Search NASA⌕ Search

SEARCH · Search NASA

Results for “DRAM”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 91 records · Page 5

Technology Development of a Solid State 266 nm Laser for NASA’s Dragonfly Mission

NASA’s Dragonfly mission is a rotorcraft lander which will explore several geologic locations on Saturn’s moon, Titan and investigate evidence of surface-level prebiotic chemistry as well as search for chemical signatures of water-based and/or hydrocarbon-based life. To perform molecular composition investigations in-situ, the payload includes the Dragonfly Mass Spectrometer (DraMS), being developed at NASA’s Goddard Space Flight Center (GSFC). DraMS will utilize laser desorption mass spectrometry (LDMS) to interrogate surface samples and measure the organic composition. Enabling this science capability is the Throttled Hydrocarbon Analysis by Nanosecond Optical Source (THANOS) laser being developed at NASA-GSFC. The THANOS laser is comprised of a solid state, passively Q-Switched Nd:YAG oscillator which is frequency converted to 266 nm and utilizes a RTP high voltage electro-optic for pulse energy control. The laser outputs <2.0 ns pulses with a maximum energy of approximately 200 uJ which can be emitted in 1 - 50 shot bursts at 100 Hz while performing LDMS science operations. While operating, the laser has the capability to throttle its UV pulse energy output from full attenuation to maximum energy to provide varying levels of fluence on samples in the DraMS instrument. This paper reports the technology development and space qualification effort of the THANOS laser including vibration, thermal vacuum cycling, radiation as well as Titan atmospheric composition optical damage testing performed at NASA-GSFC from 2019 through 2022.

Matthew W Mullin↗

Collective Memory Transfers for Multi-Core Chips

Future performance improvements for microprocessors have shifted from clock frequency scaling towards increases in on-chip parallelism. Performance improvements for a wide variety of parallel applications require domain-decomposition of data arrays from a contiguous arrangement in memory to a tiled layout for on-chip L1 data caches and scratchpads. How- ever, DRAM performance suffers under the non-streaming access patterns generated by many independent cores. We propose collective memory scheduling (CMS) that actively takes control of collective memory transfers such that requests arrive in a sequential and predictable fashion to the memory controller. CMS uses the hierarchically tiled arrays formal- ism to compactly express collective operations, which greatly improves programmability over conventional prefetch or list- DMA approaches. CMS reduces application execution time by up to 32% and DRAM read power by 2.2×, compared to a baseline DMA architecture such as STI Cell.

97 MATHEMATICS AND COMPUTING↗

A New Class of Single Event Hard Errors

This paper reports on hard errors induced by single ions in dynamic memories. For ions with atomic number below 80, hard errors in DRAMs appear to be similar to the hard errors reported in previous work on SRAMs. One feature of these hard errors is that they tend to recover gradually with time, because of annealing, and are thus partially recoverable. However, for gold ions, a second type of hard error was discovered which is not recoverable, and appears to be due to catastrophic internal shorting rather than small changes in leakage current. Thus, nonrecoverable errors will likely occur even in devices which eliminate the extreme sensitivity to leakage current that is inherent in 4-T SRAMs and DRAMs. It is important to understand the mechanism that is responsible for nonrecoverable errors, and investigate the effect of device scaling.

single event hard errors DRAM SRAM ions gold ions ↗

SharP Unified Memory Allocator: An Intent-Based Memory Allocator for Extreme-Scale Systems

The pre-exascale systems will soon be deployed with a deep, complex memory hierarchy composed of many heterogeneous memories. This presents multiple challenges for users including: how to allocate data objects with locality between memories and devices for the various memories in these systems, which includes DRAM, High-bandwidth Memory (HBM), and non-volatile random access memory (NVRAM), and how to perform these allocations while providing portability for their application. Currently, the user can make use of multiple, disjoint libraries to allocate data objects on these memories. However, it is difficult to obtain locality between memories and devices when using libraries that are unaware of each other. This paper presents the Unified Memory Allocator (UMA) of the SHARed data-structure centric Programming abstraction (SharP) library, which provides a unified interface for memory allocations across DRAM, HBM, and NVRAM and is extensible to support future memory types. In addition, the SharP UMA allows for portability between systems by supporting both explicit and implicit, intent-based memory allocations. To demonstrate the ease of use of the SharP UMA, we have extended both Open MPIand OpenSHMEM-Xto support SharP. We validate this work by evaluating the performance implications and intent-based approach with synthetic benchmarks as well as adaptations of the Graph500 benchmark.

Aderholdt, William↗

DeStager: feature guided in-situ data management in distributed deep memory hierarchies

In-situ analytics have been increasingly adopted by leadership scientific applications to gain fast insights into massive output data of simulations. With the current practice, systems buffer the output data in DRAM for analytics processing, constraining it to DRAM capacity un-used by the simulation. The rapid growth of data size requires alternative approaches to accommodating data-rich analytics, such as using solid-state disks to increase effective memory capacity. For this purpose, this paper explores software solutions for exploring the deep memory hierarchies expected on future high-end machines. Leveraging the fact that many analytics are sensitive to data features (regions-of-interest) hidden in the data being processed, the approach incorporates the knowledge of the data features into in-situ data management. It uses adaptive index creation/refinement to reduce the overhead of index management. In addition, it uses data features to predict data skew and improve load balance through controlling data distribution and placement on distributed staging servers. The experimental results show that such feature-guided optimizations achieve substantial improvements over state-of-the-art approaches for managing output data in-situ.

Computer Science↗

Key Technologies for Ultra High Dose CMOS Applications

The trend towards shrinking advanced microelectronic Logic and DRAM devices will require ultra high dose implantation. One ultra high dose application in DRAM, being rapidly adopted in production is Dual Poly Gate (DPG). Three main challenges existed for the adoption of this high dose dual poly gate (DPG) doping applications: monitoring of high dose implantation, photoresist stripping and maintaining high throughput. In this paper we present how these challenges have been addressed. VSEA's plasma doping (PLAD) tool offers several unique advantages for DPG applications. When compared to conventional or molecular beam line implanters or other immersion techniques, PLAD delivers 3 to 7 times higher throughput (compared to traditional ion implanter) without dopant penetration through the thin doped polysilicon layer into the gate oxide. It also improves P{sup +} poly silicon DPG device properties at superior throughput. In this work we demonstrate how hot spray photoresist strip processing eliminates the need for multiple-tools required for wet+ash+wet process. In addition to PLAD's patented in-situ dose control metrology we also demonstrate an ex-situ high dose implantation metrology using spectroscopic ellipsometer (SE) and spectroscopic reflectometer (SR). The technique shows good correlation (R{sup 2}{approx}0.99) between implant dose and damaged layer thickness.

75 CONDENSED MATTER PHYSICS, SUPERCONDUCTIVITY AND↗

Improvement of Contact Resistance with Molecular Ion Implantation

Basic characteristics of ClusterBoron (B{sub 18}H{sub 22}) implantation were investigated for improving contact resistance in DRAM devices. Generally, {sup 49}BF{sub 2} has been widely used for contact implant application in DRAM manufacturing because of its higher productivity compared to monomer boron ({sup 11}B). However, because of limited activation in a low thermal budget ({approx}800 deg. C) anneal, the sheet resistance was saturated for doses over 5x10{sup 15} ions/cm{sup 2}. Although many investigations have been reported, such as {sup 30}BF implant mixed implant with monomer boron etc., no practical solution has been found for dramatic improvement of contact resistance in a productive manner. B{sub 18}H{sub 22} was developed to overcome the productivity limitations encountered in low energy, high dose boron implantation and the limited activation of {sup 49}BF{sub 2} due to co-implanted fluorine. In this study, basic characterization of the B{sub 18}H{sub 22} contact implant was performed through sheet resistance, SIMS (Secondary Ion Mass Spectrometry) and XTEM (cross-sectional transmission electron microscopy). The B{sub 18}H{sub 22} implants showed lower sheet resistance than conventional {sup 49}BF{sub 2} for 5x10{sup 15} ions/cm{sup 2} on bare wafer tests. Through XTEM study, we found the activation behavior of both B{sub 18}H{sub 22} and {sup 49}BF{sub 2} were directly related with the amorphous layer thickness and residual defects from low thermal budget anneal. PMOS contact resistance in the sub-70 nm device by B{sub 18}H{sub 22} implantation showed considerable improvement (about 30%), showing B{sub 18}H{sub 22} could replace the BF{sub 2} for contact implant in contact resistance implant.

36 MATERIALS SCIENCE↗

Radiation Source Localization Using Surrogate Models Constructed from 3-D Monte Carlo Transport Physics Simulations

Recent research has focused on the development of surrogate models for radiation source localization in a simulated urban domain. We employ the Monte Carlo N-Particle (MCNP) code to provide high- delity simulations of radiation transport within an urban domain. The model is constructed to employ a source location (x, y, z) as input and return the estimated count rate for a set of speci ed detector locations. Because MCNP simulations are computationally expensive, we develop e cient and accurate surrogate models of the detector responses. We construct surrogate models using Gaussian processes (GP) and neural networks (NN) that we train and verify using the MCNP simulations. The trained surrogate models provide an e cient framework for Bayesian inference and experimental design. We employ Delayed Rejection Adaptive Metropolis (DRAM), a Markov Chain Monte Carlo (MCMC) algorithm, to infer the location and intensity of an unknown source. The DRAM results yield a posterior probability distribution for the source's location conditioned on the observed detector count rates. The posterior distribution exhibits regions of high and low probability within the simulated environment identifying potential source locations. In this manner, we can quantify the source location to within at least one of these regions of high probability in the considered cases. Employing these methods, we are able to reduce the space of potential source locations by at least 60%.

46 INSTRUMENTATION RELATED TO NUCLEAR SCIENCE AND ↗

Exploring Memory Hierarchy to Improve Scientific Data Read Performance, In: 2015 IEEE International Conference on Cluster Computing

Improving read performance is one of the major challenges with speeding up scientific data analytic applications. Utilizing the memory hierarchy is one major line of researches to address the read performance bottleneck. Related methods usually combine solide-state-drives(SSDs) with dynamic random-access memory(DRAM) and/or parallel file system(PFS) to mitigate the speed and space gap between DRAM and PFS. However, these methods are unable to handle key performance issues plaguing SSDs, namely read contention that may cause up to 50% performance reduction. In this paper, we propose a framework that exploits the memory hierarchy resource to address the read contention issues involved with SSDs. The framework employs a general purpose online read algorithm that able to detect and utilize memory hierarchy resource to relieve the problem. To maintain a near optimal operating environment for SSDs, the framework is able to orchastrate data chunks across different memory layers to facilitate the read algorithm. Compared to existing tools, our framework achieves up to 50% read performance improvement when tested on datasets from real-world scientific simulations.

Computer Science↗

Variation Tolerant and Energy-Efficient Charge Domain Compute-in-Memory Array with Binary and Multi-Level Cell Ferroelectric FET

Here, in this work, we present a variation-tolerant and energy-efficient charge-domain Ferroelectric FET (FeFET) based Compute-in-Memory (CiM) array design that is compatible with both binary and multi-level cell memory sensing. We demonstrate that: 1) by exploiting FeFET as a nonvolatile switch, its high ON/OFF ratio in the subthreshold region can suppress the error introduced by the inaccurate ON state conductance, thus realizing robust CiM operations, unlike the current-domain CiM design where the computation results is highly sensitive to the device conductance variation; 2) by leveraging a dense dynamic random access memory (DRAM)-like 1FeFET1C cell structure, the proposed design benefits from the existing high density DRAM establishment while also significantly relaxing the capacitor retention and transistor leakage requirement; 3) the charge-domain CiM supports both binary FeFET with minimum overhead and MLC FeFET with tolerable latency for MLC state sensing, whose efficacy is validated experimentally on both cell-level and array-level; 4) the proposed CiM shows much better device variation resilience than conventional current-domain CiM, and also improves inference accuracy. Macro-level evaluation results demonstrate significantly higher energy efficiency and area efficiency compared to prior CiM works.

Duan, Jiahui [University of Notre Dame, IN (United↗

NVMalloc: Exposing an Aggregate SSD Store as a Memory Partition in Extreme-Scale Machines

DRAM is a precious resource in extreme-scale machines and is increasingly becoming scarce, mainly due to the growing number of cores per node. On future multi-petaflop and exaflop machines, the memory pressure is likely to be so severe that we need to rethink our memory usage models. Fortunately, the advent of non-volatile memory (NVM) offers a unique opportunity in this space. Current NVM offerings possess several desirable properties, such as low cost and power efficiency, but suffer from high latency and lifetime issues. We need rich techniques to be able to use them alongside DRAM. In this paper, we propose a novel approach for exploiting NVM as a secondary memory partition so that applications can explicitly allocate and manipulate memory regions therein. More specifically, we propose an NVMalloc library with a suite of services that enables applications to access a distributed NVM storage system. We have devised ways within NVMalloc so that the storage system, built from compute node-local NVM devices, can be accessed in a byte-addressable fashion using the memory mapped I/O interface. Our approach has the potential to re-energize out-of-core computations on large-scale machines by having applications allocate certain variables through NVMalloc, thereby increasing the overall memory capacity available. Our evaluation on a 128-core cluster shows that NVMalloc enables applications to compute problem sizes larger than the physical memory in a cost-effective manner. It can bring more performance/efficiency gain with increased computation time between NVM memory accesses or increased data access locality. In addition, our results suggest that while NVMalloc enables transparent access to NVM-resident variables, the explicit control it provides is crucial to optimize application performance.

Computer Science↗

Co-design of Advanced Architectures for Graph Analytics using Machine Learning

A graph is an excellent way of representing relationships among entities. We can use graph analytics to synthesize and analyze such relational data, and extract relevant features that are useful for various tasks such as machine learning. Considering the crucial role of graph analytics in various domains, it is important and timely to investigate the right hardware configurations that can achieve optimal performance for graph workloads on future high-performance computing systems. Design space exploration studies facilitate the selection of appropriate configurations (e.g. memory) to achieve a desired system performance. Recently, the approach of accelerating graph analytics using persistent non-volatile memory has gained a lot of attention. Traditional system simulators such as Gem5 and NVMain can be used to explore the design space of these advanced memory architectures for graph workloads. However, these simulators are slow in execution thus limiting the efficiency of design space exploration studies. To overcome this challenge, we proposed a machine learning based approach to co-design advanced memory architectures for graph workloads. We tested our approach with DRAM, non-volatile memory, and hybrid memory (DRAM+NVM) using a breadth first search benchmark algorithm. Our results showed the applicability of the proposed machine learning based approach to the co-design of the advanced memory architectures. In this paper, we provide recommendations on selecting advanced memory architectures to achieve desired performance for graph workloads. We also discuss the performances of different machine learning models that were considered in this study.

Kurte, Kuldeep↗

Machine learning based design space exploration for hybrid main-memory design

We develop a machine learning (ML) based design space exploration (DSE) method that builds predictive models for various responses of a hybrid main-memory system. To overcome the challenges associated with latency, capacity, and power of memory systems in future extreme-scale machines, the hybrid memory architectures are being considered in which novel non-volatile memory (NVM) systems augment the traditional DRAM. However, way before their actual design and implementation, these emerging hybrid memory systems need to be simulated and analyzed to fully understand their capabilities and limitations. As the conventional architectural-level memory simulators require significant amounts of computational costs and time, we propose to utilize ML techniques for developing various memory-response models that can instantly provide a predicted response corresponding to any new memory configuration. Specifically, in this work, we apply four supervised ML techniques to build regression models for memory latency, bandwidth, power, and total read/write responses. The training and validation data for the ML methods are generated using NVMain memory simulator for DRAM, NVM, and their hybrid combinations. We demonstrate the results of the ML based memory-DSE method in terms of the learning curve characteristics for hyperparameter tuning and the statistical error analyses of the designed predictive models.

Sen, Satyabrata↗

3D-ReG: A 3D ReRAM-based Heterogeneous Architecture for Training Deep Neural Networks

Deep neural network (DNN) models are being expanded to a broader range of applications. The computational capability of traditional hardware platforms cannot accommodate the growth of model complexity. Among recent technologies to accelerate DNN, resistive memory (ReRAM)-based processing-in-memory (PIM) emerged as a promising solution for DNN inference due to its high efficiency for matrix-based computation. We face two major technical challenges in extending the use of ReRAM-based accelerators for training: (1) full-precision data is essential in back-propagation; (2) the need to support both feed-forward and back-propagation aggravates the data-movement burden. We propose a heterogeneous architecture named as 3D-ReG, which leverages full-precision GPU to ensure training accuracy and low-overhead 3D integration to provide low-cost data movements. Moreover, we introduce conservative and aggressive task-mapping schemes, which partition the computation phases in different ways to balance execution efficiency and training accuracy. We evaluate 3D-ReG implemented with two 3D integration technologies, through-silicon vias (TSVs) and monolithic inter-tier vias (MIVs), and compare them with GPU-only and PIM-only counterparts. Various GPU-only platforms using two main-memory technologies (DRAM, ReRAM) and three interconnect technologies (2D, TSV, MIV) are evaluated as well. Experimental results show that 3D-ReG can achieve on average 5.64× training speedup and 3.56× higher energy efficiency compared with the GPU with DRAM as main memory, at the cost of 0.05%–3.39% accuracy drop. We define a new metric, gain-loss ratio (GLR), which quantitatively evaluates the capability of a DNN training hardware in terms of the model accuracy and hardware efficiency. The results of our comparison show that the aggressive task-mapping scheme on MIV-based 3D-ReG outperforms the other methods.

Computer Science↗

Rutger's CAM2000 chip architecture

This report describes the architecture and instruction set of the Rutgers CAM2000 memory chip. The CAM2000 combines features of Associative Processing (AP), Content Addressable Memory (CAM), and Dynamic Random Access Memory (DRAM) in a single chip package that is not only DRAM compatible but capable of applying simple massively parallel operations to memory. This document reflects the current status of the CAM2000 architecture and is continually updated to reflect the current state of the architecture and instruction set.

Smith, Donald E.↗

Wide-bandwidth high-resolution search for extraterrestrial intelligence

A third antenna was added to the system. It is a terrestrial low-gain feed, to act as a veto for local interference. The 3-chip design for a 4 megapoint complex FFT was reduced to finished working hardware. The 4-Megachannel circuit board contains 36 MByte of DRAM, 5 CPLDs, the three large FFT ASICs, and 74 ICs in all. The Austek FDP-based Spectrometer/Power Accumulator (SPA) has now been implemented as a 4-layer printed circuit. A PC interface board has been designed and together with its associated user interface and control software allows an IBM compatible computer to control the SPA board, and facilitates the transfer of spectra to the PC for display, processing, and storage. The Feature Recognizer Array cards receive the stream of modulus words from the 4M FFT cards, and forward a greatly thinned set of reports to the PC's in whose backplane they reside. In particular, a powerful ROM-based state-machine architecture has been adopted, and DRAM has been added to permit integration modes when tracking or reobserving source candidates. The general purpose (GP) array consists of twenty '486 PC class computers, each of which receives and processes the data from a feature extractor/correlator board set. The array performs a first analysis on the provided 'features' and then passes this information on to the workstation. The core workstation software is now written. That is, the communication channels between the user interface, the backend monitor program and the PC's have working software.

Horowitz, Paul↗

Combining Video Memory Operations

Designs of video random-access memory (VRAM) integrated circuits operating under control by external logic circuits simplified according to concept of combining two memory operations performed separately heretofore. Eliminates need for DRAM-refresh timers and counters, reducing amount of circuitry needed to control VRAM thereby reducing time needed to design VRAM. Simplification also reduces time needed to redesign DRAM-refresh logic circuitry when adapting VRAM design to another VRAM for which timing specifications different. Concept can be applied to VRAM clocking data out to display unit continuously.

Kania, Michael J.↗

The Impact on Space Radiation Requirements and Effects on ASIMS

The evolution of highly miniaturized electronic and mechanical systems will be accompanied by new problems and issues regarding the radiation response of these systems in the space environment. In this paper we discuss some of the more prominent radiation problems brought about by miniaturization. For example, autonomous micro-spacecraft will require large amounts of high density memory, most likely in the form of stacked, multichip modules of DRAM's, that must tolerate the radiation environment. However, advanced DRAM's (16 to 256 Mbit) are quite susceptible to radiation, particularly single event effects, and even exhibit new radiation phenomena that were not a problem for older, less dense memory chips. Another important trend in micro-spacecraft electronics is toward the use of low-voltage microelectronic systems that consume less power. However, the reduction in operating voltage also caries with it an increased susceptibility to radiation. In the case of application specific integrated microcircuits (ASIM's), advanced devices of this type, such as high density field programmable gate arrays (FPGA's) exhibit new single event effects (SEE), such as single particle reprogramming of anti-fuse links. New advanced bipolar circuits have been shown recently to degrade more rapidly in the low dose rate space environment than in the typical laboratory total dose radiation test used to qualify such devices. Thus total dose testing of these parts is no longer an appropriately conservative measure to be used for hardness assurance. We also note that the functionality of micromechanical Si-based devices may be altered due to the radiation-induced deposition of charge in the oxide passivation layers.

Barnes, C.↗