Link to the University of Pittsburgh Homepage
Link to the University Library System Homepage Link to the Contact Us Form

Multi-port Memory Design for Advanced Computer Architectures

Zhao, Yirong (2013) Multi-port Memory Design for Advanced Computer Architectures. Master's Thesis, University of Pittsburgh. (Unpublished)

This is the latest version of this item.

[img]
Preview
PDF
Primary Text

Download (1MB) | Preview

Abstract

In this thesis, we describe and evaluate novel memory designs for multi-port on-chip and off-chip use in advanced computer architectures. We focus on combining multi-porting and evaluating the performance over a range of design parameters. Multi-porting is essential for caches and shared-data systems, especially multi-core System-on-chips (SOC). It can significantly increase the memory access throughput. We evaluate FinFET voltage-mode multi-port SRAM cells using different metrics including leakage current, static noise margin and read/write performance. Simulation results show that single-ended multi-port FinFET SRAMs with isolated read ports offer improved read stability and flexibility over classical double-ended structures at the expense of write performance. By increasing the size of the
access transistors, we show that the single-ended multi-port structures can achieve equivalent write performance to the classical double-ended multi-port structure for 9% area overhead. Moreover, compared with CMOS SRAM, FinFET SRAM has better stability and standby power. We also describe new methods for the design of FinFET current-mode multi-port
SRAM cells. Current-mode SRAMs avoid the full-swing of the bitline, reducing dynamic power and access time. However, that comes at the cost of voltage drop, which compromises
stability. The design proposed in this thesis utilizes the feature of Independent Gate (IG) mode FinFET, which can leverage threshold voltage by controlling the back gate voltage, to merge two transistors into one through high-Vt and low-Vt transistors. This design not only reduces the voltage drop, but it also reduces the area in multi-port current-mode SRAM design. For off-chip memory, we propose a novel two-port 1-read, 1-write (1R1W) phasechange memory (PCM) cell, which significantly reduces the probability of blocking at the bank levels. Different from the traditional PCM cell, the access transistors are at the top and connected to the bitline. We use Verilog-A to model the behavior of Ge2Sb2Te5 (GST: the storage component). We evaluate the performance of the two-port cell by transistor
sizing and voltage pumping. Simulation results show that pMOS transistor is more practical than nMOS transistor as the access device when both area and power are considered. The estimated area overhead is 1.7�, compared to single-port PCM cell. In brief, the contribution we make in this thesis is that we propose and evaluate three different kinds of multi-port memories that are favorable for advanced computer architectures.


Share

Citation/Export:
Social Networking:
Share |

Details

Item Type: University of Pittsburgh ETD
Status: Unpublished
Creators/Authors:
CreatorsEmailPitt UsernameORCID
Zhao, Yirongyiz52@pitt.eduYIZ52
ETD Committee:
TitleMemberEmail AddressPitt UsernameORCID
Committee ChairMohanram, Kartikkmram@pitt.eduKMRAM
Committee MemberLevitan, Stevenlevitan@pitt.eduLEVITAN
Committee MemberJones, Alexakjones@pitt.eduAKJONES
Committee MemberLi, Helenhal66@pitt.eduHAL66
Date: 24 September 2013
Date Type: Publication
Defense Date: 23 July 2013
Approval Date: 24 September 2013
Submission Date: 26 July 2013
Access Restriction: No restriction; Release the ETD for access worldwide immediately.
Number of Pages: 59
Institution: University of Pittsburgh
Schools and Programs: Swanson School of Engineering > Electrical Engineering
Degree: MS - Master of Science
Thesis Type: Master's Thesis
Refereed: Yes
Uncontrolled Keywords: Multi-port FinFET current-mode SRAM two-port PCM
Date Deposited: 24 Sep 2013 20:04
Last Modified: 15 Nov 2016 14:14
URI: http://d-scholarship.pitt.edu/id/eprint/19516

Available Versions of this Item


Metrics

Monthly Views for the past 3 years

Plum Analytics


Actions (login required)

View Item View Item