George Liang-Tai Chiu

Overview of the Blue Gene/L system architecture

The Blue Gene®/L computer is a massively parallel supercomputer based on IBM system-on-a-chip technology. It is designed to scale to 65,536 dual-processor nodes, with a peak performance of 360 teraflops. This paper describes the project objectives and provides an overview of the system architecture that resulted. We discuss our application-based approach and rationale for a low-power, highly integrated design. The key architectural features of Blue Gene/L are introduced in this paper: the link chip component and five Blue Gene/L networks, the PowerPC® 440 core and floating-point enhancements, the on-chip and off-chip distributed memory system, the node- and system-level design for high reliability, and the comprehensive approach to fault isolation.

The IBM Blue Gene/Q Compute Chip

Blue Gene/Q aims to build a massively parallel high-performance computing system out of power-efficient processor chips, resulting in power-efficient, cost-efficient, and floor-space- efficient systems. Focusing on reliability during design helps with scaling to large systems and lowers the total cost of ownership. This article examines the architecture and design of the Compute chip, which combines processors, memory, and communication functions on a single chip.

Noncontact high-speed waveform measurements with the picosecond photoelectron scanning electron microscope

The authors have replaced the electron gun and beam blanking system of a conventional voltage contrast scanning electron microscope by a pulse-laser/photocathode combination, resulting in a source producing electron pulses of order 1 ps in duration at a 100 MHz repetition rate and with a peak brightness of 3 10/sup 8/ A/cm/sup 2/ sr at 1.8 keV. This novel instrument has demonstrated stroboscopic noncontact waveform measurements on metal interconnect lines in different environments with a temporal resolution between than 5 ps, a voltage resolution of 3 mV/(Hz)/sup 1/2/, and a spatial resolution of 0.1 mu m. These measurements are achieved with extraction fields above the sample of about 1 kV/mm. >

Picosecond photoelectron scanning electron microscope for noncontact testing of integrated circuits

A scanning electron microscope which uses an ultrashort pulsed laser/photocathode combination as an electron source produces electron pulses of order 1 ps in duration at a 100‐MHz repetition rate and with a peak brightness of 3×108 A/cm2 steradian at 1.8 keV. By using this instrument in the voltage contrast mode, without contact with the samples, we have been able to measure electrical pulses propagating on coplanar transmission lines with, simultaneously, a temporal resolution of 5 ps, a voltage resolution of 3 mV/(Hz)1/2, and a spatial resolution of 0.1 μm. These measurements are achieved with extraction fields above the sample of about 1 kV/mm.

Projection display throughput: efficiency of optical transmission and light-source collection

The optical system for a projection display based on three miniature reflective spatial light modulators (SLMs) is described. The total projection display light throughput is a function not only of the optical system efficiency but also of the light-collection and light-coupling efficiency referred to here as the lamp-SLM coupling. The optical system efficiency is the transmission of the optical components in the projection display. These are examined in detail through measurements and estimates of the components in the system. The various optical components include UV-IR filtering, illumination optics, polarization optics, color separation and recombination optics, SLM efficiency, and projection optics. The lamp-SLM coupling, which is the amount of usable light that can be collected from a particular lamp coupled to the projection optical system, is determined by the light-source luminance, the efficiency of the light-collection optics, and the optical system etendue. For small SLMs, less than 50 mm diagonal, for example, the lamp-SLM coupling efficiency falls off rapidly with SLM size and optical system f-number. The dependence of this coupling efficiency on SLM size is determined from measurements of the light-collection efficiency as a function of aperture size, where the apertures are used to simulate SLMs of the same dimensions. A variety of arc lamps were investigated for use in the projection display based on IBM reflective SLM devices. The lamp-SLM coupling dependence on arc gap was determined. The measurements are used to compare various lamps and to estimate directly the throughput for the complete projection system. The SLMs used in the projection display are liquid crystal devices which utilize only one polarization of light while discarding the second. Converting the discarded polarization into useful light can in principle double the throughput of the projector. However, polarization conversion results in doubling of the size of the light source and thus produces less efficient lamp-SLM coupling, particularly for long-arc-gap lamps. Measurements and analysis of throughput enhancement by polarization conversion are presented, and the dependence on arc gap and optical system etendue is discussed.

High-numerical-aperture optical designs

This paper is an overview of the designs of high-numerical-aperture lenses for optical projection lithography at the IBM Thomas J. Watson Research Center.

Overview of optical interconnect technology

High performance computers have been striving for higher speed better connectivity and higher throughput since their inception. Continuous advancement in the performance of active devices is placing an increasingly heavier demand on passive interconnects. This paper reviews optical interconnect technology in light of recent developments and suggests opportunities in datacom applications.

The IBM Blue Gene project

This paper provides a short overview of the IBM Blue Gene® project and an introduction to all of the papers in this issue of the IBM Journal of Research and Development.

Tracking the Performance Evolution of Blue Gene Systems

IBM’s Blue Gene supercomputer has evolved through three generations from the original Blue Gene/L to P to Q. A higher level of integration has enabled greater single-core performance, and a larger concurrency per compute node. Although these changes have brought with them a higher overall system peak-performance, no study has examined in detail the evolution of performance across system generations. In this work we make two significant contributions – that of providing a comparative performance analysis across Blue Gene generations using a consistent set of tests, and also in providing a validated performance model of the NEK-Bone proxy application. The combination of empirical analysis and the predictive performance model enable us to not only directly compare measured performance but also allow for a comparison of system configurations that cannot currently be measured. We provide insights into how the changing characteristics of Blue Gene have impacted on the application performance, as well as what future systems may be able to achieve.

Ultrashort Electron-pulse Probing of Integrated Circuits

The performance of devices and circuits is advancing at a rapid pace with the advent of submicron design ground rules and sub-50 ps switching times. The requirements to probe the internal nodes of these ultrafast, ultrasmall, and ultradense circuits give rise to great challenges for high-speed electron-beam testing. In this paper, we review the steps which have allowed electron beam testing to achieve simultaneously: 5 ps temporal resolution, 0·1 wm spot size and 3 mV/ iHz voltage sensitivity. The resulting newly developed instrument, called the picosecond photoelectron scanning electron microscope (PPSEM), is capable of measuring the state-of-the-art bipolar and FET circuits, as shown in this paper.