Superscale Processor Block Diagram-Parallel Processing-Assignments, Exercises of Parallel Computing and Programming

This assignment was assigned by Prof. Rasul Rangarajan at Deenbandhu Chhotu Ram University of Science and Technology for Parallel Processing course. It includes: Table, Energy, Efficient, Systems, Competitor, Superscalar, Processor, Cache, Control

Typology: Exercises

2011/2012

Uploaded on 07/23/2012

parama
parama šŸ‡®šŸ‡³

4.1

(12)

56 documents

1 / 7

Toggle sidebar

This page cannot be seen from the preview

Don't miss anything!

bg1
QUESTION # 01
LIST
NAME OF
BEST
CHOICE
PROCESSING
SPEED
(MHz)
TOTAL
CORES
PROCESSOR
FAMILY
OS
FAMILY
INTERCONNECT
FAMILY
APP
AREA
COUNTRY
NOVEMBER
2011
K
COMPUTER
2000
705024
Sparc
Linux
Custom
Research
JAPAN
JUNE 2011
K
COMPUTER
2000
548352
Sparc
Linux
Custom
Research
JAPAN
NOVEMBER
2010
Chinese
Tianhe-1A
system
2930
186368
Intel
EM64T
Linux
Proprietary
Research
CHINA
JUNE 2010
The Jaguar
system
2600
224162
AMD
x86_64
Linux
Proprietary
Research
USA
NOVEMBER
2009
The Jaguar
system
2600
224162
AMD
x86_64
Linux
Proprietary
Research
USA
JUNE 2009
Roadrunner
3200
129600
Power
Linux
Infiniband
Not
Specified
USA
NOVEMBER
2008
Roadrunner
3200
129600
Power
Linux
Infiniband
Not
Specified
USA
JUNE 2008
Roadrunner
3200
122400
Power
Linux
Infiniband
Not
Specified
USA
NOVEMBER
2007
BlueGene/L
700
212992
Power
CNK/SLES
9
Proprietary
Not
Specified
USA
JUNE 2007
BlueGene/L
700
212992
Power
CNK/SLES
9
Proprietary
Not
Specified
USA
ANALYSIS
We note from the Table that Fujitsu has been dominating the Number one spot since 2011. Fujitsu
had System Share of 0.8% and Maximum Cores of only 7488, however over the years they have
improved. The K Computer is four times as powerful as its nearest competitor. The K Computer is
also one of the most energy efficient systems.
We also note that USA is being taken over by China and Japan in Super Computing. Before 2010, it
was all USA however the ā€˜Jaguar System’ built by Chinese Company NUDT took over the Famous
Roadrunner System.
docsity.com
pf3
pf4
pf5

Partial preview of the text

Download Superscale Processor Block Diagram-Parallel Processing-Assignments and more Exercises Parallel Computing and Programming in PDF only on Docsity!

QUESTION # 01

LIST NAME OF BEST CHOICE

PROCESSING SPEED (MHz)

TOTAL CORES

PROCESSOR FAMILY

OS FAMILY

INTERCONNECT FAMILY

APP AREA

VENDOR COUNTRY

NOVEMBER

K

COMPUTER

2000 705024 Sparc Linux Custom Research Fujitsu JAPAN

JUNE 2011 K

COMPUTER

2000 548352 Sparc Linux Custom Research Fujitsu JAPAN

NOVEMBER

Chinese Tianhe-1A system

2930 186368 Intel EM64T

Linux Proprietary Research NUDT CHINA

JUNE 2010 The Jaguar system

2600 224162 AMD

x86_

Linux Proprietary Research Cray Inc. USA

NOVEMBER

The Jaguar system

2600 224162 AMD

x86_

Linux Proprietary Research Cray Inc. USA

JUNE 2009 Roadrunner 3200 129600 Power Linux Infiniband Not Specified

IBM USA

NOVEMBER

Roadrunner 3200 129600 Power Linux Infiniband Not Specified

IBM USA

JUNE 2008 Roadrunner 3200 122400 Power Linux Infiniband Not Specified

IBM USA

NOVEMBER

BlueGene/L 700 212992 Power CNK/SLES 9

Proprietary Not Specified

IBM/DOE USA

JUNE 2007 BlueGene/L 700 212992 Power CNK/SLES 9

Proprietary Not Specified

IBM/DOE USA

ANALYSIS

We note from the Table that Fujitsu has been dominating the Number one spot since 2011. Fujitsu had System Share of 0.8% and Maximum Cores of only 7488, however over the years they have improved. The K Computer is four times as powerful as its nearest competitor. The K Computer is also one of the most energy efficient systems.

We also note that USA is being taken over by China and Japan in Super Computing. Before 2010, it was all USA however the ā€˜Jaguar System’ built by Chinese Company NUDT took over the Famous Roadrunner System.

QUESTION

Block Diagram of Superscalar Processor (P4)

( Diagram via www.intel.com )

Features

ļ‚· Super Scalar Processor

ļ‚· Simultaneously fetches multiple instructions

ļ‚· 20 stage pipleline

ļ‚· Increased Size of L2 Cache compared to old architectures

Question # 03

UMA

IntelCore i7 can used in UMA Mode within a single processor Node

( Diagram via www.intel.com )

Features

ļ‚· Totally Brand New Architecture of Platform ļ‚· Better Performance Due to QuickPath Interconnect ļ‚· Intel Turbo Boost Technology ļ‚· Introduction of New Smart Cache (L3) ļ‚· PCI Express upgraded to 2.

NUMA

Intel E8870 Chipset

( DIAGRAM via www.intel.com )

Features

ļ‚· System Bus and other internal architecture is well aligned

ļ‚· Support of PCI X for extra improvement in performance

ļ‚· Allows usage of Full Bandwidth with Introduction of a new Pre-Fetch Engine

CLUSTER SYSTEM

Many Processors can be used together in a cluster. I will take example of a TILE64 Processor

(BLOCK DIAGRAM via www.tilera.com)

Features

ļ‚· A total of 8 x 8 processor Cores (All Identical)

ļ‚· Implementation of VLIW Pipeline to support Parallelism

ļ‚· Cache Size of 5MB

ļ‚· Extremely Efficient in Power Consumption