Speed SRAM Research Articles

Reflections on Ethics, Human Intelligence and Artificial Intelligence (AI)

The history of AI is very long, which began in the mid-50s of the last century, since the creation of the Internet for military purposes (Darpanet) which became due to its great utility in the commercial and application form in all the fields of knowledge and human endeavor. These programs that initially ran in the operating systems of the time in the primitive Internet, have evolved along with the technological advances in programming languages, operating systems and software with such innovation capabilities that it has allowed their architecture with transistors and increasingly smaller chips, high-speed processing of text, number and matrix tables with formulas; coupled with the speed of RAM and hard disk memory. In this platform, AI has grown to the unimaginable, until it becomes a program with codes and algorithms that learn and re-map themselves to carry out their pre-established task more efficiently, a situation that when it happens, means improvement of the program but results and type of programming with codes and algorithms that are no longer known by the programmer, so in view of the risk of deviation from the pre-established objectives and taking care of ethical laws and regulations, the programs of this software must install filters at the beginning, during the process and at the end, as signals alarm when there are deviations such that they have ethical implications for humans. This interaction of Human Intelligence (HI) with AI has had negative as well as positive disagreements, for which in principle, the adaptation of labor and human rights standards and laws was sufficient, but now with the great development of AI, The establishment of ethical standards in the development of such programs is mandatory. At the moment, the standard already formally exists: Barcelona Declaration for the Adequate Development and Use of Artificial Intelligence in Europe.

Parallelizing the queries in a bucket-brigade quantum random access memory

Quantum algorithms often use quantum RAMs (QRAM) for accessing information stored in a database-like manner. QRAMs have to be fast, resource efficient and fault-tolerant. The latter is often influenced by access speeds, because shorter times introduce less exposure of the stored information to noise. The total execution time of an algorithm depends on the QRAM access time which includes: 1) address translation time, and 2) effective query time. The bucket brigade QRAMs were proposed to achieve faster addressing at the cost of exponentially many ancillae. We illustrate a systematic method to significantly reduce the effective query time by using Clifford+T gate parallelism. The method does not introduce any ancillae qubits. Our parallelisation method is compatible with the surface code quantum error correction. We show that parallelisation is a result of advantageous Toffoli gate decomposition in terms of Clifford+T gates, and after addresses have been translated, we achieve theoretical $\mathcal{O}(1)$ parallelism for the effective queries. We conclude that, in theory: 1) fault-tolerant bucket brigade quantum RAM queries can be performed approximately with the speed of classical RAM; 2) the exponentially many ancillae from the bucket brigade addressing scheme are a trade-off cost for achieving exponential query speedup compared to quantum read-only memories whose queries are sequential by design. The methods to compile, parallelise and analyse the presented QRAM circuits were implemented in software which is available online.

Trilayer Graphene as a Candidate Material for Phase-Change Memory Applications

ABSTRACTThere is pressing need in computation of a universal phase change memory consolidating the speed of RAM with the permanency of hard disk storage. A potentiated scanning tunneling microscope tip traversing the soliton separating a metallic, ABA-stacked phase and a semiconducting ABC-stacked phase in trilayer graphene has been shown to permanently transform ABA-stacked regions to ABC-stacked regions. In this study, we used density functional theory (DFT) calculations to assess the energetics of this phase-change and explore the possibility of organic functionalization using s-triazine to facilitate a reverse phase-change from rhombohedral back to Bernal in graphene trilayers. A significant deviation in the energy per simulated atom arises when s-triazine is adsorbed, favoring the transformation of the ABC phase to the ABA phase once more. A phase change memory device utilizing rapid, energy-efficient, reversible, field-induced phase-change in graphene trilayers could potentially revolutionize digital memory industry.

Mixture of Trees Probabilistic Graphical Model for Video Segmentation

We present a novel mixture of trees probabilistic graphical model for semi-supervised video segmentation. Each component in this mixture represents a tree structured temporal linkage between super-pixels from the first to the last frame of a video sequence. We provide a variational inference scheme for this model to estimate super-pixel labels, their corresponding confidences, as well as the confidences in the temporal linkages. Our algorithm performs inference over full video volume which helps to avoid erroneous label propagation caused by using short time-window processing. In addition, our proposed inference scheme is very efficient both in terms of computational speed and use of RAM and so can be applied in real-time video segmentation scenarios. We bring out the pros and cons of our approach using extensive quantitative comparisons on challenging binary and multi-class video segmentation datasets.

Selective Forward Body Bias for High Speed and Low Power SRAMs

Leakage power consumption is a large fraction of the total power consumption in contemporary VLSI designs. Since memories occupy a large fraction of the total area of many high-performance ICs, it is crucial to reduce the leakage energy of memories. This problem is particularly aggravated for memories implemented in the 45 nm technology node, since these processes have significantly higher leakage power. For these memories, leakage is a significant problem not only from a power point of view, but also from a performance degradation standpoint. In this paper, we quantify this problem and provide a solution, using a 45 nm 512 KByte SRAM as a design example. We show that implementing the SRAM as a monolithic memory results in increased delay as well as power. We illustrate a methodology to optimally reduce leakage power and improve performance in memories by splitting the memory array into word line groups (WLGs) and forward body bias a WLG selectively when it is accessed. We present a derivation of optimal number of WLGs and the forward body bias voltage value, and show that our approach results in a 9.2% access time reduction, and a 53.4% reduction in power during a read operation. Our approach also achieves an 18% reduction in power during a write operation and a 69% leakage power improvement. The area overhead of our scheme is 7.2% compared to a monolithic memory. Our approach does not compromise on the static noise margin (SNM). From an architectural standpoint, we also present a strategy to decide when a WLG should be left forward biased between non-successive accesses, in order to minimize the total power consumption of memory.

Switching and reliability issues of magnetic tunnel junctions for high-density memory device

Abstract Magnetoresistive random access memory (MRAM) is one of the promising universal memories, which combines the high speed of static RAM, high-density of dynamic RAM, and non-volatility of FLASH memory. However, MRAM faces several challenges prior to the appearance at the commercial market on a large scale. The most important challenge among them will be an issue, related to the magnetic switching. The conventional MRAM structures and writing methods cause several issues to be overcome for high packing density without cross-talk. Furthermore, small ferromagnetic elements will require high magnetic field, which will increase the power consumption of the devices. This article will review new writing schemes, such as thermal-assisted switching and spin-transfer switching methods. In addition, reliability characteristics and thermal stability of magnetic tunnel junctions will be presented, because ensuring the reliability of magnetic tunnel junctions emerges as an another challenging problem for the successful application of the new writing schemes to the high-density MRAM devices in the next generation.

Investigation of extrusion ratio effect on mechanical behaviour of extruded alloy AA-6101 from the billets homogenised-rapid quenched and as-cast conditions

This article describes the effect of extrusion ratio in direct extrusion process on mechanical behaviours of alloy AA-6101. The alloy is used mainly as feedstock in manufacturing of AAAC type aluminium alloy conductors and extrusions for compression joints and connectors in electrical industry. Extrusion ratio effect was examined by extruding the billets at different billet temperatures. The billet structures have been prepared as homogenised-rapid quenched (HRQ) and unhomogenised-slow cooled (UHSC). Keeping up constant the forward speed of the main ram and the exit temperature of the extrusions, AA-6101 billets were extruded directly by using three different dies with diameters 10, 30 and 50 mm. Mechanical measurements of the specimens were made after T-1 and T-6 treatments. Extrusion ratio effect on tensile strength, elongation and hardness of extrusions from the AA-6101 billets was appreciated on the graphics.

Investigation extrusion ratio effect on mechanical behaviour of extruded alloy AA-6063

This article describes effect of extrusion ratio (ER) in direct extrusion process on mechanical behaviours of the alloy 6063 prepared in two different textures as homogenised and rapid quenched (HRQ) and unhomogenised and slow cooled (UHSC) conditions. In the study, AA-6063 alloy was directly extruded as rod by using three different single hole flat type dies with diameters 10, 30 and 50 mm. Forward speed of the main ram of 1600 t extrusion press was kept to be constant when performing of the tests. For suitable pre-heating of the billets, several pre-tests have been performed to keep exit temperatures of the profiles as constant. Six billets from the alloy AA-6063 were prepared as-cast UHSC condition. Another six billets were applied homogenisation and than cooled fast. Billets with two different textures were tested by extruding different ERs and different exit temperatures 510 and 460 °C, respectively. Then the samples from two different textures of the extrusions from the alloy AA-6063 were exposed to both T-1 and T-6 treatments. Consequently, organising various comparable graphics by data from mechanical tests, effects of ER versus to ultimate tensile strength, hardness and elongation percent have been examined.

Magnetic resonance imaging (MRI) as a method to investigate movement of water during the extrusion of pastes.

To assess the potential of Magnetic Resonance Imaging (MRI) as a method of detecting water movement during the extrusion of pastes. Plug samples were made from mixtures of model materials and microcrystalline cellulose with two water contents at two different ram speeds to simulate ram extrusion. The extrusion process was stopped at different stages and analyzed for water distribution using MRI to assess the influence of water content and the speed of ram on water movement as the extrusion process progresses. Two types of water movement were detected: vertical and radial. When extruding at the faster ram speed, water moved predominantly in the vertical direction, whereas when extruding at a slower ram speed it moved predominantly in the radial direction. At the beginning of the extrusion process a greater water movement in the wetter formulations was observed. MRI appears to be a useful approach to non-invasive water mapping, and is expected to contribute towards a greater understanding of the role of water in the extrusion of pastes.

The Modern Direct-Hydraulic System

The first hydraulic press invented by Joseph Bramah in 1795 employed the direct-hydraulic system; i.e. hydraulic pressure was directly supplied to the press cylinder by a hydraulic pump and, therefore, the pressure exerted by the press ram was directly proportional to the pressure supplied by the pump, and the speed of the press ram was directly proportional to the delivery of the pump. Later developments in the use of hydraulic power resulted in the invention of the hydraulic accumulator to store liquid under pressure. With the accumulator system the speed and pressure exerted by the press ram are not controlled by the pump, and in fact they cannot be controlled with any precision; also there is considerable wastage of power when the press ram is operating at less than full pressure. The advent of the high-speed reciprocating ram pump has produced the modern direct-hydraulic system in which the press and pump form one self-contained unit. The power to drive the pump is in direct proportion to the work done by the press, and the speed and pressure exerted by the press ram can be precisely controlled. The author considers that a saving of at least 75 per cent in electric power can be made by conversion from the accumulator system to the direct-hydraulic system. Indicator diagrams are reproduced in the paper to show the saving in power which can be achieved by the direct-hydraulic system, and a comparison is made between the power consumption, in kilowatt-hours, of a direct-hydraulic cartridge-drawing press and a mechanical double-rack press doing the same operation. The paper includes a number of illustrations of direct-hydraulic presses, ranging from those of Bramah to present-day types.