How fast are Supercomputers ? Covering: FLOPS, How fast?, TOP500

Slides:



Advertisements
Similar presentations
Clusters, Grids and their applications in Physics David Barnes (Astro) Lyle Winton (EPP)
Advertisements

Michael Brost. History Theory Examples Future Man communicating with Man Man communicating with Machine
Challenge: Computing for the Life Sciences Gaining insight into protein science and protein folding mechanisms Simulating (thousands of) macromolecules.
 The k-computer is the world’s fastest supercomputer.  The k-computer named for the japnese word “kei" which stands for 10 quardillion.  K-computer.
Super Computers By Phuong Vo.
Andrew Canning and Lin-Wang Wang Computational Research Division LBNL
1 Computational models of the physical world Cortical bone Trabecular bone.
Parallel computer architecture classification
Istituto Tecnico Industriale "A. Monaco"
IBM aims to reclaim supercomputer title By Jatin Chopra.
PRIYADHARSHINI S SUPERCOMPUTERS. OVERVIEW  The term is commonly applied to the fastest high-performance systems in existence at the time of their construction.
Commodity Computing Clusters - next generation supercomputers? Paweł Pisarczyk, ATM S. A.
Last Lecture The Future of Parallel Programming and Getting to Exascale 1.
BY MANISHA JOSHI.  Extremely fast data processing-oriented computers.  Speed is measured in “FLOPS”.  For highly calculation-intensive tasks.  For.
SUPERCOMPUTERS A Brief History by Jenny Grant. What is a Supercomputer ? Definition: – Any computer able to process information at the highest capacity.
This project and the research leading to these results has received funding from the European Community's Seventh Framework Programme.
SUPERCOMPUTERS By: Cooper Couch. WHAT IS A SUPERCOMPUTER? In the most Basic sense a supercomputer is one, that is at the forefront of modern processing.
Parallel Processing1 Parallel Processing (CS 676) Overview Jeremy R. Johnson.
ASU/TGen Computational Facility.
Claude TADONKI Mines ParisTech – LAL / CNRS / INP 2 P 3 University of Oujda (Morocco) – October 7, 2011 High Performance Computing Challenges and Trends.
CS 240A Applied Parallel Computing John R. Gilbert Thanks to Kathy Yelick and Jim Demmel at UCB for.
MD240 - Management Information Systems Sept. 13, 2005 Computing Hardware – Moore's Law, Hardware Markets, and Computing Evolution.
Supercomputers Daniel Shin CS 147, Section 1 April 29, 2010.
The Problem With The Linpack Benchmark 1.0 Matrix Generator Jack J. Dongarra and Julien Langou International Journal of High Performance Computing Applications.
Pop Quiz! What new technology has the United States and Japan engaged in the virtual equivalent of the space race?
Top500: Red Storm An abstract. Matt Baumert 04/22/2008.
UNL Computer Science & Engineering Cluster Computing David R. Swanson Beowulf and Bombs.
Arquitectura de Sistemas Paralelos e Distribuídos Paulo Marques Dep. Eng. Informática – Universidade de Coimbra Ago/ Machine.
Lecture 1: Introduction to High Performance Computing.
Scientific Computing on Smartphones David P. Anderson Space Sciences Lab University of California, Berkeley April 17, 2014.
1 CHAPTER 2 COMPUTER HARDWARE. 2 The Significance of Hardware  Pace of hardware development is extremely fast. Keeping up requires a basic understanding.
ICPP 2014 Keynotes Summary 09/24. Data Centric Systems: The Next Paradigm in Computing Speaker: Dr. Tilak Agerwala ◦ Vice President, Data Centric Systems.
© Fujitsu Laboratories of Europe 2009 HPC and Chaste: Towards Real-Time Simulation 24 March
ICT in Weather Forecasting
Supercomputers Special Course of Computer Architecture H.Amano.
Copyright © The McGraw-Hill Companies, Inc. Permission required for reproduction or display. Parallel Programming in C with MPI and OpenMP Michael J. Quinn.
Cross Council ICT Conference May High Performance Computing Ron Perrott Chairman High End Computing Strategy Committee Queen’s University Belfast.
Unit 1 Unit 1 An overview of Computer System Classification of Computers.
Introduction to Computing: Lecture 2 Computer Systems.
Welcome to KISTI KISTI-CCIN2P3 FKPPL Workshop December 1, 2008 Minsun LEE.
- Rohan Dhamnaskar. Overview  What is a Supercomputer  Some Concepts  Couple of examples.
Computing Environment The computing environment rapidly evolving ‑ you need to know not only the methods, but also How and when to apply them, Which computers.
October 12, 2004Thomas Sterling - Caltech & JPL 1 Roadmap and Change How Much and How Fast Thomas Sterling California Institute of Technology and NASA.
CS 240A Applied Parallel Computing John R. Gilbert Thanks to Kathy Yelick and Jim Demmel at UCB for.
2009/4/21 Third French-Japanese PAAP Workshop 1 A Volumetric 3-D FFT on Clusters of Multi-Core Processors Daisuke Takahashi University of Tsukuba, Japan.
A look at computing performance and usage.  3.6GHz Pentium 4: 1 GFLOPS  1.8GHz Opteron: 3 GFLOPS (2003)  3.2GHz Xeon X5460, quad-core: 82 GFLOPS.
CLIM Fall 2008 What are the Roles of Satellites & Supercomputers in Studying Weather and Climate? CLIM 101.
Computing Environment The computing environment rapidly evolving ‑ you need to know not only the methods, but also How and when to apply them, Which computers.
CS 240A Applied Parallel Computing John R. Gilbert Thanks to Kathy Yelick and Jim Demmel at UCB for.
CS591x -Cluster Computing and Parallel Programming
Today's Software For Tomorrow's Hardware: An Introduction to Parallel Computing Rahul.S. Sampath May 9 th 2007.
Parallel Computers Today Oak Ridge / Cray Jaguar > 1.75 PFLOPS Two Nvidia 8800 GPUs > 1 TFLOPS Intel 80- core chip > 1 TFLOPS  TFLOPS = floating.
David P. Anderson Space Sciences Laboratory University of California – Berkeley A Million Years of Computing.
Parallel Computers Today LANL / IBM Roadrunner > 1 PFLOPS Two Nvidia 8800 GPUs > 1 TFLOPS Intel 80- core chip > 1 TFLOPS  TFLOPS = floating point.
Presented by Leo Pleše ScienceUp.org
Computer Operations Part 2.
Parallel computer architecture classification
Super Computing By RIsaj t r S3 ece, roll 50.
TOP-10-Destinations German Exports - Plastics and Rubber Machinery
32nd TOP500 List SC08, Austin, TX.
Parallel Computers Today
Physics-based simulation for visual computing applications
Course Description: Parallel Computer Architecture
with Computational Scientists
$1M a year for 5 years; 7 institutions Active:
Pioneering the Computing & Communication Services for Academic Studies
The C&C Center Three Major Missions: In This Presentation:
Benchmark software for HPC systems
HISTORY OF THE EARTH ANSWER SHEET
CSE 102 Introduction to Computer Engineering
Presentation transcript:

How fast are Supercomputers ? Covering: FLOPS, How fast?, TOP500 Dr. Show’ichi Furuyama National Institute of Technology, Toyama College, Japan Good morning everyone. Thank you all for coming to my lecture again. Today we will about the abilities of super computers. “How fast are Supercomputers?” is today’s title. Please keep your questions and comment until the end of my lecture. We will be covering three main points today. First is FLOPS, Second is “how fast?” and third is TOP 500. Have you ever heard “supercomputer” ? And do you know What/Where/How fast is the fastest supercomputer in the world? By the end of my talk, you’ll understand these topics. OK, let’s get started.

How many “+” and “-” can you find? Please answer the following calculation ! and please measure your calculation time! 2.0 - 1.9 - 6.3 + 9.4 = 3.0 - 4.5 + 1.2 + 2.3 = 1.3 + 0.9 - 6.2 + 3.0 - 4.6 = At first, I would like you to answer this easy calculation. I prepared three questions. Please measure your calculation time by your watch. Ready, set, go ! (30 seconds waiting…) Anyone finished? Ok next. Please count “+” and “-” operations in the calculation? The no.1 has three operations. How many operations do you find in three equations ? 10 ! that’s right! ---- Well done / nice work / excellent / How many “+” and “-” can you find?

How many operations could you calculate in one second ? 𝟏𝟎 [𝒐𝒑𝒆𝒓𝒂𝒕𝒊𝒐𝒏𝒔] 𝑪𝒂𝒍𝒄. 𝒕𝒊𝒎𝒆 [𝒔𝒆𝒄] = ?? FLoating-point Operations Per Second (FLOPS) = 𝑵. 𝒐𝒇 𝑶𝒑𝒆𝒓𝒂𝒕𝒊𝒐𝒏𝒔 𝑪𝒂𝒍𝒄. 𝒕𝒊𝒎𝒆 [𝒔𝒆𝒄] +, −, ×, ÷ Your next task is “How many operations could you calculate in one second?” You can calculate it by the equations. What is your result? The numeric value is called FLOPS. Flops means Floating-point operations per second. Floating-point operation means operation for decimal fraction. In the case they are plus, minus, multiply and division. In short, it is how many operations computer can calculate. The value is indicator of processing ability of computers. Anyway, how much processing ability of computer is required for a simulation?

Inside of Ocean Simulation Making computational cells Calculate equations in all cells 𝑎 11 𝒙 𝟏 + 𝑎 12 𝒙 𝟐 ⋯+ 𝑎 1𝑖 𝒙 𝒊 +⋯+ 𝑎 1𝑛 𝒙 𝒏 = 𝒃 𝟏 ⋮ 𝑎 𝑖 1 𝒙 𝟏 + 𝑎 𝑖 2 𝒙 𝟐 ⋯+ 𝑎 𝑖 𝑖 𝒙 𝒊 +⋯+ 𝑎 𝑖 𝑛 𝒙 𝒏 = 𝒃 𝒊 𝑎 𝑛1 𝒙 𝟏 + 𝑎 𝑛2 𝒙 𝟐 ⋯+ 𝑎 𝑛𝑖 𝒙 𝒊 +⋯+ 𝑎 𝑛𝑛 𝒙 𝒏 = 𝒃 𝒏 𝑎 11 𝑎 12 ⋯⋯⋯⋯ 𝑎 1𝑛 𝑎 21 𝑎 22 ⋯⋯⋯⋯ 𝑎 2𝑛 ⋮ ⋱ ⋮ ⋮ ⋱ ⋮ ⋮ ⋱ ⋮ ⋮ ⋱ ⋮ ⋮ ⋱ ⋮ 𝑎 𝑛1 𝑎 𝑛2 ⋯⋯⋯⋯ 𝑎 𝑛𝑛 𝒙 𝟏 𝒙 𝟐 𝒙 𝟑 ⋮ 𝒙 𝒊 ⋮ 𝒙 𝒏−𝟏 𝒙 𝒏 = 𝒃 𝟏 𝒃 𝟐 𝒃 𝟑 ⋮ 𝒃 𝒊 ⋮ 𝒃 𝒏−𝟏 𝒃 𝒏 In the last class, I told you what supercomputers calculate? I showed this slide. These are computational cells for an ocean simulation. We made simultaneous equations on each computational cells. Finally I introduced that the supercomputer calculates a huge matrix. However the equations are general form. So the next, let’s see the actual equations for the ocean simulation.

Equations for Ocean Simulation 𝜼 𝒊,𝒋 𝒌+𝟏 = 𝜼 𝒊,𝒋 𝒌 − 𝜟𝒕 𝜟𝒙 × 𝑴 𝒊+𝟏/𝟐,𝒋 𝒌+𝟏/𝟐 − 𝑴 𝒊−𝟏/𝟐,𝒋 𝒌+𝟏/𝟐 − 𝜟𝒕 𝜟𝒚 × 𝑵 𝒊,𝒋+𝟏/𝟐 𝒌+𝟏/𝟐 − 𝑵 𝒊,𝒋−𝟏/𝟐 𝒌+𝟏/𝟐 8 operations 𝑴 𝒊+𝟏/𝟐,𝒋 𝒌+𝟏/𝟐 = 𝑴 𝒊+𝟏/𝟐,𝒋 𝒌−𝟏/𝟐 −𝒈 ×𝑫 𝒊+𝟏/𝟐,𝒋 𝒌 × 𝜟𝒕 𝜟𝒙 × 𝜼 𝒊+𝟏,𝒋 𝒌 − 𝜼 𝒊,𝒋 𝒌 6 operations 𝑵 𝒊,𝒋+𝟏/𝟐 𝒌+𝟏/𝟐 = 𝑵 𝒊,𝒋+𝟏/𝟐 𝒌−𝟏/𝟐 −𝒈 ×𝑫 𝒊,𝒋+𝟏/𝟐 𝒌 × 𝜟𝒕 𝜟𝒚 × 𝜼 𝒊,𝒋+𝟏 𝒌 − 𝜼 𝒊,𝒋 𝒌 Please look at these equations. The ocean simulation uses them. You don’t have to understand the meanings of equations at the moment. But please count the number of operations in these equations. The eta means water elevation. How many operations can you see in it? Yes, you can see eight operations in it. And following equations are for the M and N. Those mean fluxes. The quantity are product of velocity and volume. They had 6 operations each other. As a result it is 20 operations in each computational cell. If your FLOPS is nearly 1, you need to calculate these equations for nearly 20 seconds. In the ocean simulation case, we used over 12K cells, so about 300K operations are required for one time step simulation. If your FLOPS is ‘1’, you need 3.5 days for the calculation. We strongly want to use fast computer for the simulation ! By the way, how fast is computers?… lets see a history of the supercomputers. ----- That’s correct / yes, that’s right / close not quite 6 operations Total 20 operations/cell Total Num. of Cells: 15,000  → 300,000 operations/timestep If your FLOPS is ‘1’, you need 300,000 [sec] → 3.5 [days]

History of Super Computer Name FLOPS Year ENIAC 300 1946 The 1st computer (USA) Cray-1 250,000,000 1976 Cray (USA) FACOM VP-100 1982 Fujitsu (JPN) CM-5 65,500,000,000 1993 TM (USA) NWT 170,400,000,000 1994 NAL & Fujitsu (JPN) T3E 1,020,000,000,000 1998 Earth Simulator 35,860,000,000,000 2002 JAMSTEC & NEC (JPN) Blue Gene 280,600,000,000,000 2005 LANL & IBM (USA) MFLOPS GFLOPS TFLOPS OK next, let’s focus on the history of super computer. ENIAC is the first supercomputer in the history. It was made in 1946 and it had 300 flops ability. It was nearly 300 times faster than we came out. In the mid 70s to mid 80s the calculation ability exceeded mega flops. And in the early 90s, the ability achieved giga flops level. You can see the CM-5 in left photo. Personally the CM-5 is a memorable machine for me because I used the machine for my master’s thesis. In the late 90s to early 2000, the T3E reached tera flops. I used the machine for my doctoral thesis. So I have a lot of memories of the machine. It is a photo of T3E. By the way, I asked you which CPU is in your laptop. A lot of answers were intel Core i7. How fast is your laptop? It is from 50 – 224 GFLOPS. The ability is similar to CM-5, it was a fastest supercomputer in the world in 1993. It was very big like as a photo and 1,024 CPUs were connected, but now you can use the same ability computer personally and you can use when you want to use it. It is impressive situation for me because I used CM-5. ©www.top500.org, LANL ©Cray Thinking Machine CM-5 Cray T3E How fast is Core i7 ? Your laptop : 50 – 224 GFLOPS

1st – 5th Ranking of Top 500 List (November 2013) Country System Rmax (TFLOPS) 1 China Tianhe-2 (MilkyWay-2) 33862.7 2 USA Titan 17590.0 3 Sequoia 17173.2 4 Japan K 10510.0 5 Mira 8586.6 Top500 List – November 2013 K : 103 M : 106 G : 109 T : 1012 P : 1015 Finally I will introduce the latest supercomputing ability. Every June and November, the top 500 fastest supercomputers are listed. 2014 June’s result will be opened on the 23rd of this month. So I’ll show the November 2013 list. The fastest computer is MilkyWay-2 in China. This ability is 33 peta Flops. Peta is 10 to the power of 15, and the K in Japan is the 4th fastest computer. It is 10 peta flops. In the 3rd section of lecture, I would like to talk about the K computer.

K Computer Origin of Name : 一京 = 1016 = 10 Peta (C) RIKEN 𝐾 𝑐𝑜𝑚𝑝𝑢𝑡𝑒𝑟 𝑌𝑜𝑢𝑟 𝑙𝑎𝑝𝑡𝑜𝑝 = 10,000,000 [𝐺𝐹𝐿𝑂𝑃𝑆] 224 [𝐺𝐹𝐿𝑂𝑃𝑆] = By the way, do you know the origin of name of K computer? In Japan, the unit of 10 to the power of 16 is called KEI. The engineers named the computer in the hope to achieve 10 peta flops. When I heard the story, I felt strong passion of Japanese enginner. Let’s check the K’s ability. Your laptop has 224 giga flops. The K is 10 peta flops, so about 44 thousand times faster than your laptop. Can you imagine the calculation speed? And next. The comparing with your ability. In one second, K can calculate 10 peta operations. On the other hand, you can calculate 1 operation. When will we finish the task for 1 second of K computer. It will be 317 m years after…. It is incredible power of K. ----- June 2011 & Nov. 2011 fastest machine in TOP500 Fujitsu : http://jp.fujitsu.com/about/tech/k/?fstory=1106b06-01 864 Racks in 50m x 60m room (basketball coat 28x15 m, 6-7 times bigger) 44,642 (times faster) 𝐾 𝑐𝑜𝑚𝑝𝑢𝑡𝑒𝑟 𝑌𝑜𝑢 = 10 [𝑃𝐹𝐿𝑂𝑃𝑆] 1 [𝐹𝐿𝑂𝑃𝑆] → 317,098,082 [years]

Summary FLOPS (FLoating-Point Operations Per Second) How many ‘+ - x / ‘ in 1 second ? Indicator of ability. How fast is that ? You : ~ 1 FLOPS ? History of Supercomputers Laptop : 200 GFLOPS TOP500 in November 2013 Top : Milky-way 2 (China) 33 PFLOPS 4 th : K computer (Japan) 10 PFLOPS 44,000 times faster than laptop 317 million years human’s task for K’s 1 second task OK, I’ve now come to the end of my talk. I would like to summarize the lecture. The first I talked to you today about FLOPS. It was a indicator of ability for computer. And we calculated about ourselves ability. It was 0.5 – 1.0 flops. Next I talked about history of supercomputers and we get information of our laptop’s ability. Our laptop’s ability is estimated about 200 gflops. And finally I introduced latest result of fastest computer in the world. It is Milky-way 2 in China. It has 33 pflops ability. And K is 4th fastest computer and the ability is 10 pflops. It is 44,000 times faster than laptop. And when you explain the supercomputer to anybody, please remember what I said “Supercomputer’s 1 second task is similar with 317 million years task for person….” Thank you all for listening , it was a pleasure being here today. Thanks again.

http://www.top500.org/