Performance and power efficiency of
Dell PowerEdge servers with E5-2600 v2
This white paper shows the performance and power efficiency improvements of Dell™ PowerEdge™ and PowerEdge C servers with the Intel® Xeon® processor E5-2600 v2 product family.
Solutions Performance Analysis
Dell | Enterprise Solutions Group
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
ii
This document is for informational purposes only and may contain typographical errors and technical
inaccuracies. The content is provided as is, without express or implied warranties of any kind.
©2013 Dell Inc. All rights reserved. Dell and its affiliates cannot be responsible for errors or omissions in
typography or photography. Dell, the Dell logo, and PowerEdge are trademarks of Dell Inc. Intel and Xeon are
registered trademarks of Intel Corporation in the U.S. and other countries. Microsoft, Windows, and Windows
SQL Server are either trademarks or registered trademarks of Microsoft Corporation in the United States
and/or other countries. Other trademarks and trade names may be used in this document to refer to either the
entities claiming the marks and names or their products. Dell disclaims proprietary interest in the marks and
names of others.
Members of Dell’s Solutions Performance Analysis team contributing to this whitepaper are: Louis Barton,
Brian Bassett, John Beckett, Burton Culley, Burton Finley, Waseem Raja, Donald Russell, Daniel Turner, Bruce
Wagner and Juergen Zimmermann.
October 2013 | Version 1.11
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
iii
Contents
Executive summary ....................................................................................................................................................... 5
Introduction ..................................................................................................................................................... 5
Key findings ...................................................................................................................................................... 5
Power efficiency with E5-2600 v2 ................................................................................................ 5
Performance with E5-2600 v2 ....................................................................................................... 5
Methodology .................................................................................................................................................................. 6
Integer performance .................................................................................................................................................... 7
SPEC CPU2006 integer tests .......................................................................................................... 7
Java server tests ................................................................................................................................. 9
Floating point performance ....................................................................................................................................... 11
SPEC CPU2006 floating point tests ............................................................................................. 11
HPC performance tests .................................................................................................................. 13
Memory subsystem performance ........................................................................................................................... 15
Workloads performance ............................................................................................................................................ 16
Database performance ................................................................................................................................ 16
VDI ....................................................................................................................................................................17
Virtualization .................................................................................................................................................. 18
Big data ........................................................................................................................................................... 19
Power efficiency .......................................................................................................................................................... 20
Summary ....................................................................................................................................................................... 22
Appendix A — Test configurations .......................................................................................................................... 23
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
iv
Tables
Table 1: Benchmark configurations ........................................................................................................................ 23
Figures
Figure 1: Performance improvement running SPECint_rate_base2006 .......................................................... 7
Figure 2: Performance improvement running SPECint_base2006 ................................................................... 8
Figure 3: Performance improvement running SPECjbb2005 ............................................................................. 9
Figure 4: Performance improvement running SPECjbb2013 ............................................................................ 10
Figure 5: Performance improvement running SPECfp_rate_base2006 .......................................................... 11
Figure 6: Performance improvement running SPECfp_base2006 .................................................................. 12
Figure 7: Performance improvement running Linpack ....................................................................................... 13
Figure 8: Performance improvement running SPEC OMP2012 ....................................................................... 14
Figure 9: Performance improvement running STREAM ..................................................................................... 15
Figure 10: Performance improvement running a database workload ............................................................ 16
Figure 11: Performance improvement running Login VSI ...................................................................................17
Figure 12: Performance improvement running VMmark 2.5 ............................................................................. 18
Figure 13: Performance improvement running HiBench K-Means Clustering benchmark ....................... 19
Figure 14: Power efficiency improvement of the R720 running the SPECpower_ssj2008 benchmark . 20
Figure 15: Power efficiency improvement of the M620 running the SPECpower_ssj2008 benchmark 21
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
5
Executive summary
Introduction
Dell PowerEdge and PowerEdge C servers are now available with the Intel Xeon processor E5-2600 v2
product family, code named “Ivy Bridge-EP”. These new Xeon processors feature up to 12 cores and are
fabricated on a 22-nanometer process for improved power efficiency.
In order to show customers the performance and power efficiency improvements that E5-2600 v2 processors
bring to PowerEdge servers, Dell’s Solutions Performance Analysis team performed a series of benchmarks
using the new processors then compared the results to those obtained using the Intel Xeon processor E5-
2600 product family, code named “Sandy Bridge-EP”.
Based on the results of performed testing, PowerEdge servers with E5-2600 v2 processors perform 23% to
50% better and have as much as 52% better power efficiency, compared to the same servers running E5-2600
processors.
Key findings
Power efficiency with E5-2600 v2
The PowerEdge R720 is the world’s most power-efficient server, achieving a 54% higher overall score
in the industry standard SPECpower_ssj2008 benchmark with two E5-2660 v2 processors.
The PowerEdge M620 is once again the world’s most power-efficient blade server, achieving a 51%
higher overall score in the industry standard SPECpower_ssj2008 benchmark with two E5-2660 v2
processors.
Performance with E5-2600 v2
The PowerEdge R720 server with two E5-2697 v2 processors supported 50% more virtual desktops
using the industry standard Login VSI benchmark.
The PowerEdge R620 server with two E5-2697 v2 processors achieved 48% higher gigaFLOPS
running the Linpack scientific-computing benchmark.
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
6
Methodology
To highlight the performance improvements gained by PowerEdge servers running E5-2600 v2 processors,
few configuration changes were made between those runs and the runs with E5-2600 processors that were
used for comparison. However, as PowerEdge servers with E5-2600 v2 processors support up to 16 DIMMs
running at 1866MT/s, that speed memory was used in most benchmarks, except as noted in Appendix A.
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
7
Integer performance
SPEC CPU2006 integer tests
The industry standard SPEC CPU2006 benchmark is described on SPEC.org as:
CPU2006 is SPEC's next-generation, industry-standardized, CPU-intensive benchmark suite, stressing
a system's processor, memory subsystem and compiler. SPEC designed CPU2006 to provide a
comparative measure of compute-intensive performance across the widest practical range of
hardware using workloads developed from real user applications.
The integer portion of the benchmark is particularly good at measuring a server’s ability to run general
business applications. In Figure 1, we see a 41% improvement in the SPECint_rate benchmark with E5-2600 v2
processors.
Figure 1: Performance improvement running SPECint_rate_base2006
938
666
0
200
400
600
800
1000
SPECint_rate_base2006
R620 with E5-2697 v2
R620 with E5-2690
41%
http://www.spec.org/cpu2006/
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
8
Even the SPECint_base_2006 benchmark, which only runs a single copy of each benchmark, benefits with the
upgrade to E5-2667 v2, as seen in Figure 2.
Figure 2: Performance improvement running SPECint_base2006
63.0
51.2
0.0
10.0
20.0
30.0
40.0
50.0
60.0
70.0
SPECint_base2006 (higher is better)
M620 with E5-2667 v2
M620 with E5-2680
23%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
9
Java server tests
The SPECjbb2005 and SPECjbb2013 are also useful for measuring a server’s ability to run general business
apps. As described on SPEC.org:
SPECjbb2005 (Java Server Benchmark) is SPEC's benchmark for evaluating the performance of server
side Java. …SPECjbb2005 evaluates the performance of server side Java by emulating a three-tier
client/server system (with emphasis on the middle tier).
SPECjbb2005 is a multithreaded benchmark, and as Figure 3 shows, scores 42% higher with E5-2697 v2.
Figure 3: Performance improvement running SPECjbb2005
2,120,679
1,493,529
0
500,000
1,000,000
1,500,000
2,000,000
2,500,000
SPECjbb2005 score (BOPS, higher is better)
M620 with E5-2697 v2
M620 with E5-2680
42%
http://www.spec.org/jbb2005/
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
10
SPECjbb2005 was retired on October 1, 2013, for the newer SPECjbb2013 benchmark, which is described as
follows:
A usage model based on a worldwide supermarket company with an IT infrastructure that handles a
mix of point-of-sale requests, online purchases and data-mining operations.
As Figure 4 shows, the performance improvement with SPECjbb2013 is 33%.
Figure 4: Performance improvement running SPECjbb2013
59,651
42,431
0
10,000
20,000
30,000
40,000
50,000
60,000
70,000
SPECjbb2013 score (SPECjbb2013-MultiJVM max-jOPS, higher is better)
R720 with E5-2697 v2
R720 with E5-2690
33%
http://www.spec.org/jbb2013/http://www.spec.org/jbb2013/
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
11
Floating point performance
SPEC CPU2006 floating point tests
Floating point performance is important to those running science, simulations and HPC workloads. SPEC
CPU2006 contains a suite of floating point tests which when in “rate” (multithreaded) mode, show a 39%
performance improvement with E5-2697 v2, as seen in Figure 5.
Figure 5: Performance improvement running SPECfp_rate_base2006
681
491
0
100
200
300
400
500
600
700
800
SPECfp_rate_base2006 (higher is better)
T620 with E5-2697 v2
T620 with E5-2690
39%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
12
SPECfp_base2006 runs a single copy of each benchmark. In Figure 6 we see a 26% performance
improvement with the E5-2667 v2 processor.
Figure 6: Performance improvement running SPECfp_base2006
104
82.8
0
20
40
60
80
100
120
SPECfp_base2006
M620 with E5-2667 v2
M620 with E5-2680
26%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
13
HPC performance tests
The Linpack benchmark is used to measure a system’s floating point processing power by solving linear
equations. We ran the Intel Optimized Linpack Benchmark against a single node using both generations of
processors. The E5-2697 v2 shows a 48% performance improvement, as seen in Figure 7.
Figure 7: Performance improvement running Linpack
516.4
349.7
0.0
100.0
200.0
300.0
400.0
500.0
600.0
Linpack performance (gigaFLOPS, higher is better)
R620 with E5-2697 v2
R620 with E5-2690
48%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
14
SPEC OMP2012 is another scientific computing benchmark, described on SPEC.org as:
SPEC OMP2012 is designed for measuring performance using applications based on the OpenMP 3.1
standard for shared-memory parallel processing. The benchmark suite includes 14 scientific and
engineering application codes, covering everything from computational fluid dynamics (CFD) to
molecular modeling to image manipulation.
Figure 8 shows that running E5-2697 v2 processors on the PowerEdge R720 provides a 38% uplift compared
to the E5-2690.
Figure 8: Performance improvement running SPEC OMP2012
6.90
5.01
0.00
1.00
2.00
3.00
4.00
5.00
6.00
7.00
8.00
SPEC OMP2012 (SPECompG_peak2012, higher is better)
R720 with E5-2697 v2
R720 with E5-2690
38%
http://www.spec.org/omp2012/
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
15
Memory subsystem performance
Many workloads benefit from greater memory bandwidth. Dell PowerEdge servers running E5-2600 v2
processors support up to 16 DIMMs running at 1866MT/s. In Figure 9, the STREAM memory bandwidth
benchmark shows a 27% improvement thanks to the faster memory and additional cores of the E5-2697 v2.
Figure 9: Performance improvement running STREAM
101.5
80.1
0
20
40
60
80
100
120
STREAM performance (Gigabytes per second, higher is better)
R720 with E5-2697 v2
R720 with E5-2690
27%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
16
Workloads performance
Database performance
To measure the performance improvements brought by E5-2600 v2 processors to database workloads, we
connected a PowerEdge R720 server to a Dell PowerVault™ MD1220 storage enclosure using a PERC H810
RAID controller. Four 400GB solid state drives were installed in the MD1220 and configured in RAID 10 to host
a Microsoft® SQL Server® 2012 SP1 database.
We installed the 64-bit version of Quest Benchmark Factory® for Databases v6.9.1 and used it to create a
230GB test database. We then ran the Benchmark Factory Online Transaction Processing (OLTP) workload.
Figure 10 shows the peak transactions per second (TPS) achieved by the R720 with both sets of processors.
For this workload, the E5-2697 v2 processors provide 21% improved performance.
Figure 10: Performance improvement running a database workload
2934
2432
0
500
1000
1500
2000
2500
3000
3500
SQL performance (transactions per second, higher is better)
R720 with E5-2697 v2
R720 with E5-2690
21%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
17
VDI
Administrators looking to maximize their investment in Virtual Desktop Infrastructure (VDI) often use the Login
VSI load testing tool to see how many virtual desktops a server can support while maintaining acceptable
response time. Using Login VSI set to the medium workload, the following results were generated.
Figure 11: Performance improvement running Login VSI
191
127
0
50
100
150
200
250
Login VSI (maximum sustained desktops, higher is better)
R720 with E5-2697 v2
R720 with E5-2690
50%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
18
Virtualization
Administrators running in virtualized environments can see in Figure 12 that virtualization performance on the
PowerEdge R720 is up to 42% greater when running VMware® VMmark® 2.5 with the E5-2697 v2 processor.
Figure 12: Performance improvement running VMmark 2.5
16.19
11.38
0
2
4
6
8
10
12
14
16
18
VMmark 2.5 score (higher is better)
R720 with E5-2697 v2
R720 with E5-2690
42%
http://www.vmware.com/products/vmmark/
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
19
Big data
The Apache™ Hadoop® software library is described by the Apache Hadoop Project Team as:
…a framework that allows for the distributed processing of large data sets across clusters of computers
using simple programming models.
Dell tests the performance of our servers using HiBench 2.2. While many of the sub-benchmarks of HiBench
are bound by storage performance thus seeing minimal improvement from processor upgrades, the K-Means
Clustering sub-benchmark was completed in 25% less time when the processors in our four-node C8220
cluster were upgraded to E5-2640 v2 processors and 1866MT/s DIMMs.
Figure 13: Performance improvement running HiBench K-Means Clustering benchmark
18,456
24,573
0
5,000
10,000
15,000
20,000
25,000
30,000
Hadoop K-Means Clustering (time to process baseline workload in seconds, lower is better)
4 x C8220 with E5-2640 v2
4 x C8220 with E5-264025%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
20
Power efficiency
The SPECpower_ssj2008 benchmark measures the performance of a server or servers running a server-side
Java workload, and measures the power consumed during the workload to produce a performance-per-watt
metric.
The greater power efficiency and improved performance of E5-2600 v2 processors leads to a 54% increase in
the PowerEdge R720’s overall score on the SPECpower_ssj2008 benchmark. The score of 8,458
SPECpower_ssj2008 overall ssj_ops/watt makes the PowerEdge R720 the world’s most power-efficient
server.
Figure 14: Power efficiency improvement of the R720 running the SPECpower_ssj2008 benchmark
8,458
5,506
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
R720 SPECpower_ssj2008 overall score (ssj_ops per watt, higher is better)
R720 with E5-2660 v2
R720 with E5-2660
54%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
21
The Dell PowerEdge M610 was the most power efficient blade server of the previous generation, with its score
of 3,348 SPECpower_ssj2008 overall ssj_ops/watt being the best among all blade servers running Intel Xeon
X5600 processors. Similarly, the PowerEdge M620 scored the highest among all blade servers using E5-2600
processors (4,982 SPECpower_ssj2008 overall ssj_ops/watt) and the PowerEdge M520 scored the highest
among all blade servers using E5-2400 processors (5,521 SPECpower_ssj2008 overall ssj_ops/watt).
Based on all blade SPECpower scores submitted to spec.org at the time of this whitepaper’s writing, the
PowerEdge M620 is once again the world’s most power-efficient blade server, with a score of 7,525
SPECpower_ssj2008 overall ssj_ops/watt with E5-2660 v2 processors.
Figure 15: Power efficiency improvement of the M620 running the SPECpower_ssj2008 benchmark
7,525
4,649
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
M620 SPECpower_ssj2008 overall score (ssj_ops per watt, higher is better)
16 x M620 with E5-2660 v2
16 x M620 with E5-2670
51%
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
22
Summary
Performance benefits with E5-2600 v2 processors will depend on the workload being run. Scientific
workloads that scale linearly with additional cores, such as Linpack, can see performance improvements of as
much as 48%. Likewise, the VDI benchmark Login VSI saw a 50% increase in the number of virtual desktops
supported.
Workloads with a heavy storage component benefit somewhat less from the additional processing power
brought by the upgrade of E5-2600 v2 processors. Even so, the IO bound Quest Benchmark Factory with SQL,
saw a 21% improvement in transactions per second running an OLTP workload.
Power-efficiency improvements gained from a move to E5-2600 v2 processors will likewise depend on the
workload being run. However, in the industry-standard SPECpower_ssj2008 benchmark, the gains were
impressive. The performance per watt of the R720 improved by 52%, making it the most power-efficient server
in the world. And, the performance per watt of the M620 improved by 51% compared to E5-2600, making it
the world’s most power-efficient blade server.
The E5-2600 v2 product family is available for purchase on all PowerEdge 12G servers that supported E5-
2600 processors. The enhanced performance and world record power efficiency make the new processors a
worthy enhancement to the already formidable 12G server lineup.
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
23
Appendix A — Test configurations
Table 1: Benchmark configurations
Benchmark PowerEdge Platform
E5-2600 v2 product family
E5-2600 v2 product family score
E5-2600 product family
E5-2600 product family score
Improvements with v2
SPECint_rate_base2006 R620 E5-2697 v2 938 E5-2690 666 41%
SPECint_base2006 M620 E5-2667 v2 63 E5-2680 51.2 23%
SPECfp_rate_base2006 T620 E5-2697 v2 681 E5-2690 491 39%
SPECfp_base2006 M620 E5-2667 v2 104 E5-2680 82.8 26%
STREAM R720 E5-2697 v2 101.5 E5-2690 80.1 27%
Linpack R620 E5-2697 v2 516.4 E5-2690 349.7 48%
SPECpower_ssj2008 (modular)
M620 E5-2660 v2 7,525 E5-2670 4,982 51%
SPECpower_ssj2008 (rack)
R720 E5-2660 v2 8,458 E5-2660 5,506 52%
SPECjbb2005 (modular) M620 E5-2697 v2 2,120,679 E5-2680 1,493,529 42%
SPECjbb2013 max-jOPS R720 E5-2697 v2 59,651 E5-2690 42,431 33%
Login VSI R720 E5-2697 v2 191 E5-2690 127 50%
SPEC OMP 2012 (peak) R720 E5-2697 v2 6.90 E5-2690 5.01 38%
VMmark 2.5 R720 E5-2697 v2 16.19 E5-2690 11.38 42%
SAP M620 E5-2697 v2 9,695 E5-2690 6,500 49%
SQL (Benchmark Factory)
R720XD E5-2697 v2 2,934 E5-2690 2,432 21%
Hadoop C8220 E5-2640 v2 18,456 E5-2640 24,573 33%
For SPECpower_ssj2008, we used 1600MT/s UDIMMs with E5-2600 v2 and 1333MT/s UDIMMs with
E5-2600
Performance and power efficiency of Dell PowerEdge servers with E5-2600 v2
24
For all other benchmarks, we used 1866MT/s RDIMMs with E5-2600 v2 and 1600MT/s RDIMMs with
E5-2600
For SQL workload testing with Quest Benchmark Factory, we used a PERC H710 mini with 2 x 200GB
SAS SSD RAID 1 for the operating system, and a PERC H810 connected to a PowerVault MD1220 with
4 x 400GB SAS SSD in RAID 10 for the database, and 4 x 300GB 10k SAS in RAID 10 for database logs
For the Hadoop K-Means Clustering test, we used the LSI 2008 controller and 12, 1TB SATA HDDs