Modular High-Performance Computing Using Chiplets

IF 1.8 4区计算机科学 Q3 COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS

Computing in Science & Engineering Pub Date : 2024-06-20 DOI:10.1109/mcse.2023.3341749

Bapi Vinnakota, John M. Shalf

{"title":"Modular High-Performance Computing Using Chiplets","authors":"Bapi Vinnakota, John M. Shalf","doi":"10.1109/mcse.2023.3341749","DOIUrl":null,"url":null,"abstract":"The performance growth rate of high-performance computing (HPC) systems has fallen from 1000× to just 10× every eleven years. The HPC world, like large cloud service provider data centers, has turned to heterogeneous acceleration to deliver continued performance growth through specialization. Chiplets offer a new, compelling approach to scaling performance through adding workload-specific processors and massive bandwidth to memory into computing systems. If design and manufacturing challenges are resolved, chiplets can offer a cost-effective path for combining die from multiple function-optimized process nodes, and even from multiple vendors, into a single application-specific integrated circuit (ASIC). This article explores opportunities for building and improving the performance of bespoke HPC architectures using open-modular “chiplet” building blocks. The hypothesis developed is to use chiplets to extend the functional and physical modularity of modern HPC systems to within the semiconductor package. This planning can reduce the complexity and cost of assembling chiplets into an ASIC product and make it easier to build multiple product variants.","PeriodicalId":10553,"journal":{"name":"Computing in Science & Engineering","volume":"117 1","pages":""},"PeriodicalIF":1.8000,"publicationDate":"2024-06-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computing in Science & Engineering","FirstCategoryId":"94","ListUrlMain":"https://doi.org/10.1109/mcse.2023.3341749","RegionNum":4,"RegionCategory":"计算机科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q3","JCRName":"COMPUTER SCIENCE, INTERDISCIPLINARY APPLICATIONS","Score":null,"Total":0}

引用次数: 0

Abstract

The performance growth rate of high-performance computing (HPC) systems has fallen from 1000× to just 10× every eleven years. The HPC world, like large cloud service provider data centers, has turned to heterogeneous acceleration to deliver continued performance growth through specialization. Chiplets offer a new, compelling approach to scaling performance through adding workload-specific processors and massive bandwidth to memory into computing systems. If design and manufacturing challenges are resolved, chiplets can offer a cost-effective path for combining die from multiple function-optimized process nodes, and even from multiple vendors, into a single application-specific integrated circuit (ASIC). This article explores opportunities for building and improving the performance of bespoke HPC architectures using open-modular “chiplet” building blocks. The hypothesis developed is to use chiplets to extend the functional and physical modularity of modern HPC systems to within the semiconductor package. This planning can reduce the complexity and cost of assembling chiplets into an ASIC product and make it easier to build multiple product variants.

查看原文本刊更多论文

使用芯片组的模块化高性能计算

高性能计算（HPC）系统的性能增长率已从 1000 倍下降到每 11 年仅 10 倍。与大型云服务提供商数据中心一样，高性能计算领域已转向异构加速，通过专业化实现性能的持续增长。通过在计算系统中添加工作负载专用处理器和大规模内存带宽，Chiplet 提供了一种全新的、引人注目的性能扩展方法。如果能解决设计和制造方面的难题，芯片组就能提供一条具有成本效益的途径，将来自多个功能优化工艺节点，甚至来自多个供应商的芯片组合成单一的专用集成电路（ASIC）。本文探讨了使用开放模块化 "芯片组 "构件构建和提高定制高性能计算架构性能的机会。本文提出的假设是利用芯片将现代高性能计算系统的功能和物理模块化扩展到半导体封装内。这种规划可以降低将芯片组组装到 ASIC 产品中的复杂性和成本，并使构建多种产品变体变得更加容易。

本文章由计算机程序翻译，如有差异，请以英文原文为准。

求助全文

约1分钟内获得全文求助全文

来源期刊

Computing in Science & Engineering 工程技术-计算机：跨学科应用

CiteScore

4.20

自引率

0.00%

发文量

审稿时长

6-12 weeks

期刊介绍： Physics, medicine, astronomy -- these and other hard sciences share a common need for efficient algorithms, system software, and computer architecture to address large computational problems. And yet, useful advances in computational techniques that could benefit many researchers are rarely shared. To meet that need, Computing in Science & Engineering presents scientific and computational contributions in a clear and accessible format. The computational and data-centric problems faced by scientists and engineers transcend disciplines. There is a need to share knowledge of algorithms, software, and architectures, and to transmit lessons-learned to a broad scientific audience. CiSE is a cross-disciplinary, international publication that meets this need by presenting contributions of high interest and educational value from a variety of fields, including—but not limited to—physics, biology, chemistry, and astronomy. CiSE emphasizes innovative applications in advanced computing, simulation, and analytics, among other cutting-edge techniques. CiSE publishes peer-reviewed research articles, and also runs departments spanning news and analyses, topical reviews, tutorials, case studies, and more.