Digital PDFs

EB-N0102-51

November 1992

11 pages

Original

81.9kB

Document:	Alpha AXP Workstation Family Performance Brief - OpenVMS
Order Number:	EB-N0102-51
Revision:
Pages:	11
Original Filename:	https://web-docs.gsi.de/~kraemer/COLLECTION/ftp.cuhk.edu.hk/pub/doc/digital/alpha-axp-openvms-workstation.pdf

OCR Text

Alpha AXP Workstation Family
Performance Brief - OpenVMS

DEC 3000 Model 500 AXP Workstation

DEC 3000 Model 400 AXP Workstation

INSIDE
Benchmark results:
• SPEC
Digital Equipment Corporation
November 20, 1992
Second Edition
EB-N0102-51

• LINPACK
• Dhrystone
• X11perf
• DN&R Labs CPU2

Alpha AXP Workstation Family
Performance Brief - OpenVMS
Introducing Digital’s Alpha AXP Workstation Family
This document presents the results of industry-standard benchmarks on the DEC 3000 Model
400 AXPTM and DEC 3000 Model 500 AXP workstations in the DEC OpenVMS AXPTM
operating system environment. Both workstations are members of Digital’s breakthrough
64-bit RISC architecture.
DEC 3000 Model 400 AXP Workstation
The DEC 3000 Model 400 AXP workstation is Digital’s entry-level, desktop workstation,
and it runs at a CPU clock speed of 133 MHz. This workstation allows for expansion of
memory, storage, I/O, and graphics.
The DEC 3000 Model 400 AXP workstation satisfies the performance needs of technical
users developing or deploying software and of commercial users doing financial analysis,
network management, publishing, and database services.
DEC 3000 Model 500 AXP Workstation
The DEC 3000 Model 500 AXP workstation is the most powerful system in Digital’s 64-bit
Alpha AXPTM workstation family, and it runs at a CPU clock speed of 150 MHz. The DEC
3000 Model 500 AXP workstation uses advanced CPU, TURBOchannel, and graphics
technologies. This workstation is available in a deskside or rackmountable configuration.
The DEC 3000 Model 500 AXP workstation is the system of choice for such
high-performance technical and commercial applications as mechanical CAD, scientific
analysis, medical imaging, animation and visualization, financial analysis, and insurance
processing.

Digital’s Alpha AXP Workstation Family Performance
The performance of the Alpha AXP workstation family was evaluated using
industry-standard benchmarks. These benchmarks allow comparison across vendors.
Performance characterization is one "data point" to be used in conjunction with other
purchase criteria such as features, service, and price. Features may include resource sharing
with VMSclusters, multi-vendor integration with Network Application Support (NAS), and
network management with DECmcc.
For more information on Digital’s Alpha AXP workstation family, please contact your local
Digital sales representative.

Digital Equipment Corporation

November 20, 1992

Alpha AXP Workstation Family
Performance Brief - OpenVMS
Table 1 Digital’s Alpha AXP Workstation Family Benchmark Results
Benchmark

DEC 3000 Model 400 AXP

DEC 3000 Model 500 AXP

SPECmark89

108.1

121.5

SPECint92
SPECfp92

65.3
112.2

74.3
126.0

LINPACK 64-bit Double-Precision
100X100 (MFLOPS)
1000x1000 (MFLOPS)

26.4
70.8

30.2
79.9

Dhrystone V1.1 (instructions/second) 228,310.0
V2.1 (instructions/second) 249,625.6

257,731.0
281,214.8

X11perf (2D Kvectors/second)
X11perf (2D Mpixels/second)

564.0
27.4

636.0
31.0

DN&R Labs CPU2 (MVUPs)

179.4

197.5

Notes: The performance information in this report is for guidance only. System
performance is highly dependent upon application characteristics. Individual work
environments must be carefully evaluated and understood before making estimates of
expected performance. This report simply presents the data, based on specified benchmarks.
Competitive information is based on the most current published data for those particular
systems and has not been independently verified.
We chose the competitive systems (shown with the Alpha AXP workstations in the following
charts and tables) based on comparable or close CPU performance and coupled with
comparable expandability capacity, mostly memory and disk. Although we do not present
price comparisons in this report, system price was a secondary factor in our competitive
choices.
The Alpha AXP performance information presented in this brief is the latest measured results
as of the date published. Digital has an ongoing program of performance engineering across
all products. As system tuning and software optimizations continue, Digital expects the
performance of its workstations to increase. As more benchmark results become available,
Digital will publish reports containing the new and updated benchmark data.

November 20, 1992

Digital Equipment Corporation

Alpha AXP Workstation Family
Performance Brief - OpenVMS
SPEC Benchmark Suites
SPEC (Standard Performance Evaluation Corporation) was formed to identify and create
objective sets of applications-oriented tests, which can serve as common reference points and
be used to evaluate performance across multiple vendors’ platforms.
SPEC Release 1
In October 1989, SPEC introduced SPEC Release 1, a benchmark suite that measures
CPU-intensive, single stream performance of uniprocessor systems. SPEC Release 1 consists
of ten portable programs similar to those found in technical environments. Four programs
are written in C and primarily test integer performance. The remaining six programs are
written in FORTRAN and measure floating-point performance. SPECmark89 is the metric
for this suite.
SPECmark89 represents the geometric mean of the ten benchmark SPECratios. The
SPECratio for a benchmark is the quotient derived from dividing the SPEC Reference Time
by a particular machine’s corresponding run time. The SPEC Reference Time is the time that
it takes a DEC VAX 11/780 to run each benchmark (in seconds).
Figure 1 SPECmark89 Benchmark Results

SPECmark89

160.0
146.8
140.0
126.2

121.5

120.0

108.1
100.0
87.6

92.3

86.6

80.0
69.0

73.7

70.2

60.0
40.0
20.0

0.0
DEC 3000/400 AXP
HP 9000/730
HP 9000/715/50
IBM RS 6000/350
IBM RS 6000/580
DEC 3000/500 AXP
HP 9000/750
HP 9000/735/755
IBM RS 6000/560
SGI INDIGO (R4000)

Digital Equipment Corporation

November 20, 1992

Alpha AXP Workstation Family
Performance Brief - OpenVMS
SPEC CINT92 and CFP92
In January 1992, SPEC announced the availability of the CINT92 and CFP92 benchmark
suites. CINT92, the integer suite, contains six real-world application benchmarks written in
C. The geometric mean of the suite’s six SPECratios is the SPECint92 figure. CFP92
consists of fourteen real-world applications; two are written in C and twelve in FORTRAN.
Five of the fourteen programs are single precision, and the rest are double precision.
SPECfp92 equals the geometric mean of this suite’s fourteen SPECratios.
CINT92 and CFP92 have different workload characteristics. Each suite provides
performance indicators for different market segments. SPECint92 is a good base indicator of
CPU performance in a commercial environment. SPECfp92 may be used to compare
floating-point intensive environments, typically engineering and scientific applications.
Figure 2 SPEC CINT92 Benchmark Results
SPECint92
100.0

80.0

80.0
74.3
65.3

59.1

60.0
52.0

57.6

53.3

51.1
43.2

40.0

36.5

35.4

20.0

0.0
DEC 3000/400 AXP
HP 9000/730
HP 9000/715/50
IBM RS 6000/350
IBM RS 6000/580
SUN SPARC 10-41
DEC 3000/500 AXP
HP 9000/750
HP 9000 735/755
IBM RS 6000/560
SGI INDIGO (R4000)

November 20, 1992

Digital Equipment Corporation

Alpha AXP Workstation Family
Performance Brief - OpenVMS

Figure 3 SPEC CFP92 Benchmark Results
SPECfp92

160.0
150.6
140.0
126.0

124.7

120.0
112.2
100.0

97.6
86.7

84.8

80.0
72.1

74.2
65.1
60.3

60.0

40.0

20.0

0.0
DEC 3000/400 AXP
HP 9000/730
HP 9000 715/50
IBM RS 6000/350
IBM RS 6000/580
SUN SPARC 10-41
DEC 3000/500 AXP
HP 9000/750
HP 9000 735/755
IBM RS 6000/560
SGI INDIGO (R4000)

Digital Equipment Corporation

November 20, 1992

Alpha AXP Workstation Family
Performance Brief - OpenVMS
LINPACK 100x100 and 1000x1000
LINPACK is a linear equation solver written in FORTRAN. LINPACK programs consist of
floating-point additions and multiplications of matrices. The LINPACK benchmark suite
consists of two benchmarks.
1.

100x100 LINPACK solves a 100x100 matrix of simultaneous linear equations. Source
code changes are not allowed so that the results may be used to evaluate the compiler’s
ability to optimize for the target system.

1000x1000 LINPACK solves a 1000x1000 matrix of simultaneous linear equations.
Vendor optimized algorithms are allowed.

The LINPACK benchmarks measure the execution rate in MFLOPS (millions of
floating-point operations per second). When running, the benchmark depends on
memory-bandwidth and gives little weight to I/O. Therefore, when LINPACK data fit into
system cache, performance may be higher.
Figure 4 LINPACK 100x100 and 1000x1000 Double-Precision Benchmark
Results
MFLOPS
100.0
100x100

84.0
79.9

80.0

1000x1000

70.8

100x100
61.0
60.0
23.7

1000x1000

23.7

40.8

38.1

40.0

31.3

30.2
26.4

22.4
18.8

20.0
13.2

N/A

0.0

N/A

DEC 3000/400 AXP
HP 9000/730
HP 9000/715/50
IBM RS 6000/350
IBM RS 6000/580
DEC 3000/500 AXP
HP 9000/750
HP 9000 735/755
IBM RS 6000/560
SUN SPARC 10-41

November 20, 1992

Digital Equipment Corporation

Alpha AXP Workstation Family
Performance Brief - OpenVMS
Dhrystone
Developed as an Ada program in 1984, the Dhrystone benchmark was rewritten in C in 1986.
It measures processor and compiler efficiency and is representative of systems programming
environments. Dhrystones are most commonly expressed in Dhrystone instructions per
second and in integer MIPS (millions of instructions per second). For V1.1, one Dhrystone
MIP equals the number of Dhrystone instructions per second performed by a DEC VAX
11/780 (1757 Dhrystone instructions/second).
Dhrystone V1 and V2 vary considerably. Version 1.1 contains sequences of code segments
that calculate results never used later in the program. These code segments are known as
"dead code." Compilers able to identify the dead code can eliminate these instruction
sequences from the program. These compilers allow a system to complete the program in
less time and result in a higher Dhrystones rating. Dhrystones V2 was modified to execute
all instructions.
Note: The Dhrystone benchmark is small and fits completely in most system caches. Level
of compiler optimization, as well as the particular hardware architecture, can affect results.
Figure 5 Dhrystone V1.1 Benchmark Results
Dhrystones/second
(in thousands)
300.0
281.2

250.0

249.6

V1.1

V2.1

V1.1

V2.0

257.7

228.3
215.5
200.0

194.2

192.4

150.0
133.7
121.6
100.0

133.7
122.3

131.6
109.4
97.0

106.5

50.0

N/A

0.0

N/A

DEC 3000/400 AXP
HP 9000/730
HP 9000/715/50
IBM RS 6000/350
SUN SPARC 10-41
DEC 3000/500 AXP
HP 9000/750
HP 9000/735/755
IBM RS 6000/560

Digital Equipment Corporation

November 20, 1992

Alpha AXP Workstation Family
Performance Brief - OpenVMS
X11perf Benchmark
X11perf tests various aspects of X server performance including simple 2D graphics,
window management functions, and X-specific operations. Other non-traditional graphics
include CopyPlane and various stipples and tiles.
X11perf employs an accurate client-server synchronization technique to measure graphics
operations’ completion times. X11perf tests both graphics primitive drawing speeds and
window environment manipulation.
Table 2 contains the two most commonly requested performance metrics from X11perf tests
for 2D graphics systems: X11perf 10-pixel line tests and X11perf Copy 500x500 from
pixmap to window tests. The 10-pixel line results are shown in units of 2D Kvectors/second
drawing rate, and the Copy 500x500 from pixmap to window are shown in units of 2D
Mpixels/second fill rate (1 Mpixel equals 1,048,576 pixels).
Table 2 X11perf Benchmark Results
Workstation

2D Kvectors/second

2D Mpixels/second

DEC 3000 Model 500 AXP

636.0

31.0

DEC 3000 Model 400 AXP

564.0

27.4

HP 9000 Model 735/755 CRX

971.0

HP 9000 Model 730 CRX

908.0

38.1

HP 9000 Model 720 CRX

868.0

22.8

HP 9000 Model 715/50 C

860.0

SGI Crimson Elan (R4000)

389.0

9.3

np = not published

November 20, 1992

Digital Equipment Corporation

Alpha AXP Workstation Family
Performance Brief - OpenVMS
DN&R Labs CPU2
DN&R Labs CPU2, a benchmark from Digital Review & News magazine, is a floating-point
intensive series of FORTRAN programs and consists of thirty-four separate tests. The
benchmark is most relevant in predicting the performance of engineering and scientific
applications. Performance is expressed as a multiple of MicroVAX II Units of Performance
(MVUPs).

Figure 6 DN&R Labs CPU2 Benchmark Results
MVUPs

250.0

197.5

200.0
179.4

150.0

100.0

88.2

86.8
71.4

50.0

0.0
DEC 3000/400 AXP

DEC 3000/500 AXP

HP 9000/730

IBM RS 6000/350

Digital Equipment Corporation

IBM RS 6000/560

November 20, 1992

Alpha AXP Workstation Family
Performance Brief - OpenVMS
References
System and Vendor

Sources

DEC 3000 Model 400 AXP Workstation
DEC 3000 Model 500 AXP Workstation
HP 9000 Model 720

All benchmarking performed by Digital Equipment Corporation.
All benchmarking performed by Digital Equipment Corporation.
X11perf benchmark results from Workstation Laboratories, Inc.,
Volume 13, Chapter 11 (4/1/91).
SPEC benchmark results reported by HP (11/10/92).
Dhrystone and LINPACK 100x100 benchmark results reported in "HP
Apollo 9000 Series 700 Performance Brief" (1/92).
LINPACK 1000x1000 reported by Dongarra, J., "Performance
of Various Computers Using Standard Linear Equations
Software" (9/28/92).
DN&R Labs CPU2 and X11perf benchmark results reported by
Workstation Laboratories, Inc., Volume 16, Chapter 12
(3/1/92).
SPEC benchmark results reported by HP (11/10/92).
Dhrystone and LINPACK 100x100 benchmark results reported in "HP
Apollo 9000 Series 700 Performance Brief" (1/92).
SPEC, LINPACK, Dhrystone, and X11perf benchmark results reported
in "HP Apollo 9000 Series 700 Workstation Systems
Performance Brief" (11/92).
SPEC benchmark results reported by IBM (9/92).
Dhrystone and DN&R Labs CPU2 benchmark results reported in
Workstation Laboratories, Inc., Volume 16, Chapter 20
(3/1/92).
LINPACK benchmark results reported by IBM (9/92) and Dongarra, J.,
"Performance of Various Computers Using Standard Linear
Equations Software" (9/28/92).
SPEC benchmark results reported in SPEC Newsletter (9/92).
LINPACK benchmark results reported by IBM (9/92).
X11perf benchmark results reported in Workstation Laboratories Inc.,
Volume 17, Chapter 21 (5/1/92).
SPEC benchmark results reported in SPEC Newsletter (9/92).
SPEC benchmark results reported in SPEC Newsletter (9/92).
LINPACK and Dhrystone benchmark results reported by SUN
(11/10/92).

HP 9000 Model 730

HP 9000 Model 750

HP 9000 Models 715/50 and 735, and 755

IBM RS 6000 Models 350 and 560

IBM RS 6000 Models 580 and 980
SGI Crimson Elan (R4000)
SGI INDIGO (R4000)
SUN SPARC 10 Model 41

Digital believes the information in this publication is accurate as of its publication date; such information is subject to change without notice.
The following are trademarks of Digital Equipment Corporation: AXP, Alpha AXP, the AXP logo, the AXP signature, DEC, OpenVMS, and
TURBOchannel.
HP is a registered trademark of Hewlett-Packard Company. IBM and RS 6000 are trademarks of International Business Machines Corporation.
SPEC, SPECratio, SPECmark89, SPECint92, and SPECfp92 are trademarks of Standard Performance Evaluation Corporation. X11 is a
trademark of The Massachusetts Institute of Technology. SUN and SPARC are trademarks of Sun Microsystems, Inc.
Printed in U.S.A. EB-N0102-51. Copyright © 1992 Digital Equipment Corporation. All Rights Reserved.

November 20, 1992

Digital Equipment Corporation