HPC Intel® Xeon® Processor 5600 and 7500 Series Gold Presentation ...

bugenigmaΛογισμικό & κατασκευή λογ/κού

30 Οκτ 2013 (πριν από 3 χρόνια και 7 μήνες)

125 εμφανίσεις

(Stimulating Innovation)
2

Illuminate. Insight. Solve.




April 2009

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Meeting Today’s HPC Challenges


Versatile, scalable solutions to enable innovation

Performance

Adaptability

Integration and
Consistency

CPU Bound

Memory

Bandwidth

Bound

Memory

Capacity

Bound

Flexibility /

Cost Sensitive

For demonstration purposes only

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Xeon
®

7500 Series

CPU Bound

Memory Bandwidth

Memory Capacity

Xeon
®

5600 Series

Cost Sensitive

HPC Platform Decision Criteria

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

Processor 5600 Series

Building on Xeon® 5500 Leadership Capabilities


Xeon
®

5600


Xeon
®

5600

3.46 GHz 130W, 4C

3.33 GHz 130W, 6C

3.06 GHz 95W, 4C

2.93 GHz 95W, 6C

Greater
performance

at the same power

Greater performance
for bandwidth
sensitive apps.

Up to 2
-

1333 MHz
DIMMs per channel.

Nehalem Micro
-
architecture + 32nm CPU = greater
perf
. and
perf
./watt

Providing more
performance for data
intensive workloads

Up to 6 cores, Up to 12MB
Cache.

Delivering more into
the same
package…increasing
perf
./$

PCI Express* 2.0

ICH 10/10R

Intel
®

82599
10GbE
Controller

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

5600

A New Generation of Intelligent Processors

Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate perf
orm
ance of Intel products as measured by those tests. Any
difference in system hardware or software design or configuration may affect actual performance. Buyers should consult other
sou
rces of information to evaluate the performance of systems or
components they are considering purchasing. For more information on performance tests and on the performance of Intel product
s,
visit http://www.intel.com/performance/resources/limits.htm
Copyright © 2010, Intel Corporation. * Other names and brands may be claimed as the property of others.


Knows Where to Put the Speed, Knows How to Save Energy

HPC Applications,
SPECfp
*_rate & STREAM

Bandwidth Sensitive apps

Frequency Sensitive apps

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Xeon
®

5500


Xeon
®

5600

SKU Transition

Advanced


6.4 GT/s QPI


8MB / 12MB


DDR3 1333


Turbo Boost


HT

Standard


5.86 GT/s QPI


8MB / 12MB


DDR3 1066


Turbo Boost


HT

Basic


4.8 GT/s QPI


4M cache


DDR3 800

80W

L5640 6C

2.26 GHz

Xeon
®

5500


Usage

Xeon
®

5600

(Xeon 5600
-
EP)

X5570

2.93 GHz

X5560

2.80 GHz

X5550

2.66 GHz

E5540

2.53 GHz

E5530

2.40 GHz

E5520

2.26 GHz

E5506

2.13 GHz

E5504

2.00 GHz

95W

95W

95W

80W

80W

80W

80W

80W

E5502

1.86 (2C)

80W

X5670 6C

2.93 GHz

X5660 6C

2.80 GHz

X5650 6C


2.66 GHz

E5640 4C

2.66 GHz

E5630 4C

2.53 GHz

E5620 4C

2.40 GHz

95W

95W

95W

80W

80W

X5680 6C

3.33 GHz

130W

E5507

2.26 GHz (4C)

E5506

2.13GHz (4C)

80W

80W

E5503

2.00 GHz (2C)

80W

60W

80W

L5630 4C

2.13 GHz

40W

L5609 4C

1.86 GHz

40W

Xeon
®

5600 (WSM
-
EP) SKUs


4C Xeon
®

5500 SKUs


Low Power Options

X5677 4C

3.46 GHz

130W

Freq
-
Optimized

X5667 4C

3.06 GHz

95W

Maximum

Performance

6 Cores

4 Cores

(freq optimized)

Cost
-
Optimized

Higher Frequency

Xeon
®

E5500 SKUs

Best Price

Performance

Higher Freq

More Cache

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

5600
Sku

Benefits

SPECint_rate_base2006* (Linux)

Intel® Xeon® 5600 Series

Xeon 5680

Xeon 5506

Xeon 5630

Xeon 5650

Xeon 5640

Xeon 5660

Xeon 5620

Xeon 5507

Advanced

Standard

Basic

Xeon 5670

Even better
ROI
opportunity with Xeon® 5600 (WSM
-
EP)

Differentiation:

Frequency

QuickPath

CPU Cache

TurboBoost

Hyper
-
threading

6
core

Up to 39%

Performance

Boost

Up to 46%

Performance

Boost

Performance Boost

Xeon 5500 lineup

Xeon® 5500 source: Intel internal measurements submitted to
www.spec.org

for SPECint_rate_2006 on March 16, 2009. Turbo Boost and Intel HT are both ON. Complete configuration details
at
www.spec.org
.


Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate perf
orm
ance of Intel products as measured by those tests. Any
difference in system hardware or software design or configuration may affect actual performance. Buyers should consult other
sou
rces of information to evaluate the performance of systems or
components they are considering purchasing. For more information on performance tests and on the performance of Intel product
s,
visit
Intel Performance Benchmark Limitations
.


* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

7500
-

Super Node

Delivering large memory, innovative solutions to HPC

2.66 GHz 130W, 6C

2.26 GHz 130W, 8C

Meeting the needs of
frequency sensitive
HPC users

Higher Frequency

Ideal for large
memory and
memory bandwidth
sensitive workloads

Support for Node Controller


8X
the bandwidth*

4X the capacity*

The

super node” delivers the necessary compute, memory and memory
bandwidth performance to solve big science faster.

50% More Cache*/ 33%
More Cores*

Providing more
performance for data
intensive workloads

Scalable to 256 sockets
for large, scalable
workloads

*See legal disclaimer in backup

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Workload
Requirements


Maximum scalability


Large memory & cache


Minimize node to node

communication


Single system image

Transforming HPC with Super Node Scalability

Xeon
®

7500

Manufacturing CAE

Energy

Life Science

Weather

FSI

Transformative Designs by:

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Basic

Standard

Advanced

Using Advanced Intel
®

Xeon
®

7500
Processors for HPC

Selection Guidance

X7560

X7550

X6550

X7542

E7540

E7530

E6540

E7520

Frequency

2.26 GHz

2.0 GHz

2.0 GHz

2.66 GHz

2.0 GHz

1.86 GHz

2.0 GHz

1.86 GHz

Cores

8

8

8

6

6

6

6

4

Cache

24M

18M

18M

18M

18M

12M

18M

18M

QPI

6.4 GT/s

6.4 GT/s

6.4 GT/s

5.86 GT/s

6.4 GT/s

5.8 GT/s

6.4 GT/s

4.8 GT/s

Power
Rqmt

130W

130W

130W

130W

105W

105W

105W

95W

Scalable

w/o NC

Yes

Yes

No

Yes

Yes

Yes

No

Yes

Scalable

to
xx w/o NC

8S

8S

NA

8S

8S

4S

NA

4S

Optimized for HPC

Maximize Cores, scalability,
frequency and price

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Nehalem
-
EX delivers up to 2.4x average / 2.7x max HPC app performance**

Intel
®

Xeon® 7500 Performance Summary

4S HPC Benchmarks Comparison to 5600 Series

HPC Benchmark Suite

Relative Geometric Mean Scores by
segment. Higher is better.

Maximum performance
observed in suite**

See legal disclaimer in backup

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

X7542 performs and average of ~0.9x on HPC apps at a lower CPU price**

Intel® Xeon® Processor 7500 series
-
based Server

4S Nehalem
-
EX HPC Benchmarks Top
-
Bin vs. “Frequency Optimized” SKU

HPC Benchmark Suite

Relative Average Scores by segment.

Higher is better.

Source: Intel Internal measurements Dec 2009. See backup for additional
details

** proposed 1Ku list price vs. X7560

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

BACK UP

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Cluster Ready applications
execute on top of the
architecture

Intel® Cluster Ready
Architecture

One Architecture Meant To Maximize Your Product’s Potential
Reach

Wide variety of system
software capable of providing
compliant software stacks

The architecture promotes
seamless utilization of a wide
range of network, storage and
hardware solutions.

Intel
®

Cluster Ready certified solutions are easily paired with registered applications

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

What does it mean to certify a solution?

Certification means designing a solutions that is compliant with the Intel
Cluster Ready specification and is repeatable. It also means that certified
systems can run multiple certified applications on the same certified cluster.



Get the Intel® Cluster Ready Seal of
Approval

Certifying Every System as Intel Cluster Ready

Develop
cluster recipes
with Intel
®

Cluster Ready
specification

Intel
®

Cluster Ready Hardware System Certification Process

Certify your
system is
Intel Cluster
Ready
compliant

Integrate Intel
Cluster Ready
into
manufacturing
process

Prepare sales
and marketing
for Intel
Cluster Ready

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.


Validates the system
contains the architecture


Provides detailed
diagnostic information in
an easy
-
to
-
read format


100+ checks for a wide
array of cluster
evaluations or create
your own

More About Intel® Cluster Checker

Helping You Get to Market Faster and Support Customers Better


Helps isolate system
problems from application
problems


Intel Cluster Checker
dashboard clearly depicts
system health


Helps reduce the time
spent troubleshooting and
helps minimize the
number of support
personnel needed


* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel 10 Gigabit Ethernet for HPC

HPC Fabric Needs?

Niantic (82599)


Dual

10GbE NIC or LOM


L2 I/F, small packet
perf
.


Integrated KR
interfaces


Storage:
iSCSI
,

FCoE


Integrated KR/SFI


Power: 5.2W typical

NE020 (
NetEffect
)


iWARP

10GbE NIC


RDMA apps over Ethernet


iWARP

I/F for RDMA apps


Single
-
digit µs latency


OS: RH/SLES OFED,W2K8


Power: 8.5W typical

SFP+ NIC


Great
perf
. &
driver support

-

Wall Street

BladeCenter

10GbE
Mezz
.

-

IBM

T100 Efficiency

in T250 Super
-

computer

-

Biomedical


Firm

10GbE RDMA

Messaging s/w

-

NYSE

Intel® Ethernet

“It Just Works”

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel®
Ethernet Server Cluster

Adapters w/ iWARP


Low
-
latency 10Gb Ethernet


iWARP

(RDMA over Ethernet)


5

s latency for <10 connections


<25

s for 64 connections


User
-
space placement avoids copies & context switches


Ethernet alternative to
InfiniBand


Standards
-
based Ethernet


Full Interoperability with Ethernet infrastructure


Uses existing IP manageability IT practices


OS Support


Linux: OFED 1.4.1 support in Red Hat & SUSE


Windows: Microsoft Windows HPC Server 2008


Adapter Feature Support


PCI Express v1.1 x8 Interface


256MB local memory


ECC protected


PXE Boot*, Wake on LAN


LAN Interfaces: CX4 or SFP+

Ethernet

iWARP


PCIe

Sys Driver


Application

OS Stack

IO Lib

S/W

H/W

Basic

Network

Controller

Flow

TCP/IP

Kernel

User

Avoid

Copies

& Kernel

iWARP

Network

Controller

Flow

Without

iWARP

With

iWARP

Shipping

Now!

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Top500 Class Supercomputer w/
iWARP

Where Would it Land on November 2009 Top500 List?


155th in overall performance


73th in overall efficiency


Next most Efficient Ethernet Cluster


70% & 20.6TF which ranks #486



Cluster Hardware Details


4032 cores, 1008 Sockets,
Intel® Xeon 5550 2.66GHz


504 servers, Dell R610


12TB of system memory
(24GB/node)


40TB of storage (80GB/node)


10GbE
iWARP

Fabric

>
Intel’s
NetEffect

Server
Cluster Adapter DA

>
Arista

Networks 7148SX low
-
latency switches


2.3 to 1 Oversubscribed
Fabric


Cluster Software Details


Intel MPI 3.2.1


uDAPL

1.2


35.81 TeraFLOPs

84.14% Efficiency

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Financial Messaging Performance

CLUSTER DETAILS:

12 servers/nodes

Red Hat

1 Publisher

12 Subscribers

Arista

71245 24
-
port 10Gb Switch

SFP+
Twinax

Cabling


NODE DETAILS:

2x Intel E5540 2.53GHz

16GB 1066MHz DDR3

NYSE Data Fabric™ Messaging

NetEffect

Server Cluster Adapter

(10Gb Ethernet)

Red Hat 5.3

OFED 1.4.1

Performance tests and ratings are measured using specific computer systems and/or components and reflect the approximate perf
orm
ance of
Intel products as measured by those tests. Any difference in system hardware or software design or configuration may affect a
ctu
al
performance. Buyers should consult other sources of information to evaluate the performance of systems or components they are

co
nsidering
purchasing. For more information on performance tests and on the performance of Intel products, visit
http://www.intel.com/performance/resources/benchmark_limitations.htm.

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel Ethernet Resources


Intel Ethernet


www.intel.com/go/ethernet


10GbE vs. 1GbE Power/Cost Calculator:


www.event
-
management
-
online.de/LAD/calculator.aspx


iWARP

Super
-
computer Whitepaper


download.intel.com/support/network/
sb
/inteliwarp_researchinst.pdf


* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

LEGAL DISCLAIMERS

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

A Family of Processors

to Meet Your HPC Needs

Powerful, Intelligent, Adaptable


Average of 3x performance claim based on geometric mean of four industry
-
standard,
common enterprise benchmarks (
SPECjbb
*2005,
SPECint
*_rate_base2006,
SPECfp
*_rate_base2006, and TPC Benchmark* E) comparing best published / submitted
results on 4
-
socket (4S) Intel Xeon processor X7560

based server platform to best
published 4S Intel Xeon processor X7460

based server platform as of March 26, 2010.



BLACKSCHOLES*:
Baseline Configuration and Score on Benchmark:
-

Intel pre
-
production system with two Intel® Xeon® processor X5570 (2.93 GHz, 8 MB last level
cache, 6.4 GT/sec QPI), 24GB memory (6x4GB DDR3
-
1333), 4 x 150GB 10K RPM SATA
RAID0 for scratch, Red Hat* EL 5 Update 4 64
-
bit OS. Source: Intel internal testing as of
February 2010. SunGard v3.0 source code compiled with Intel v11.0 compiler. Elapsed
time to run benchmark: 18.74 seconds.


* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

5600

A New Generation of Intelligent Processors


STREAM:
Baseline Configuration and Score on
Stream
-
MP
Benchmark:
-

Intel pre
-
production system with two Intel® Xeon® processor X5570 (2.93 GHz, 12MB L3, 6.4
GT/s, Quad
-
core, 130W TDP), C3 Disabled, C6 Enabled, Turbo Disabled, HT Disabled, NUMA Enabled, 24 GB (6x4GB DDR3
-
1333 DR regist
ered ECC), 1x150GB 10K RPM
SATA HDD,
Red
Hat

EL5
-
U4
kernel

2.6.18
-
164.el5 experimental.8
). Source: Intel internal testing TR1012. Score of workloads:
36588.0 MB/s


New Configuration and Score on Stream
-
MP Benchmark:
-

Intel pre
-
production system with two Intel® Xeon® processor X5680 (3.33 GH
z, 12MB L3, 6.4 GT/s, Hex
-
core,
130W TDP), C3 Disabled, C6 Enabled, Turbo Disabled, HT Disabled, NUMA Enabled, 24 GB (6x4GB DDR3
-
1333 DR registered ECC), 1x150G
B 10K RPM SATA HDD,
Red
Hat

EL5
-
U4
kernel

2.6.18
-
164.el5 experimental.8
). Source: Intel internal testing as of Feb 2010. Score of workloads: 43000
MB/s
.



CAE:
Baseline Configuration and Score on
CAE

Vertical: 2
-
socket server with Intel® Xeon® processor X5570 (2.93 GHz, 8MB LLC, 6.4 GT/S QPI, 95W TDP), Turbo
Enabled, HT Enabled or Disabled for best performance on each application, 24 GB memory (6x4GB DDR3
-
1333 registered ECC), 4x150GB

10K RPM SATA HDD RAID0 for
scratch, Red Hat* EL 5.4 64
-
bit OS (2.6.18
-
164
-
el5). Source: Intel internal testing as of January 2010. Geometric mean score for

nine applications


New Configuration and Score on CAE Vertical: Intel pre
-
production 2
-
socket server with Intel® Xeon® processor X5680 (3.33 GHz,
12MB LLC, 6.4 GT/S QPI, 130W TDP),
Turbo Enabled, HT Enabled or Disabled for best performance on each application, 24 GB memory (6x4GB DDR3
-
1333 registered ECC), 4
x150GB 10K RPM SATA HDD RAID0
for scratch, Red Hat 5.4 64
-
bit OS (2.6.18
-
164
-
el5). Source: Intel internal testing as of January 2010. Geometric mean score for

nine applications



SPECfp_rate:
25% gain on SPECfp_rate_base2006 over previous generation supported by the following:


Baseline Configuration and Score: Score 197, Referenced as published at
http://www.spec.org/cpu2006/results/res2009q2/cpu2006
-
20090511
-
07354.html



New Configuration and Score: Dell PowerEdge R710 system with two Intel® Xeon® processor X5680 (3.33 GHz, 12MB L3, 6.4 GT/s, s
ix
-
core, 130W TDP), Maximum
Performance Power Management mode, Data Reuse Disabled, C1E Enabled, Turbo Enabled, HT Enabled, 48 GB (12x4GB DDR3
-
1333 register
ed ECC), 1x146GB 15K RPM
SAS HDD, SUSE Linux Enterprise Server 11 (2.6.27.19
-
5
-
smp).


Source: Submitted to www.spec.org for publication as of March 15 20
10.


Geomean

Score of 17 workloads:
248.



LIFE SCIENCES:
Baseline Configuration and Score on
Life Sciences
Vertical: 2
-
socket server with Intel® Xeon® processor X5570 (2.93 GHz, 8MB LLC, 6.4 GT/S QPI,
95W TDP), Turbo Enabled, HT Enabled or Disabled for best performance on each application, 24 GB memory (6x4GB DDR3
-
1333 register
ed ECC), 4x150GB 10K RPM SATA
HDD RAID0 for scratch, Red Hat* EL 5.4 64
-
bit OS (2.6.18
-
164
-
el5). Source: Intel internal testing as of January 2010. Geometric
mean score for ten applications


New Configuration and Score on Life Sciences Vertical: Intel pre
-
production 2
-
socket server with Intel® Xeon® processor X5680 (
3.33 GHz, 12MB LLC, 6.4 GT/S QPI, 130W
TDP), Turbo Enabled, HT Enabled or Disabled for best performance on each application, 24 GB memory (6x4GB DDR3
-
1333 registered E
CC), 4x150GB 10K RPM SATA HDD
RAID0 for scratch, Red Hat 5.4 64
-
bit OS (2.6.18
-
164
-
el5). Source: Intel internal testing as of January 2010. Geometric mean sco
re for ten applications



LINPACK:
Baseline Configuration and Score on
Linpack

Benchmark:
-

Supermicro* pre
-
production system with two Intel® Xeon® processor X5570 (2.93 GHz, 8MB L3,
6.4 GT/s, Quad
-
core, 95W TDP), BIOS rev 02/23/2009, C3 Disabled, C6 Enabled, Turbo Enabled, HT Disabled, NUMA Enabled, 24 GB (6x
4GB DDR3
-
1333 DR registered
ECC), 1x150GB 10K RPM SATA HDD,
Red
Hat

EL5
-
U3
kernel

2.6.18
-
128.el5 for x86_64
). Source: Intel internal testing TR1011A. Score of workloads:
91
GFlops
.


New Configuration and Score on
Linpack

Benchmark:
-

Supermicro* pre
-
production system with two Intel® Xeon® processor X5680 (3.33 GHz, 12MB L3, 6.4 GT/s, Hex
-
core, 130W TDP), BIOS rev 01/06/2010, C3 Disabled, C6 Enabled, Turbo Enabled, HT Disabled, NUMA Enabled, 24 GB (6x4GB DDR3
-
1333
DR registered ECC), 1x150GB
10K RPM SATA HDD,
Red
Hat

EL5
-
U4
kernel

2.6.18
-
164.el5 experimental.8
). Source: Intel internal testing as of Feb 2010. Score of workloads:
146
GFlops
.




BLACKSCHOLES*:
Baseline Configuration and Score on Benchmark:
-

Intel pre
-
production system with two Intel® Xeon® processor X5570 (2.93 GHz, 8 M
B last level
cache, 6.4 GT/sec QPI), 24GB memory (6x4GB DDR3
-
1333), 4 x 150GB 10K RPM SATA RAID0 for scratch, Red Hat* EL 5 Update 4 64
-
bit O
S. Source: Intel internal testing
as of February 2010. SunGard v3.0 source code compiled with Intel v11.0 compiler. Elapsed time to run benchmark: 18.74 second
s.


New Configuration and Score on Benchmark:
-

Intel pre
-
production system with two Intel® Xeon® processor X5680 (3.33 GHz, 12 MB l
ast level cache, 6.4 GT/sec QPI),
24GB memory (6x4GB DDR3
-
1333), 4 x 150GB 10K RPM SATA RAID0 for scratch, Red Hat* EL 5 Update 4 64
-
bit OS. Source: Intel interna
l testing as of February 2010.
SunGard v3.0 source code compiled with Intel v11.0 compiler. Elapsed time to run benchmark: 11.51 seconds.



* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Performance Publication Opportunities

Westmere
-
EP

Platform

Target Segment

Benchmark

Target
Score

Target
Date

Server Benchmarks

Integer

SPECint_rate2006

352

Floating point

SPECfp_rate2006

247

Java

SPECjbb2005 (Oracle JVM)

760K

SPECjbb2005 (Sun
HotSpot

JVM)

835K

SPECjbb2005 (IBM J9
JVM)

890K

SPECjvm2008

458

Web

SPECweb2005/2009

13181PHP

ERP

SAP
-
SD (MS)

4420

Business Intelligence

TPC
-
H

TBD

Database

TPC
-
C

903K

TPC
-
E

1200

Power

SPECpower_ssj2008

45%

Virtualization

VMMark

32.2

SPECvirt_sc2009

?

HPC

Manufacturing

LS
-
Dyna

TBD

Fluent

TBD

Abaqus

TBD

HPC

OMP

SPECOMP2001

TBD

SPECMPI2007

TBD

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Nehalem
-
EP to Westmere
-
EP Comparison

Nehalem
-
EP

Westmere
-
EP

Cores

2 or 4 cores

4 and 6 core SKUs

Cache Size

4 or 8 MB

Up to 12 MB

Max DPC & config

2
-
3 RDIMMs


or 2 UDIMMs

Same max (3DPC) as
NHM
-
EP;

(LV DDR3 up to 1066 MHz
2DPC)

Intel
®

TXT (formerly LT
-
SX)

No

Yes

Virtualization Technology

EPT, VPID, etc.

adds Real Mode support
and transition latency
reduction

New Instructions

(baseline)

AES
-
NI, PCMULQDQ

TDP

Server/Workstation: 130W, 95W, 80W, 60W, <=45W



Max DIMMs per channel supported is memory speed dependent

“Drop
-
in” replacement for existing Intel 5500 Chipset platforms

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Nehalem
-
EX Super Node

Delivering large memory, innovative solutions to HPC

**Xeon 7500 offers performance, feature and other metrics “up to”
the levels shown

1.

8X the bandwidth
: Per Intel internal memory BW measurement
3.2.10 vs. previous generation Intel® Xeon® processor 7400 series
based solutions

2.

4x the capacity:

Previous generation Intel ®7300 Chipset
supported up to 8 DIMMs per channel x 4 x 16GB DIMMs = 512GB.
Intel® Xeon 7500 series
glueless

8 socket configuration can support
16 DIMMs per socket x 8 sockets x 16GB DIMMs = 2,048GB. 4X the
memory capacity of previous generation

3.

50% more Cache/ 33% more Cores
: Previous generation Xeon
7400 offers: 16MB cache and 6 cores. Xeon 7500 offers up to
24MB
cache (50
% increase) and 8 cores (33% increase)

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

HPC “Super
-
Node”

Nehalem
-
EX 4S


Source: Intel Internal measurements
November2009.



Performance tests and ratings are measured using specific computer
systems and/or components and reflect the approximate performance of
Intel products as measured by those tests. Any difference in system
hardware or software design or configuration may affect actual
performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are
considering purchasing. For more information on performance tests
and on the performance of Intel products, Go to:
http://www.intel.com/performance/resources/benchmark_limitations.htm
. Copyright © 2010, Intel Corporation. * Other names and brands may
be claimed as the property of others.



Source: Intel Internal measurements November2009. See backup for additional
details

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

5 Month Single Core Refresh ROI Claim


Back Up


5 month ROI claim estimated based on comparison between 2S Single Core Intel®
Xeon® 3.80 with 2M L2 Cache and 2S Intel® Xeon® X5670 based servers. Calculation
includes analysis based on performance, power, cooling, electricity rates, operating
system annual license costs and estimated server costs. This assumes 8kW racks, $0.10
per kWh, cooling costs are 2x the server power consumption costs, operating system
license cost of $900/year per server, per server cost of $7200 based on estimated list
prices and estimated server utilization rates. All dollar figures are approximate.
Performance and power comparisons are based on estimated server side java
benchmark results (Intel Corporation Jan 2010). Platform power was estimated during
the steady state window of the benchmark run and at idle. Performance gain compared
to baseline was 15x.


Baseline platform: Intel server platform with two 64
-
bit Intel Xeon Processor 3.80Ghz with 2M L2
Cache, 800 FSB, 8x1GB DDR2
-
400 memory, 1 hard drive, 1 power supply, Microsoft* Windows*
Server 2003
Ent
. SP1, Oracle*
JRockit
* build P27.4.0
-
windows
-
x86_64 run with 2 JVM instances


New platform: Intel server platform with two six
-
core Intel® Xeon® processor X5670, 2.93 GHz,
12MB L3 cache, 6.4QPI, 12 GB memory (6x2GB DDR3
-
1333), 1 hard drive, 1 power supply,
Microsoft Windows Server 2008 64 bit SP2, Oracle*
JRockit
* build P28.0.0
-
29 run with 2 JVM
instances


Performance tests and ratings are measured using specific computer systems and/or
components and reflect the approximate performance of Intel products as measured by
those tests. Any difference in system hardware or software design or configuration may
affect actual performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are considering purchasing.
For more information on performance tests and on the performance of Intel products,
visit Intel Performance Benchmark Limitations.

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Single Core Energy Efficient Refresh Calculation Details

2005

2010

Delta / Notes

Product

Intel® Xeon® 3.8GHz
with 2M cache

Intel® Xeon® X5670
(2.93GHz)

Performance

per Server

1

Up to 15x increase

Intel internal measurements on a
server side java benchmark as of
Jan 2010

Server Power

Idle / Active Power

228W idle / 382W
active

118W idle / 375W
active

Server idle for 16 hours per day
and active for 8 hours per day

# Servers needed

315

21

~ 15:1 server consolidation

# Racks needed

15 racks

1 rack

15:1 Rack
Consolidation

Annual Server
kWhr

772,904

Estimated:

38,184

Estimated 93% lower
energy costs

Total Annual
Energy Costs

$154,581

Estimated:

$7,637

$146,944
electricity cost reduction per year.
Assumes $0.10/kWhr and 2x cooling factor

Operating System
Licensing Costs

$283,500

$18,900

$264,600 less per year

Assumes a RHEL 1yr license at $900

Source www.dell.com as of 12/16/08

Estimated
Annual Cost Savings of $411,544

Cost of new HW

n/a

$151,200

Assume $7,200 per server

Estimated Payback Period of 5 months


For Backup, but not required as part of disclaimer

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

15 Month Dual Core Refresh ROI Claim


Back up


15 month ROI claim estimated based on comparison between 2S Dual Core Intel®
Xeon® 5160 (3.0GHz) and 2S Intel® Xeon® X5670 based servers. Calculation includes
analysis based on performance, power, cooling, electricity rates, operating system
annual license costs and estimated server costs. This assumes 8kW racks, $0.10 per
kWh, cooling costs are 2x the server power consumption costs, operating system license
cost of $900/year per server, per server cost of $7200 based on estimated list prices
and estimated server utilization rates. All dollar figures are approximate. Performance
and power comparisons are based on
estimated

server side java* benchmark results
(Intel Corporation Jan 2010). Platform power was
estimated

during the steady state
window of the benchmark run and at idle. Performance gain compared to baseline was
5x.


Baseline platform: Intel server platform with two dual core Intel® Xeon® Processor 5160,
2.93GHz, 1333MHz FSB, 8x2GB FBDMIMM DDR2
-
667 memory, 1 hard drive, 1 power supply,
Microsoft* Windows* Server 2003 Ent. SP1, Oracle* JRockit* build P27.4.0
-
windows
-
x86_64 run
with 2 JVM instances


New platform: Intel server platform with two six
-
core Intel® Xeon® processor X5670, 2.93 GHz,
12MB L3 cache, 6.4QPI, 12 GB memory (6x2GB DDR3
-
1333), 1 hard drive, 1 power supply,
Microsoft Windows Server 2008 64 bit SP2, Oracle* JRockit* build P28.0.0
-
29 run with 2 JVM
instances


Performance tests and ratings are measured using specific computer systems and/or
components and reflect the approximate performance of Intel products as measured by
those tests. Any difference in system hardware or software design or configuration may
affect actual performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are considering purchasing.
For more information on performance tests and on the performance of Intel products,
visit Intel Performance Benchmark Limitations.

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Dual Core
Energy Efficient

Refresh Calculation Details

2005

2010

Delta / Notes

Product

Intel Xeon 5100 series
(3.00GHz)

Intel® Xeon® X5670
(2.93GHz)

Performance

per Server

1

Up to 5x increase

Intel internal measurements on a
server side java benchmark as of
Jan 2010

Server Power

Idle / Active Power

252W idle / 354W
active

118W idle / 375W
active

Server idle for 16 hours per day
and active for 8 hours per day

# Servers needed

112

21

5:1 server consolidation

# Racks needed

5 racks

1 rack

5:1 Rack
Consolidation

Annual kWhr

281,883

Estimated:

38,184

Estimated 83% lower
energy costs

Annual Energy
Costs

$56,376

Estimated:
$7,637

$45,120
electricity cost reduction per year.
Assumes $0.10/kWhr and 2x cooling factor

OS Licensing Costs

$100,800

$18,900

$75,600 less per year

Assumes a RHEL 1yr license at $900

Source www.dell.com as of 12/16/08

Estimated
Annual Cost Savings of $120,720

Cost of new HW

n/a

$151,200

Estimated Payback Period of 15 months

For Backup, but not required as part of disclaimer

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

EP/EX Memory Comparison

What’s New in Expandable Servers


Source: Intel Internal measurements
Sep 2009
.
See backup for additional
details


Performance tests and ratings are measured using specific computer
systems and/or components and reflect the approximate performance of
Intel products as measured by those tests. Any difference in system
hardware or software design or configuration may affect actual
performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are
considering purchasing. For more information on performance tests and
on the performance of Intel products, visit
http://www.intel.com/performance/resources/limits.htm Copyright ©
2010, Intel Corporation.

* Other names and brands may be claimed as the property of others.


* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

HPC “Super
-
Node”

Nehalem
-
EX 4S


Source: Intel Internal measurements
November2009. See backup for additional details


Performance tests and ratings are measured using specific computer
systems and/or components and reflect the approximate performance of
Intel products as measured by those tests. Any difference in system
hardware or software design or configuration may affect actual
performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are
considering purchasing. For more information on performance tests
and on the performance of Intel products, Go to:
http://www.intel.com/performance/resources/benchmark_limitations.htm
. Copyright © 2010, Intel Corporation. * Other names and brands may
be claimed as the property of others.



* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel
®

Xeon
®

5600 Sell
-
up Benefits

SPECint_rate_base2006* (Linux)

Xeon® 5500 source: Intel internal measurements submitted to
www.spec.org

on March 16, 2009. Turbo Boost and Intel HT are
both ON. Complete configuration details at
www.spec.org
.

Xeon® 5600 source: Internal Intel estimates based on preliminary
testing with WSM
-
EP 3.20 GHz SKUs. Assumes 5% performance
scaling for Standard SKUs, and 35% for 6C SKUs.

Performance tests and ratings are measured using specific computer
systems and/or components and reflect the approximate
performance of Intel products as measured by those tests. Any
difference in system hardware or software design or configuration
may affect actual performance. Buyers should consult other sources
of information to evaluate the performance of systems or
components they are considering purchasing. For more information
on performance tests and on the performance of Intel products, visit
Intel Performance Benchmark Limitations
.



* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Source: Intel Internal measurements Dec 2009. See backup for
additional details

** vs. X5680

Intel® Xeon® 7500 Performance Summary

4S HPC Benchmarks Comparison to 5600 Series

Performance tests and ratings are measured using specific computer
systems and/or components and reflect the approximate performance of
Intel products as measured by those tests. Any difference in system
hardware or software design or configuration may affect actual
performance. Buyers should consult other sources of information to
evaluate the performance of systems or components they are
considering purchasing. For more information on performance tests and
on the performance of Intel products, Go to:
http://www.intel.com/performance/resources/benchmark_limitations.htm
.
Copyright © 2010, Intel Corporation. * Other names and brands may be
claimed as the property of others.

HPC Benchmark Suite

* Other names and brands may be claimed as the property of others.




Copyright © 2010, Intel Corporation.

Intel® Xeon® Processor 7500 series
-
based
Server

4S Nehalem
-
EX HPC Benchmarks Top
-
Bin vs. “Frequency Optimized” SKU


Average

performance of common kernel benchmarks shown by
segment


Compares top
-
bin to “frequency
-
optimized” Nehalem
-
EX SKU


X7560 is baseline


Average 0.9x scaling with X7542


Better per core memory bandwidth contributor


Great price performance value with X7542 on HPC
-
type
applications