Microprocessors Deliver Teraflops Intel showed parts of modules from TFLOPS : A massively parallel computer that will perform math operation at a peak rate of 1.8 trillion/second - teraflops. An order of magnitude faster than any existing computer. $45.5 million project funded by DOE. Specialty chips to handle volume of data flashing between the P6s. Tackle complex problems w/ a "Russian doll" strategy centered on the P6 processors. Grouped in pairs that share a common memory, forming nodes or computation units; pairs of nodes share communication pathways on a single CPU. Arrays of boards fit into 80 cabinets . Coordinating 50 billion bytes of data each second are special purpose network chips, one for each board. It uses a lightweight kernet. SToring a complete copy of the processors' UNIX O/S on every node. . The lightweight kernel reduces the burden by limiting the full operating system to only a handful of nodes, leaving the rest to subsist on a minimum set of instructions. Other areas of parallel computing are likely to benefit from these innovations. /*---------------------------------------------------------------------*/ The Future of The Behavioral And Social Sciences 1. Integrate current data sets 2. Improve the coverage of longitudinal surveys. 3. Experiment with new methodologies to study nonlinear, dynamic systems. 4 Develop comparable international research to determine both the incidence and prevalence of behavioral and social phenomena. 5 Integrate quantitative and qualitative research methods more systematically to advance new theory. /*---------------------------------------------------------------------*/ Negative Viscosity in a Magnetic Fluid Magnetic fuild in a colloidal dispersion of a single-domain magnetic particles of about 10 nanometers. Ultrastable magnetic fluids remain liquid wehn highly magnetized, even in the most intense applied magnetic fields. MFs (magnetic fluids) are distinguished from ordinary fluids by the body and surface forces that arise yielding new fluid mechanical phenomena. Magnetic fluids are colloidal solutions of manetic nanoparticles suspended ina fluid carrier. Each particle is a parmanent magnetic dipole, and when the particle is not too small, the orientation of the dipole is locked into the crystal axis of the particle. A repulsive force acts as an elastic cushion. preventing the particles from sticking to each other. Thanks to Brownian translational motion, the particles do not settle in gravitational or magnetic fields. /*---------------------------------------------------------------------*/ Surprises on the Way from 1 - 2D quantum Magnets : The ladder Matterials To make the transition from the quasi-long range order in a chain of antiferromagnetically coupled S=1/2 spins to the true long-range order that occurs in a plane, one can assemble chains to make ladders of increasing width. Surprisingly, this crossover between one and two dimensions is not at all smooth. Ladders with an even number of legs have purely short-range magnetic order and a finite energy gap to all magnetic excitations. Predictions of this ground state have now been verified experientally. Holes doped into these ladders are predicted to pair and possibly superconduct. /*---------------------------------------------------------------------*/ Group velocity in strongly scattering Media Investigation of the ballistic propagation of acoustic waves through a resonantly scattering, inhomogeneous medium indicates that although the ballistic signal remains conherent with the incident pulse, it is nevertheless strongly affected by scattering resonances. These resonances cause considerable frequency dispersion and substantially reduce the phase and group velocities. The experiemntal data are quantitateively describbed by a theoretical model that correctly accounts for the coupling between the resonant scatterres, leading to a n effective renormalization of the scattering within the medium. This approach resolves a long-standing problem in the definition of the group velocity in strongly scattering materials. --fin