28773 results about "Computer architecture" patented technology

A three-dimensional flash memory array incorporates thin film transistors having a charge storage dielectric arranged in series-connected NAND strings to achieve a 4F2 memory cell layout. The memory array may be programmed and erased using only tunneling currents, and no leakage paths are formed through non-selected memory cells. Each NAND string includes two block select devices for respectively coupling one end of the NAND string to a global bit line, and the other end to a shared bias node. Pairs of NAND strings within a block share the same global bit line. The memory cells are preferably depletion mode SONOS devices, as are the block select devices. The memory cells may be programmed to a near depletion threshold voltage, and the block select devices are maintained in a programmed state having a near depletion mode threshold voltage. NAND strings on more than one layer may be connected to global bit lines on a single layer. By interleaving the NAND strings on each memory level and using two shared bias nodes per block, very little additional overhead is required for the switch devices at each end of the NAND strings. The NAND strings on different memory levels are preferably connected together by way of vertical stacked vias, each preferably connecting to more than one memory level. Each memory level may be produced with less than three masks per level.

A semiconductor memory device includes a memory string coupled to a bit line, a page buffer configured to sense a sensing current of the bit line in an erase verification operation or a program verification operation, and a sensing control circuit configured to differently set a level of the sensing current in the erase verification operation and the program verification operation in order to sense the threshold voltage level of a selected memory cell of the memory string.

A novel massively parallel supercomputer of hundreds of teraOPS-scale includes node architectures based upon System-On-a-Chip technology, i.e., each processing node comprises a single Application Specific Integrated Circuit (ASIC). Within each ASIC node is a plurality of processing elements each of which consists of a central processing unit (CPU) and plurality of floating point processors to enable optimal balance of computational performance, packaging density, low cost, and power and cooling requirements. The plurality of processors within a single node may be used individually or simultaneously to work on any combination of computation or communication as required by the particular algorithm being solved or executed at any point in time. The system-on-a-chip ASIC nodes are interconnected by multiple independent networks that optimally maximizes packet communications throughput and minimizes latency. In the preferred embodiment, the multiple networks include three high-speed networks for parallel algorithm message passing including a Torus, Global Tree, and a Global Asynchronous network that provides global barrier and notification functions. These multiple independent networks may be collaboratively or independently utilized according to the needs or phases of an algorithm for optimizing algorithm processing performance. For particular classes of parallel algorithms, or parts of parallel calculations, this architecture exhibits exceptional computational performance, and may be enabled to perform calculations for new classes of parallel algorithms. Additional networks are provided for external connectivity and used for Input / Output, System Management and Configuration, and Debug and Monitoring functions. Special node packaging techniques implementing midplane and other hardware devices facilitates partitioning of the supercomputer in multiple networks for optimizing supercomputing resources.

In a non-volatile semiconductor memory system (or other type of memory system), a memory cell is programmed by changing the threshold voltage of that memory cell. Because of variations in the programming speeds of different memory cells in the system, the possibility exists that some memory cells will be over programmed. That is, in one example, the threshold voltage will be moved past the intended value or range of values. The present invention includes determining whether the memory cells are over programmed.

An exemplary cluster system according to the present invention includes a first node including a plurality of paths respectively connected to I / O slots via a switch and a second node including a plurality of paths connected to the switch, wherein the first node includes a first load measuring part which measures a first load for each path thereof, and the switch makes switching to disconnect an I / O slot from the first node and connect the I / O slot to the second node when the first load of the path connected to the I / O slot is higher than a first threshold.