Quadrics Supercomputers World Ltd.
Brief information about the company and products. Also availiable: russian version
In April 1996, Alenia Spazio (a Finmeccanica Company) and Meiko Ltd. formed "Quadrics Supercomputers World Limited", based in Bristol (UK) and Roma (Italy).
QSW consolidates two differing technologies produced in Europe, namely Quadrics system (SIMD), derived from the INFN APE100 project - a high performance array processor system suitable for certain intensive applications - and CS2 - a scaleable and more "general purpose" supercomputer solution. CS2 Systems have been installed in some important Centres of Research (CERN in Geneva, Lawrence Livermore National Laboratories in the USA) where they have been utilised in a large range of scientific applications. APE100 systems have been installed in Centres of Research for High Energy Physics as INFN, DESY and in others Centres like ENEA, CNR, University of Orsay.
QSW will utilise the present technologies and interim developments in order to supply High Performance Computer Systems, with capacity from 1,6 to 100 GigaFlops. In the meantime QSW is also developing new products, which will bring this performance up to and beyond the 1 TeraFlop regime. For the year 2000, it is foreseen that the realization of a system able to offer performances up to 10 TeraFlops.
QSW's follow-on product to the CS2 is called QM1, which will be in full production in 1998 and is based on the next generation UltraSparc processors and communication network. This represents an important achievement in terms of overall performance as well as price/performance ratio. First installations have been carried out in the frame of STRIDE program.
QSW's follow-on product to the APE 100 (APEmille) is a joint development programme between teams in Bristol and Rome in co-operation with the INFN. The APE1000 Project, will combine these skills in a new range of industrial products. This machine will have a peak computing performance of more than 1 Tflops and with a price of approximately 8000$ per GigaFlops.
The new system extends the interconnect technology derived from the MPP architecture in the delivery of scalable systems. By the end of 2000 the total number of nodes within the system will be increased to over 1000, coupled with increasing CPU count per node to 32, providing the availability of machines with a total computing power in the range of 100 TFlops by 2002
More information about QSW is availiable at http://www.quadrics.com.
© Лаборатория Параллельных информационных технологий НИВЦ МГУ
