII is the latest processor developed by Fujitsu for the SPARC64 Series. It uses the 65-nm technology of Fujitsu and it has an operating frequency of 2.5 GHz. The chip measures 21.3 mm by 20.9 mm. The chip has four built-in cores with a shared 5-MB L2 cache configuration. The operating power consumption is 135 W. ----------------------------------------------------- 4-processora na odnom kristale -pri toj ze texcnologii wsego 40 gigaflop The FX1, using a new architecture designed to extract the maximum performance from the SPARC64™ VII quad-core CPU developed by Fujitsu The FX1's performance is 40 gigaflops(4) per node ( i chip s 4 processorami) http://www.fujitsu.com/global/news/pr/archives/month/2008/20081119-01.html http://www.fujitsu.com/downloads/MAG/vol44-4/paper08.pdf wneschnij wid stojki na str. 8 7po textu str.2) http://www.fujitsu.com/downloads/SPARCE/whitepapers/sparc-architectureM3000-wp-e-081028.pdf
milstar: dlja srawnenija -chto sejtshas proektirujut w MCST -postwschik EVM dlja S-400/500 i FSB na texnologii 0.09 microna ( o 0.065 microna 2007 esche ne rassmatriwaetsja publichno) serijno wipuskaetsja na 0.13 microna СИСТЕМА НА КРИСТАЛЛЕ «МЦСТ-4R» http://www.mcst.ru/b_22-23.shtml СИСТЕМА НА КРИСТАЛЛЕ «ЭЛЬБРУС-S» http://www.mcst.ru/b_8-9.shtml Tel: +7 (495) 363-9665 | Fax: +7 (495) 363-9599 | E-mail: mcst@mcst.ru Konechno wse zawisit ot zadach *************************** wpolne dopuskaju chto dlja zadach S-400/s-500 bolee chem dostatochno 0.13 microna
milstar: Bolee swezaja informazija o rinke UNIX serverow ( Sun Sparc ,Fujitsu Sparc i ... ELbrus Sparc -poka tolko Processor dlja S-400/500 i FSB na texnologii 0.13 microna ) Unix server sales sink; IDC cites a perfect storm Sharp drop in Unix server sales seen caused by recession, pending new products and Oracle-Sun merger Patrick Thibodeau (Computerworld (US)) 28/05/2010 06:38:00 Tags: unix, supercomputing, server market, IDC, HP, Gartner The Unix server business has been hit particularly hard during the current recession. The latest market share server numbers from IDC, released Wednesday, show that users have put off purchasing new Unix systems in recent months, cutting its share of overall server spending to one of the lowest levels ever. The IDC report, released Wednesday, shows worldwide Unix revenues of $US2.3 billion, or just over 22% of total ************************************************************************************************ spending on servers, during the first 2010 quarter. ****************************************** Mirowoj rinok primerno 10 mlrd $ w god , w predelax Rossii **************************************************** MCST i swjazannie s nim 5 institutow mogut poiskat* sbit s goskompanijax (tipa ros.zeleznie dorogi) *********************************************************************************** The Unix server share of server revenue is down 10.5 percentage points from the year ago quarter. Unix servers are mid-to-high end system traditionally used to run mission critical systems, but their use has been gradually eroding as the capabilities of x86 servers increase. The Unix servers run several Unix variants, including Solaris, AIX and HP-UX. The shift from such Unix variants is perhaps most dramatic in the supercomputing business. The Top500 supercomputer list offers detailed statistics on processors, architecture and operating systems deployed in these systems. Late last year, Linux was the operating system of choice for just over 78% of the Top500 supercomputers, and even a bit higher when Red Hat and Suse-specific implementations are added. The highest ranked Unix system was IBM's AIX at 4.4%. By comparison, Linux accounted for 61% share of the Top 500 list in 2005, while the highest ranked Unix system was Hewlett-Packard's HP-UX at nearly 15%. The latest supercomputing list will be out Monday. Despite the latest numbers, it is important to note that Unix still accounts for a major share of server revenue, and remains an important part of the hardware mix. Jean Bozman, an analyst at IDC who follows this market, cited a combination of factors in the sharp drop in Unix server sales in the first quarter. First, she said, the recession delayed sales of Unix servers, which are typically replaced every five to seven years -- ************************************************************************************************ longer than most x86 systems. Unix server buyers may also be holding back as Oracle Corp. moves to fully absorb Sun Microsystems Inc. after its deal to buy the key Unix server vendor closed in January. Further, Bozman noted that Unix server users may be waiting for product upgrades from Hewlett-Packard Co., which recently announced new products in its Integrity line, and IBM, which is due to release Unix servers later this year. "It isn't just one thing, it is multiple things," said Bozman, of the Unix server sales drop. Bozman said it is too early to say whether the first quarter results indicate that users are accelerating a shift away from Unix. Overall, IDC said the worldwide server market increased by 4.6% year-over-year to $10.4 billion in the first quarter. Gartner Inc., this week put the server growth at 6% for the quarter and the total market at $10.75 billion. Both firms cited the x86 systems as key to the growth. wes* rinok s window pop izdeliami 42 mlrd $ ,no window k MCST ne imeet otnoschenija ,UNIX tolko 10 mlrd $ ----------------------------------------------------------------------------------------------------------------------------------- "It's really going to take a while to see what the shape of that will really be." Charles King, an analyst at Pund-IT Inc. in Hayward, Calif., of the broader Unix trend. Oracle may be "the biggest question mark, although the company has thrown its weight behind Sun's UltraSparc Unix systems. [Oracle] said that it will continue development, but it is going to take a while for us to really see what the shape of that is going to be," said King.
milstar: dlja sprawki SPARC64 VIIIfx maj 2009 ,0.042 microna ,8 core , 200 kwadratnix millimetrow ,128 gigaflop Srawnit 1 chip 200 kw .mm s 150 stuk 300 mm diamtrom wafer w chas ot odnoj sistemi Elbrus ot S-400/500 eto toze Sparc versija http://img.jp.fujitsu.com/downloads/jp/jhpc/sparc64viiifx-extensions.pdf maj 2009 goda Fujitsu today debuted a new super chip at a technology event in Japan that it claims to be the world’s fastest CPU. It’s called the Venus SPARC64 VIIIfx which i s capable of 128 GFLOPs--Intel's top of the line Nehalem-based Xeon 5500 for example, achieves roughly 76 GFLOPs. --------------------------------------------------------------------------------------------------------------------------------------- ZoomThe chip is supposedly built using a 45-nm process and, from what we can make it out from the original Japanese report on PC Watch (translated), has an integrated memory controller. The original report also states that Fujitsu’s chip has 2.5 times the “high speed operation” and one-third the power consumption of an Intel CPU ----------------------------------------------------------------------------------------------------------- (though it does not state which chip in specific). Being that this new chip is a SPARC64 CPU, it’ll be reserved specifically for industrial and scientific applications, such as for studying space, weather, astronomy, among --------------------------------------------------------------------------------------------------------------------------------------------- others --------- W Rossii Sparc versii takze dlja S-400/500 i Kompjuteri dlja FSB Elbrus ------------------------------- http://winfuture.de/news,49320.html Fujitsu plant 10 PetaFLOPS Supercomputer für 2011 Der japanische Elektronikkonzern Fujitsu will bereits Anfang 2011 einen neuen Supercomputer fertig stellen, der eine Leistung von 10 PetaFLOPS bieten soll. Das System wäre fast zehn Mal leistungsstärker als die bisher führenden Supercomputer. 10 w 16 = 10 Petaflops ---------------------------- Fujitsu gab seine Pläne am Dienstag anlässlich der Hot Chips Konferenz an der Universität Stanford bekannt, berichtet 'InfoWorld'. Der neue Supercomputer wird für das Japan Institute of Physical and Chemical Research (RIKEN) gebaut. Seine Basis bildet der neue Sparc64 VIIIfx Prozessor von Fujitsu, der über acht Rechenkerne verfügt. Fujitsu budet ispolzowat*w superkompjuter tot che chip ,chto i w swoej linejke Unix serverow Eto xoroschee reschenie -processor SPARC eto inzenernij proekt kotorij razrabatiwalsja pod UNIX ----------------------------------------------------------------------------------------------------------------------- w otlichii ot paketa Intel/Window dlja pop mass ----- Abstract The Sparc64 VIIIfx eight-core processor, developed for use in petascale computing systems, runs at speeds of up to 2 GHz and achieves a peak performance of 128 gigaflops while consuming as little as 58 watts of power. Sparc64 VIIIfx realizes a six-fold improvement in performance per watt over previous generation Sparc64 processors. http://ieeexplore.ieee.org/xpl/freeabs_all.jsp?isnumber=5446241&arnumber=5446249
milstar: n the third quarter of 2007, ************************** ASML began volume shipment of the XT:1900i, with a new industry benchmark of 1.35 NA, which is close to the practical limit for water-based immersion technology. This optical lithography system is capable of volume productionof ICs down to 40 nm and below and is used for high volume IC manufacturing at multiple customers worldwide **************************** In the third quarter of 2008, ASML announced an enhanced version of the XT:1900i system, the XT:1950i, with improved throughput of 148 wafers per hour, resolution of 38 nm and a scheduled overlay of 4 nm. In the first quarter of 2009, we startedshipments of XT:1950i systems, which extend the performance, imaging and overlay specifications of the successful XT:1900i system. Also by the end of 2009, ASML had received five orders for next generation EUV systems, the first of which is scheduled for shipment in the second half of 2010. EUV will provide a large “process window” and much greater shrink compared to current approaches and we expect it to be a multi-generation lithography solution. The first generation of these systems will combine a wavelength of 13.5 nm and a lens system with a NA of 0.25 to provide imaging at a resolution of 27 nm. The EUV platform is targeted for production of ICs down to 22 nm and beyond. ''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''''' 150 wafer po 300 mm w chas ot odnoj sistemi za 25-28 mln ewro Na perwoe wremja dolzno xwatit ***************************** 25-28 mln ewro + 1 komnata klassa 1 dlja kompanii Ivanova,Chemzowa ,Kudrina ne den*gi ... daze ypominat neprilichno takuju summu
milstar: popravka ploschad* spark 128 gigaflop 500 kw.mm ,a ne 200 kak ykaziwalos ranee 760 mln transistorow ,1271 nozek ,58 watt ,128 gigaflop http://img.jp.fujitsu.com/downloads/jp/jhpc/090825HotChips21.pdf
milstar: Август 2010 года Очередная поставка заказчикам вычислительных средств, использующих двухядерные российские микропроцессоры разработки ЗАО "МЦСТ" В адрес ОАО "Таганрогский завод "Прибой" поставлены вычислительные комплексы, разработанные на базе процессорных модулей МВС/С с жесткими условиями эксплуатации. Процессорный модуль МВС/С содержит четыре двухядерных микропроцессора R-500S, таким образом представляя собой одноплатную 4-х машинную систему с общим количеством процессоров, равным восьми. ---------------------------------------------------------------------------------------------------------- Он выпускается и в варианте МВС/С-К с кондуктивным охлаждением. Завод является четвертой организацией, которая применяет модули этого типа в разрабатываемых радиотехнических системах. К настоящему времени в отдельной комплектации и в составе многомашинных вычислительных комплексов им поставлено 150 модулей. http://www.mcst.ru/news.shtml
milstar: ОАО "Таганрогский завод "Прибой" - крупное промышленное предприятие Юга России, специализируещееся, главным образом на разработке и серийном изготовлении гидроакустического вооружения для ВМФ, одно из ведущих приборостроительных предприятий страны. Продукцией завода оснащены практически все надводные корабли ВМФ, использующие гидроакустические средства, от тральщиков до больших противолодочных и авианесущих кораблей, а также около 20 проектов кораблей поставлявшихся и строившихся за рубежом http://www.priboy.ru/ Открытое Акционерное Общество "Таганрогский завод «ПРИБОЙ», в течении 45 лет разрабатывало и изготавливало аппаратуру специального назначения для нужд ВМФ, рыбопоисковую аппаратуру для рыбопромысловых судов и осталось единственным предприятием в Российской Федерации, способным решать задачу создания и поставок новейших образцов техники для ВМФ na perwoj stranize Priboy realizuet Texas Instruments DSP ! ---------------------------------------------------------------------- Esli Rossija(ee wozdi) xochet soxranit nezawisimost* WPK ,neobxodimo wipuskat* rossijskie ОАО "Таганрогский завод "Прибой" реализует микросхемы DCP010515DBP-U с приемкой "5" Производитель: Texas Instrument Год выпуска 2008 - 1109 штук Год выпуска 2010 - 550 штук Обращаться по телефону: (8634)39-00-59 Морозова Е.В. http://www.priboy.ru/
milstar: The HPC Market The International Data Corporation (“IDC”), a leading HPC market analyst firm, divides the HPC technical server market into four competitive segments by selling price: • supercomputers that sell for $500,000 and up; • divisional servers that sell for $250,000 to $499,999; • departmental servers that sell for $100,000 to $249,999; and • workgroup servers that sell for under $100,000. We primarily target the supercomputer segment with our products and services and our Cray CX products target the workgroup server market. Our Custom Engineering practices target both the high-performance computing market as well as high-end niches within the technical services market. IDC estimates that in 2008 the size of the entire HPC technical server market was $9.8 billion, with $2.7 billion in the supercomputer segment, and for 2009, IDC estimates that the HPC technical server market decreased to $8.5 billion, with the supercomputer segment increasing to $2.9 billion. See “Economic Crisis Response:Worldwide Technical Computing HPC Market, 2009-2013 Forecast Update” (November 2009) and Worldwide Technical Computing Server 2009-2013 Forecast Update” (May 2009). IDC assumes that the high-end supercomputer segment will continue to grow, less affected by the general economic slowdown, being sustained somewhat by long buying cycles and by an increasing number of petascale system purchases in the next one to three years. The IDC base forecast predicts the HPC market as a whole should return to growth in 2010, with the supercomputer segment having the highest compound annual growth rate of 7.1% over the 2008 to 2013 period. http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9NDIxNjN8Q2hpbGRJRD0tMXxUeXBlPTM=&t=1 The U.S. government and its various agencies have determined that it is in the best economic and security interest of the country to establish and maintain a leadership position in the development of supercomputing technologies. The largest of such initiatives is the Defense Advanced Research Projects Agency (“DARPA”) High Productivity Computing Systems (“HPCS”) initiative, which is a multi-phase initiative under which we have received funding for our Cascade program since 2002 and have a contract to receive funding for our Cascade program into 2012 to the extent we meet certain specified milestones and contribute minimum levels of funding. The DARPA program is designed to provide support for breakthroughs in high productivity supercomputing systems for the national security, research and industrial user communities.
milstar: We shipped our first Cray XT5 system in the second half of 2008. The “Jaguar” system at Oak Ridge National Laboratory, the largest and fastest computer system in our history and the first and, to date, the only system in the world to exceed more than one petaflops sustained performance on real scientific applications, is a Cray XT5 system with over 2.3 petaflops of peak performance. Cray XT5m System. Our Cray XT5m supercomputer is designed to make our HPC technology available to more users by targeting a lower price band in the supercomputer market segment with price points starting at approximately $500,000. The Cray XT5m system incorporates a version of our Cray SeaStar network specially designed and optimized for systems with peak performance of less than 70 teraflops, providing superior bandwidth, upgradeability and manageability at prices comparable to those of commodity clusters. Offered with up to six cabinets, the Cray XT5m series features many-core (four- or six-core) AMD Opteron processors and can be air or liquid cooled through use of Cray ECOphlex technology. The Cray Linux Environment enables the use of a wide range of open source tools as well as streamlined porting of a broad set of applications from independent software vendors. AMD i Linux originalno ne jawljautsja produkziej dlja inzenerow ... dlja pop massi ###################################################### Cray CX1 System. The Cray CX1 system, purpose-built for offices, laboratories and university departments requiring workgroup servers, incorporates up to eight nodes and 16 Intel Xeon processors, either dual or quad core, and delivers up to eight cores and up to 64 gigabytes of memory per node (with up to 64 cores per chassis), with up to 24 terabytes of internal storage within a chassis. Up to three chassis can be linked with the optional 24-port Infiniband switch allowing for expansion to 192 cores. Systems can be configured with a mix of compute, storage and visualization blades to meet customers’ individual requirements. The deskside system, which uses standard office power, features either the Windows HPC Server 2008 operating system or the Red Hat Enterprise Linux operating system. List prices start at around $15,000 and range to over $100,000. Intel toze produkzija dlaj pop massi .Originalnno Intel/Window
milstar: We believe our DARPA HPCS Phase III award, which began in 2006 and is expected to provide up to $190 million of co-funding of the research and development efforts towards building a prototype “Cascade” system, validates this vision. Our Cascade development program implements our Adaptive Supercomputing vision. Our Cascade efforts are co-funded by the U.S. government. Under our funding agreement with DARPA, we are to develop a prototype system that demonstrates the functionality required for scaling to multiple sustained petaflops levels of performance on real applications. Our system involves a new system architecture that combines future processor technologies, a new high-performance network and an adaptive software layer into a single integrated system. Pursuant to our agreement with DARPA we are obligated to spend at least $285 million of our funds, with DARPA reimbursing us up to $190 million. The DARPA program is milestone-based with a specified part of the DARPA reimbursement obligation associated with each milestone. Each milestone has specific requirements for information and deliverables that we are to provide and specified minimum exit criteria demonstrating that we are making required progress towards completion of the prototype system. DARPA provides a formal acceptance of each milestone, which is required for us to invoice for the associated DARPA payment. Overall, we anticipate spending in excess of the required $285 million to complete the program. As of March 10, 2010, we had met six milestones and had received a total of $110 million in cash payments from DARPA. Six milestones remain totaling up to $80 million with the final prototype demonstration milestone scheduled for the second half of 2012. We will own the final prototype system and will provide DARPA’s mission partners access to the prototype system for a period of six months following the completion of the DARPA program. http://phx.corporate-ir.net/External.File?item=UGFyZW50SUQ9NDIxNjN8Q2hpbGRJRD0tMXxUeXBlPTM=&t=1
полная версия страницы