Digital PDFs

EC-N0763-51

April 1992

56 pages

Original

0.3MB

Document:	Digital's DECstation Family Performance Summary
Order Number:	EC-N0763-51
Revision:
Pages:	56
Original Filename:	https://web-docs.gsi.de/~kraemer/COLLECTION/DEC/decstation-Apr-92.pdf

OCR Text

Digital’s DECstation Family
Performance Summary

Version 2.1
April 1992
Order Number: EC-N0763-51

Digital Equipment Corporation
Maynard, Massachusetts
TM

_______________________________________________________
First Printing, April 1992
The information in this document is subject to change without notice and should not be construed as a commitment by Digital Equipment Corporation.
Digital Equipment Corporation assumes no responsibility for any errors that may appear in this document.
Any software described in this document is furnished under a license and may be used or copied only in accordance with the terms of such license. No responsibility is assumed for the use or reliability of software or equipment that is not supplied by Digital Equipment Corporation or its affiliated companies.
Restricted Rights: Use, duplication, or disclosure by the U.S. Government is subject to restrictions as set forth
in subparagraph (c)(1)(ii) of the Rights in Technical Data and Computer Software clause at DFARS 252.227
7013.

Copyright ©1992 Digital Equipment Corporation
All right reserved.
Printed in U.S.A.
The following are trademarks of Digital Equipment Corporation:
DEC, DECstation, DECwrite, the DIGITAL Logo, and ULTRIX.
The following are third-party trademarks:
AIX, IBM are trademarks of International Business Machines Corporation
Hewlett-Packard, HP, PC/AT are trademarks of Hewlett-Packard Company
IRIS 4D, IRIS Indigo are trademarks of Silicon Graphics Computer Systems
Microsoft is a registered trademark of Microsoft Corporation
NFS, SPARC, SPARCstation, SUN are trademarks of Sun Microsystems, Incorporated
Prestoserve is a registered trademark of Legato Systems, Inc.
SPEC is a trademark of the Standard Performance Evaluation Corporation
UNIX is a registered trademark of UNIX System Laboratories, Inc.
X11 is a trademark of The Massachusetts Institute of Technology
This document was prepared using DECwrite, Version 2.0

Contents
Part l Executive Summary ................................................................................................ 7
1

Introduction ....................................................................................................................... 8

Methodology ...................................................................................................................... 9

Summary of Relative Performance ................................................................................. 9

SPEC Benchmark Suite .................................................................................................. 17

4.1

Background .................................................................................................................... 17

4.2
4.2.1

Results and Conclusions ................................................................................................ 17
SPEC Ratios ................................................................................................................ 19

Dhrystone Integer Benchmark ....................................................................................... 21

5.1

Background .................................................................................................................... 21

5.2

Results and Conclusion .................................................................................................. 21

Whetstone Benchmark ................................................................................................... 24

6.1

Background .................................................................................................................... 24

6.2

Results and Conclusions ................................................................................................ 24

Linpack Benchmark ........................................................................................................ 26

7.1

Background .................................................................................................................... 26

7.2

Results and Conclusions ................................................................................................ 26

DR Labs CPU2 ................................................................................................................. 28

8.1

Background .................................................................................................................... 28

8.2

Results and Conclusions ................................................................................................ 28

Khornerstone Benchmarks ............................................................................................ 29

9.1

Background .................................................................................................................... 29

9.2

Results and Conclusions ................................................................................................ 29

SoftPC for ULTRIX .......................................................................................................... 30

10.1

Background .................................................................................................................... 30

10.2

Results and Conclusions ................................................................................................ 30

DECstations Graphic Options ........................................................................................ 31

2D Graphics X11perf Benchmarks ................................................................................ 32

12.1

Background .................................................................................................................... 32

Digital’s DECstation Family Performance Summary

iii

12.2

Results and Conclusions ................................................................................................ 32

3D Graphics Benchmarks .............................................................................................. 38

13.1

Background ................................................................................................................... 38

13.2

Results and Conclusions ................................................................................................ 38

Picture-Level Benchmarks ............................................................................................. 40

14.1

Background .................................................................................................................... 40

14.2

Results and Conclusions ................................................................................................ 41

A Test Configurations ....................................................................................................... A-1
B References ........................................................................................................................ B-1
Figures
Figure 4-1:
Figure 4-2:
Figure 4-3:
Figure 5-1:
Figure 5-2:
Figure 5-3:
Figure 6-1:
Figure 6-2:
Figure 7-1:
Figure 7-2:
Figure 8-1:
Figure 9-1:
Figure 12-1:
Figure 12-2:
Figure 12-3:
Figure 12-4:
Figure 12-5:
Figure 12-6:
Figure 14-1:

SPECmark Benchmark Results .......................................................................... 17
SPEC Integer Benchmark Results ....................................................................... 18
SPEC Floating Point Benchmark Results ............................................................ 18
MIPS Results from Dhrystones V1.1 Benchmark ................................................ 21
Dhrystone V1.1 Benchmark Results (Dhrystones/second) .................................. 22
Dhrystones V2.0 and V2.1 Benchmark Results (Dhrystones/second) ................. 23
Whetstones Single-precision Benchmark Results ............................................... 24
Whetstones Double-Precision Benchmark Results .............................................. 25
Linpack Single-precision Benchmark Results ...................................................... 26
Linpack Double-precision Benchmark Results ..................................................... 27
DR Labs CPU2 Benchmark Results .................................................................... 28
Khornerstones Benchmark Results ...................................................................... 29
Personal DECstations and Competitors 2D Graphics X11perf Benchmark
Results ................................................................................................................. 35
DECstation 5000/120, 125, & 133 and Competitors 2D Graphics X11perf
Benchmark Results .............................................................................................. 35
DECstation 5000/200 & 240 and Competitors 2D Graphics X11perf Benchmark
Results ................................................................................................................. 36
Personal DECstations and Competitors 2D Fill Area Graphics X11perf Benchmark
Results ................................................................................................................. 36
DECstation 5000/120, 125, & 133 and Competitors 2D Fill Area Graphics X11perf
Benchmark Results .............................................................................................. 37
DECstation 5000/200 & 240 and Competitors 2D Fill Area Graphics X11perf
Benchmark Results .............................................................................................. 37
PLB Benchmarks Results .................................................................................... 42

Tables
Table 3-1:
Table 3-2:
Table 3-3:
Table 3-4:

DECstations and Competitive Workstations CPU and FPU Benchmark Results 10
DECstations and Competitive Workstations Graphics Benchmark Results ......... 11
DECstations and Competitive Workstations SoftPC Benchmark Results ............ 14
Key to Graphs ...................................................................................................... 16

Digital’s DECstation Family Performance Summary

Table 4-1:
Table 4-2:
Table 10-1:
Table 11-1:
Table 12-1:
Table 12-2:
Table 13-1:

DECstations SPEC Ratios (SPEC Reference Time/Elapsed Time) .................... 19
Competitors SPEC ratios (SPEC Reference Time/Elapsed Time) ...................... 20
SoftPC Benchmarks Results ................................................................................ 30
Key to DECstations Graphic Options .................................................................. 31
DECstations 2D Graphics X11perf Benchmark Results ...................................... 33
Competitors 2D Graphics X11perf Benchmark Results ....................................... 34
3D Graphics Benchmark Results ......................................................................... 39

Digital’s DECstation Family Performance Summary

Part l
Executive Summary

Introduction

Introduction
Results from a large selection of benchmarks are needed to understand system performance. A workstation application depends on all aspects of the system including
the CPU, floating point processor, I/O, and graphics. Each application, however,
uses these features to different degrees. Therefore, understanding the nature of your
application environment and matching this information with benchmark results are
crucial for accurate performance prediction. Digital’s DECstation Family Performance Summary provides performance information on the DECstation family of systems and competitive computers running standard industry benchmarks.
This revision of Digital’s DECstation Family Performance Summary contains updated and new graphic benchmark results for both the DECstation family of systems
and comparable, competitive systems.
Digital’s DECstation Family Performance Summary is a technical reference document for Digital sales support personnel, customers, and other individuals who need
to understand performance characteristics of the DECstation family.
The benchmarks results appear in the following order:

•

SPEC Benchmarks

•

Dhrystone Benchmarks

•

Whetstone Benchmarks

•

Linpack Benchmarks

•

DR Labs CPU2 Benchmark

•

Khornerstone Benchmark

•

SoftPC Benchmarks

•

2D Graphics X11perf Benchmarks

•

3D Graphics Benchmark

•

Picture-Level Benchmarks

Digital’s DECstation Family Performance Summary

Summary of Relative Performance

Methodology
Wherever possible, we executed the same benchmark code on all our systems.
Competitive information is based on the most current, previously published data for
those particular systems and has not been independently verified. See Appendix A,
titled Test Configurations, for configuration, operating system and compiler versions, and source of each benchmark result.

Summary of Relative Performance
Tables 3-1 through Table 3-3 contain the benchmarking performance results of the
DECstation family of systems and competitive systems.

Digital’s DECstation Family Performance Summary

14.0
28.0

16.4

19.3

25.5

23.5

32.4

13.5

20.3

24.4

25.0

11.0

10.3

34.6

49.7

59.5

25.9

43.4

71.4

26.3

DECstation 5000/125

DECstation 5000/133

DECstation 5000/200

DECstation 5000/240

SPARCstation IPC

SPARCstation ELC

SPARCstation IPX

SPARCstation 2

HP 9000/425t

HP 9000/425e

HP 9000/705

HP 9000/710

HP 9000/720

IBM RS/6000 220

IBM RS/6000 320H

IBM RS/6000 350

SGI Personal Iris 4D/25G 12.2

31.1

DECstation 5000/120

SGI 4D/35

SGI Indigo

Table current as of April 1992

36.2

19.1

28.4

33.4

11.1

112.3

68.8

33.7

78.5

62.4

43.4

9.3

10.3

27.4

26.5

22.0

14.0

35.8

26.7

29.1

21.7

18.4

21.7

na = not available

23.6

21.8

17.5

39.5

35.4

24.6

12.2

12.3

21.7

18.0

12.8

27.9

19.5

20.9

16.1

13.8

15.7

54945

60240

28901

106382

65789

64102

100149

94339

61495

58139

45454

50075

46875

40540

27585

75557

47920

60475

47090

38116

46927

48544

56370

95420

57274

55127

87000

58574

39185

38760

35587

43478

25063

22831

64103

38168

49020

38759

32051

38760

31847

Dhrystones
per second
(Version 2)

Refer to Appendix A for test configurations

31.27

34.29

16.45

60.55

37.44

36.48

57.00

53.69

35.00

33.09

25.87

28.50

26.68

23.07

15.70

43.00

27.27

34.42

26.80

21.69

26.70

37951

DECstation 5000/25

21.60

DECstation 5000/20

18.4

16.3

Workstation
13.5

SPEC
MIPS
Dhrystones
SPEC- SPEC Floating (Dhrystone per second
mark Integer Point
Version 1.1) (Version 1.1)

22676

24331

12903

65789

25000

17544

56180

55556

10858

4417

19920

27778

23148

10204

42812

25679

33292

25627

20465

25860

20985

Whetstone
Single
(WIPS)

17921

19194

9615

65789

27778

18868

48310

34364

10753

4112

14641

19120

14663

6369

34457

20899

26724

20597

16560

20882

16865

Whetstone
Double
(WIPS)

4.30

5.11

2.94

15.49

10.38

4.79

22.90

20.94

1.69

6.10

4.34

3.61

3.20

10.80

6.81

8.79

6.67

5.31

6.60

5.32

Linpack
Single
Precision
(MFLOPS)

3.15

3.99

1.42

13.99

8.57

5.22

17.20

9.76

8.00

1.60

1.62

4.20

2.65

2.20

1.70

6.04

3.73

5.93

3.01

2.56

2.80

2.44

Linpack
Double
Precision
(MFLOPS)

26.14

29.56

16.15

71.47

28.07

55.51

48.43

7.92

8.47

21.60

22.24

18.52

10.64

47.16

28.61

37.10

26.81

22.00

27.71

22.78

DR
Labs
CPU2
(MVUPS)

Table 3-1: DECstations and Competitive Workstations CPU and FPU Benchmark Results

92032

86327

20727

88038

83582

62870

44573

42881

42753

14512

27142

29111

23909

19030

50949

27604

39606

33731

29494

32822

26689

Total
Khornerstones

Table current as of April 1992
Refer to Appendix A for test configuration

434.0

339.0

158.0

298.0

124.0

434.0

338.0

156.0

284.0

114.0

434.0

338.0

137.0

272.0

108.0

339.0

156.0

285.0

183.0

137.0

153.0

2D Vectors
X11perf.
10-pixel line
(Kvectors/sec.)²

na=not available or not applicable
nr=not reported

12.3

DECstation 5000/133 PXGT+

5.1

DECstation 5000/133 TX
18.3

14.8

DECstation 5000/133 HX

DECstation 5000/133 PXG+

8.1

DECstation 5000/133 MX

12.3

DECstation 5000/125 PXGT+

4.3

DECstation 5000/125 TX
18.3

14.6

DECstation 5000/125 HX

DECstation 5000/125 PXG+

8.0

DECstation 5000/125 MX

12.3

DECstation 5000/120 PXGT+

4.1

DECstation 5000/120 TX
18.3

14.6

DECstation 5000/120 HX

DECstation 5000/120 PXG+

7.9

18.3

DECstation 5000/120 MX

Personal DECstation 5000/25 PXG+

4.3

Personal DECstation 5000/25 TX

6.4

Personal DECstation 5000/25 board
14.7

4.1

Personal DECstation 5000/20 TX

Personal DECstation 5000/25 HX

5.7

Personal DECstation 5000/20 board

Workstation

2D Fill Area
X11perf
Copy 500x500
from pixmap to
windows
(Mpixels/sec.)¹

405

376

313

310

307

312

3D Graphics
3D Vectors
(Kvectors/sec.)

12.2:na

11.3:na

12.2:na

11.2:na

12.1:na

11.2:na

13.7:na

13.2:na

12.5:na

11.9:na

11.4:na

10.7:na

11.9:na

head

shuttle

17.1:na 21.3:na 21.1:na

16.1:na 20.9:na 19.9:na

17.0:na 21.3:na 21.1:na

16.2:na 20.9:na 20.0:na

17.0:na 21.3:na 21.1:na

16.2:na 20.9:na 20.0:na

sys_chassis cyl_head

GPC Picture-Level Benchmarks

³ Kpolygons=1,000 polygons

pc_board

¹ Mpixels=1,048,576 pixels
² Kvectors=1,000 vectors

102

3D Graphics
3D Polygons
(Kpolygons/sec.)³

Table 3-2: DECstations and Competitive Workstations Graphics Benchmark Results

20.3
30.5
8.1

DECstation 5000/240 MX

DECstation 5000/240 HX

DECstation 5000/240 TX

9.7
na
8.3
na
na
na
na

SPARCstation IPX~

SPARCstation IPX-SunPHIGS

SPARCstation 2~

SPARCstation 2 GX-SunPHIGS

SPARCstation 2 GXplus-SunPHIGS

SPARCstation 2 GS-SunPHIGS

SPARCstation 2 GT-SunPHIGS

HP 9000/425t Personal VRX~

HP 9000/425e~
Table current as of April 1992
Refer to Appendix A for test configuration

69.1

23.3

205.0

217.0

29.3

58.2

445.0

345.0

278.0

621.0

248.0

445.0

345.0

510.0

162.0

(Kvectors/sec.)²

2D Vectors
X11perf.
10-pixel line

436

401

434

400

(Kvectors/sec.)

3D Graphics
3D Vectors

8.8
49.0
na
na=not available or not applicable
~ X11perf results from Workstation Laboratories, Inc.

1.5

17.9

SPARCstation ELC~

HP 9000/425t~

5.1

SPARCstation IPC~

12.3

DECstation 5000/200 PXGT+

DECstation 5000/240 PXGT+

18.5

DECstation 5000/200 PXG+

18.5

30.3

DECstation 5000/200 HX

DECstation 5000/240 PXG+

14.0

(Mpixels/sec.)¹

DECstation 5000/200 MX

Workstation

2D Fill Area
X11perf
Copy 500x500
from pixmap to
windows

9.7: 9.7

3.6: 3.6

14.0:14.0

14.6:14.6

14.3:14.3

12.7:na

11.6:na

12.6:na

11.6:na

¹ Mpixels=1,048,576 pixels

² Kvectors=1,000 vectors

18.6:18.6

4.0: 4.0

10.7:10.7

10.8:10.8

10.7:10.7

15.3:na

13.8:na

14.9:na

13.6:na

21.9:na

20.8:na

shuttle

6.2: 6.2

9.2: 9.2

21.7:na 21.9:na

21.3:na 20.7:na

21.7:na

21.3:na

head

³ Kpolygons=1,000 polygons

15.1:15.1 29.3:29.3 17.2:17.2

7.2: 7.2

17.6:na

16.7:na

17.7:na

16.8:na

sys_chassis cyl_head

GPC Picture-Level Benchmarks
pc_board

na
nr
nr=not reported

106

(Kpolygons/sec.)³

3D Graphics
3D Polygons

Table 3-2: DECstations and Competitive Workstations Graphics Benchmark Results (cont.)

3.4

HP 9000/720 CRX24Z-HP-PHIGS

IBM RS/6000 220 POWER GTOgraPHIGS

IBM RS/6000 320H POWER GTOgraPHIGS~

IBM RS/6000 350 POWER GTOgraPHIGS

SGI 4D/35

nr=not reported

na=not available or not applicable

18.8:41.0

18.8:35.9

18.5:32.4

28.4:33.0

29.7:34.3

27.6:31.6

24.2:26.8

pc_board

Refer to Appendix A for test configuration

(Kpolygons/sec.)³

3D Graphics
3D Polygons

np=not possible

(Kvectors/sec.)

3D Graphics
3D Vectors

~ X11perf results from Workstation Laboratories, Inc.

141.0

11.3

77.4

868.0

492.0

(Kvectors/sec.)²

2D Vectors
X11perf.
10-pixel line

Table current as of April 1992

8.6

HP 9000/720 CRX24-HP-PHIGS

SGI 4D/RPC Indigo~

22.8

HP 9000/710 C-HP-PHIGS

HP 9000/720 CRX-HP-PHIGS~

3.4

(Mpixels/sec.)¹

HP 9000/710~

Workstation

2D Fill Area
X11perf
Copy 500x500
from pixmap to
windows

31.5:na

29.5:na

20.9:38.0

21.5:40.3

21.1:37.1

20.1:34.3

sys_chassis

32.6:42.5

14.4:16.0

14.3:15.8

head

np:25.3

np:24.2

np:22.4

21.1:na

15.7:na

15.6:na

shuttle

³ Kpolygons=1,000 polygons

² Kvectors=1,000 vectors

¹ Mpixels=1,048,576 pixels

11.3:17.4 27.4:35.5

11.5:17.4 27.2:34.2

11.3:17.4 27.0:31.8

31.9:na

18.9:na

11.4:na

cyl_head

GPC Picture-Level Benchmarks

Table 3-2: DECstations and Competitive Workstations Graphics Benchmark Results (cont.)

13.1

DECstation 5000/200 CX

Table current as of April 1992

Refer to Appendix A for test configuration

6.9

13.0

DECstation 5000/125 PXG

IBM PC/AT (*MHz)

9.8

13.1
12.5
13.0

DECstation 5000/120 MX

DECstation 5000/25 board @66Hz
DECstation 5000/25 HX @66Hz
DECstation 5000/25 HX @72Hz

Workstations

SoftPC
Norton SI

1761

3479

2517

2528

2806
2750
3193

SoftPC
Dhrystone

1.0

3.3

3.1

2.5

3.2
3.1
3.2

SoftPC
PC Magazine’s
Bench V4.0

Table 3-3: DECstations and Competitive Workstations SoftPC Benchmark Results

Part II
Benchmark Results

Summary of Relative Performance

We ran a variety of benchmarks on a number of DECstations. The benchmarks
were primarily CPU-intensive or graphics-intensive or both. Results from our
benchmarking activity as well as benchmark results of competitive workstations are
presented in this section. We encourage the reader to carefully consider the appropriateness of mapping these results into their own environments.
The table appearing on this page explains the system abbreviations used in the
graphs. Specific configuration details and source of benchmark results are presented in Appendix A. References are listed in Appendix B.

Table 3-4: Key to Graphs

Key to Graphs
Abbreviation

Full Product Description

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425t PVRX

Digital Personal DECstation 5000 Model 20
Digital Personal DECstation 5000 Model 25
Digital DECstation 5000 Model 120
Digital DECstation 5000 Model 125
Digital DECstation 5000 Model 133
Digital DECstation 5000 Model 200
Digital DECstation 5000 Model 240
Sun SPARCstation IPC
Sun SPARCstation ELC
Sun SPARCstation IPX
Sun SPARCstation 2
Hewlett-Packard 9000 Model 425t
Hewlett-Packard 9000 Model 425t Personal
VRX
Hewlett-Packard 9000 Model 425e
Hewlett-Packard 9000 Model 705
Hewlett-Packard 9000 Model 710
Hewlett-Packard 9000 Model 720
IBM AT 80286
IBM RISC System/6000 POWERstation 220
IBM RISC System/6000 POWERstation 320H
IBM RISC System/6000 POWERstation 350
Silicon Graphics Personal Iris 4D/25G
Silicon Graphics 4D/35
Silicon Graphics 4D/RPC Indigo

HP 9000/425e
HP 9000/705
HP 9000/720
HP 9000/720
IBM AT 80286
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

4.1

Background
Systems Performance Evaluation Cooperative (SPEC) is a nonprofit organization
formed to develop a standard suite of benchmark programs that characterize system
performance. Digital is a member of SPEC and endorses its goals.

The release 1.2 suite consists of ten compute intensive codes/programs. The performance metric used by SPEC is elapsed time. Four of the ten programs are written
in C and are classified as compute intensive. The geometric mean of these make up
the SPEC metric called SPECint. The other six programs are written in FORTRAN
and are floating point intensive. The geometric mean of these make up the SPECfp
metric. SPECmark is the geometric mean of all ten elapsed times normalized to the
VAX 11/780.

4.2

Results and Conclusions
Shown in the following three charts are the SPEC R1.2 Benchmark Suite.

Figure 4-1:

SPECmark Benchmark Results
16.3
19.1
16.4
19.3
25.5
23.5

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/705
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

32.4
13.5
20.3
24.4
25.0

11.0
10.3
34.6
49.7
59.5
25.9
43.4

71.4
12.2
31.1
26.3
0.0

20.0

40.0

60.0

80.0

100.0

SPECmark

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

Figure 4-2:

SPEC Integer Benchmark Results
13.5
15.7
13.8
16.1
20.9
19.5

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 7000/705
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

27.9
12.8
18.0
21.7
21.7
12.3
12.2
24.6
35.4
39.5
17.5
21.8
36.2
14.0
28.0
23.6
0.0

20.0

40.0

60.0

80.0

100.0

SPECint

Figure 4-3:

SPEC Floating Point Benchmark Results
18.4
21.7
18.4
21.7
29.1
26.7
35.8
14.0
22.0
26.5
27.4
10.3
9.3
43.4

62.4
78.5
33.7
68.8
112.3
11.1
33.4
28.4
0.0

20.0

40.0

60.0

80.0

100.0

120.0

140.0

SPECfp

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

4.2.1

SPEC Ratios
The SPEC Ratio for a benchmark is the quotient derived from dividing the SPEC
reference time by a particular machine’s corresponding run time. For Release 1.2,
the SPEC reference time is the time (in seconds) that it takes a VAX 11/780 machine to run each particular benchmark in the suite.

Table 4-1: DECstations SPEC Ratios (SPEC Reference
Time/Elapsed Time)

Benchmark
No. & Name

001.gcc
008.espresso
013.spice 2g6
015.doduc
020.nasa7
022.li
023.eqntott
030.matrix300
042.fpppp
047.tomcatv

Type

INT¹
INT
FP²
FP
FP
INT
INT
FP
FP
FP

DS
5000/20

DS
DS
5000/25 5000/120

DS
5000/125

DS
5000/133

DS
5000/200

DS
5000/240

11.0
14.2
9.7
15.6
17.5
15.5
13.6
47.4
18.4
17.0

12.2
17.4
11.4
18.7
19.9
17.5
16.5
57.4
22.1
19.5

13.7
17.4
11.4
18.8
20.0
17.4
16.4
57.0
21.9
19.5

16.8
22.7
14.8
25.2
25.4
22.4
22.3
82.7
28.4
27.1

18.1
18.3
14.3
20.8
27.7
23.1
18.7
66.5
25.1
26.4

23.3
28.7
16.5
30.8
32.0
32.5
28.0
99.0
38.4
33.8

12.1
14.2
9.7
15.4
17.5
15.6
13.6
47.5
18.3
17.1

¹ Integer benchmark written in C
² Floating-point benchmark written in FORTRAN

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

Table 4-2: Competitors SPEC ratios (SPEC Reference
Time/Elapsed Time)

Benchmark
No. & Name

SPARC- SPARC- SPARC- HP
HP
station station station 9000/ 9000/
Type ELC
IPX
2
705
710

001.gcc
INT¹
008.espresso INT
013.spice 2g6 FP²
015.doduc
FP
020.nasa7
FP
022.li
INT
023.eqntott
INT
030.matrix300 FP
042.fpppp
FP
047.tomcatv FP

16.6
18.0
13.4
14.0
23.6
19.0
18.4
67.7
19.1
19.7

19.9
21.7
16.1
16.6
28.4
23.0
22.2
81.5
23.0
23.8

20.0
21.7
16.5
18.2
29.1
23.1
22.3
82.6
23.8
24.9

19.6
28.3
22.9
29.7
35.6
25.4
26.0
190.9
32.8
43.9

29.0
39.8
33.2
42.5
51.7
36.7
37.2
277.6
45.3
64.1

HP
9000/
720

IBM
6000/
220

IBM
IBM
6000/ 6000/
320H 350

SGI
Indigo

36.0
43.4
44.5
47.6
64.5
37.7
41.2
323.2
78.5
66.4

14.6
19.7
14.6
18.1
32.3
14.7
22.0
191.5
25.7
34.7

19.9
32.6
20.5
34.1
26.4
42.6
29.5
48.9
81.3 129.5
21.3
35.7
25.9
43.2
407.7 675.4
54.1
87.8
76.1 124.9

23.5
22.8
19.4
22.6
30.5
25.8
22.3
86.5
19.1
23.6

¹ Integer benchmark written in C
² Floating-point benchmark written in FORTRAN

Digital’s DECstation Family Performance Summary

Dhrystone Integer Benchmark

5.1

Background
The Dhrystone benchmark was introduced in an ADA program in 1984 by Reinhold
P. Weicker. It has since been translated into C and TURBO PASCAL.

This synthetic benchmark1 measures processor and compiler efficiency. Its emphasis is on the type of data and operations encountered in a system rather than numerical programming. Dhrystones is CPU-intensive.

Dhrystones are most commonly expressed in Integer MIPS (Millions of Instructions
Per Second) where 1 MIP is the number of Dhrystones per second that can be performed by a VAX 11/780 (1757 Dhrystones/second). Please note that not all vendors base their MIPS rating on the Dhrystone benchmark.
5.2

Results and Conclusion
Following are the results of the Dhrystone benchmark tests.

Figure 5-1:

MIPS Results from Dhrystones V1.1 Benchmark
21.60
26.70
21.69
26.80

34.42
27.27
43.00
15.70
23.07
26.68
28.50
25.87
33.09
35.00
53.69
57.00
36.48
37.44
60.55
16.45
34.29
31.27
0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

MIPS
1 Synthetic benchmarks are mostly loops which time different system calls and are not based on actual applications.

Digital’s DECstation Family Performance Summary

Dhrystone Integer Benchmark

Dhrystone V1.x and V2.x 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
then eliminate those instruction sequences from the program. These compilers allow
a system to complete the program in less time and result in a higher Dhrystone rating. Dhrystone V2.x has been modified to execute all instructions.
Dhrystones V1.1 results are shown in the following graph.

Figure 5-2: Dhrystone V1.1 Benchmark Results
(Dhrystones/second)
DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/245e
HP 9000/705
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

37951
46927
38116
47090

Dhrystones/second V1.1

60475
47920
75557

27585
40540
46875
50075
45454
58139
61495
94339
100149

64102
65789
106382
28901
60240
54945
0

20000

40000

60000

80000

100000

120000 140000

Dhrystones/second

Digital’s DECstation Family Performance Summary

Dhrystone Integer Benchmark

Note: SGI 4D/25G V2 and HP 9000/705 Dhrystones V2.x results were not available.

Figure 5-3: Dhrystones V2.0 and V2.1 Benchmark Results
(Dhrystones/second)

DS 5000/20

Dhrystones/second V2

31847

DS 5000/25

38760

DS 5000/120

32051

DS 5000/125

38759

DS 5000/133

49020

DS 5000/200

38168

DS 5000/240

64103

SPARCstation IPC

22831

SPARCstation ELC

25063

SPARCstation IPX

43478

SPARCstation 2

35587

HP 9000/425t

38760

HP 9000/245e

39185

HP 9000/710

58574

HP 9000/720

87000

IBM RS/6000 220

55127

IBM RS/6000 320H

57274

IBM RS/6000 350

95420
56370

SGI 4D/35

48544

SGI Indigo

20000

40000

60000

80000

100000

120000 140000

Dhrystones/second

Digital’s DECstation Family Performance Summary

Whetstone Benchmark

6.1

Background
The Whetstone benchmark was developed in Great Britain’s National Physical
Laboratory in 1970. This synthetic benchmark was designed to represent small engineering/scientific programs.
The Whetstone benchmark has been implemented in single-precision and doubleprecision FORTRAN programs, each arranged to defeat most compiler
optimizations. The results are measured in WIPS (Whetstone Instructions Per Second).

6.2

Results and Conclusions
The single- and double-precision Whetstone benchmark results are shown in the
following two graphs.
Note: HP 9000/705 Whetstone benchmark results were not available.

Figure 6-1:

Whetstones Single-precision Benchmark Results
20985
25860
20465
25627
33292
25679
42812

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

10204
23148
27778
19920
4417
10858
55556
56180
17544
25000
65789
12903
24331
22676
0

20000

40000

60000

80000

100000

120000

Single-precision WIPS

Digital’s DECstation Family Performance Summary

Whetstone Benchmark

Figure 6-2:

Whetstones Double-Precision Benchmark Results

16865
20882
16560
20597
26724
20899

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

34457
6369
14663
19120
14641
4112
10753
34364
48310
18868
27778
65789
9615
19194
17921

10000

20000

30000

40000

50000

60000

70000

80000

Double-precision WIPS

Digital’s DECstation Family Performance Summary

Linpack Benchmark

7.1

Background
Developed at Argonne National Laboratories, Linpack is a FORTRAN benchmark
that solves a 100x100 system of linear equations. This benchmark is widely used to
compare the performance of mathematical and scientific applications where floating
point computations are prevalent. When running, this benchmark gives little weight
to I/O.

The results are measured in millions of floating point operations per second
(MFLOPS). Both single-precision and double-precision operations are reported.
7.2

Results and Conclusions
The following two charts contain Linpack MFLOPS results.
Note: HP 9000/705 single-precision Linpack results were not available.

Figure 7-1:
DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

Linpack Single-precision Benchmark Results
5.32
6.60
5.31
6.67
8.79
6.81
10.80
3.20
3.61
4.34
6.10
1.69
1.69
20.94
22.90
4.79
10.38
15.49
2.94
5.11
4.30

0.00

5.00

10.00

15.00

20.00

25.00

30.00

Single-precision MFLOPS

Digital’s DECstation Family Performance Summary

Linpack Benchmark

Figure 7-2:

Linpack Double-precision Benchmark Results

2.44
2.80
2.56
3.01

5.93
3.73
6.04
1.70
2.20
2.65
4.20
1.62
1.60
8.00
9.76
17.20
5.22
8.57
13.99
1.42
3.99
3.15
0.00

5.00

10.00

15.00

20.00

25.00

Double-precision MFLOPS

Digital’s DECstation Family Performance Summary

DR Labs CPU2

8.1

Background
The DR Labs CPU2 benchmark, a public domain benchmark from Digital Review
Magazine, is a floating-point intensive series of FORTRAN programs that include
thirty-four separate tests. The results of the CPU2 suite reflect the raw compute
speed of a processor, the efficiency of a system’s FORTRAN compiler, and to a
lesser degree, the speed of memory access. A system’s I/O capabilities do not affect
the results. 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).
8.2

Results and Conclusions
The results of the DR Labs CPU2 benchmarks are shown in the following graph.
Note: HP 9000/705 and IBM RS/6000 320H MVUPs results were not available.

Figure 8-1:

DR Labs CPU2 Benchmark Results
22.78
27.71
22.00
26.81

DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

37.10
28.61
47.16
10.64
18.52
22.24
21.60
8.47
7.92
48.43
55.51
28.07
71.47
16.15
29.56
26.14
0.00

10.00

20.00

30.00

40.00

50.00

60.00

70.00

80.00

MVUPs

Digital’s DECstation Family Performance Summary

Khornerstone Benchmarks

9.1

Background
The Khornerstone and Khornerstone2 benchmarks were developed by Workstation
Laboratories, Irving, TX. Consisting of 21 separate tests, the benchmarks rate the
overall performance of a system. The tests include both public domain (e.g., Whetstone, Dhrystone, Sieve, etc.) and proprietary routines. The purpose of these tests is
to measure single-user loads on a system and to provide one number representing
that load condition.

The Khornerstone and Khornerstone2 benchmarks differ in that the Khornerstone2
test runs 10 times as many repetitions of each test.
9.2

Results and Conclusions
The following graph has the Khornerstone and Khornerstone2 benchmark results.
Note: HP 9000/705 Khornerstone results were not available. HP 9000/425e
Khornerstone results were recorded using large buffer cache.

Figure 9-1:
DS 5000/20
DS 5000/25
DS 5000/120
DS 5000/125
DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720
IBM RS/6000 220
IBM RS/6000 320H
IBM RS/6000 350
SGI 4D/25G
SGI 4D/35
SGI Indigo

Khornerstones Benchmark Results
26689
32822
29494
33731
39606
27604
50949

19030
23909
29111
27142
14512
42753
42881
44573

62870
83582
88038

20727

20000

40000

60000

86327
92032
80000
100000

120000

Total Khornerstones

Digital’s DECstation Family Performance Summary

SoftPC for ULTRIX

10.1

Background
DEC SoftPC for ULTRIX provides the RISC user the ability to operate DOS based
software. The SoftPC product emulates an IBM PC/AT system (real mode only) using standard hardware and software. A DOS application will function in the SoftPC
environment as it functions on an IBM PC/AT so long as it does not require special
hardware, protected mode, himem.sys, or copy protected diskette.

10.2

Results and Conclusions
The performance of the DEC SoftPC facility is dependent upon the speed of the
processor and the type of application being run. Performance of graphics applications will be slower than compute-intensive applications. The benchmark results
shown below were derived by running ten cycles (on an average) of each of the following: Norton Speed Index (SI), Dhrystone Integer Benchmark, and PC Magazine’s BENCH V4.0. The BENCH program gives an index number relative to an 8
MHz IBM PC/AT.
As shown in the following table, the DECstation 5000 Model 25 running SoftPC is
approximately twice as fast as an IBM PC/AT.

Table 10-1: SoftPC Benchmarks Results

Computers

SoftPC
Norton SI

SoftPC
SoftPC
Dhrystone PC Magazine’s
per second Bench 4.0

DECstation 5000/25 board @66Hz
DECstation 5000/25 HX @66Hz
DECstation 5000/25 HX @72Hz

13.1
12.5
13.0

2,806
2,750
3,193

3.2
3.1
3.2

DECstation 5000/120 MX

9.8

2,528

2.5

DECstation 5000/125 PXG

13.0

2,517

3.1

DECstation 5000/200 CX

13.1

3,479

3.3

IBM PC/AT (8MHz)

6.9

1,761

1.0

Digital’s DECstation Family Performance Summary

DECstations Graphic Options

DECstations Graphic Options
Abbreviations used for the DECstation computer systems appearing in the following
sections are shown in the following table.

Table 11-1: Key to DECstations Graphic Options
Key to Graphic Options
Abbreviation

Full Graphics Description

Options

board
MX
TX
TX
HX
HX
HX
HX
PXG
PXG+
PXG+
PXGT
PXGT+
PXGT+

On-board 8-Plane Frame Buffer
1-Plane Monochrome Frame Buffer
24-Plane TrueColor Frame 2D Graphics
24-Plane TrueColor Frame 2D Graphics
8-Plane Color Smart Frame 2D Graphics
8-Plane Color Smart Frame 2D Graphics
8-Plane Color Smart Frame 2D Graphics
8-Plane Color Smart Frame 2D Graphics
8- or 24-Plane 3D Graphics w/wo Z-Buffer
8- or 24-Plane Dual-Width 3D Graphics
8- or 24-Plane Dual-Width 3D Graphics
24-Plane 3D Graphics w/wo Z-Buffer
24-Plane 3D Graphics
24-Plane 3D Graphics

(1024x768@72Hz)
(1280x1024@72Hz)
(1280x1024@66Hz)
(1280x1024@72Hz)
(1024x768@72Hz)
(1024x864@60Hz)
(1280x1024@66Hz)
(1280x1024@72Hz)
(1280x1024@66Hz)
(1280x1024@66Hz)
(1280x1024@72Hz)
(1280x1024@66Hz)
(1280x1024@66Hz)
(1280x1024@72Hz)

Digital’s DECstation Family Performance Summary

Tested
Yes
Yes
Yes
No
No
No
No
Yes
Yes
Yes
No
Yes
Yes
No

2D Graphics X11perf Benchmarks

12.1

Background
Developed by Digital and submitted to the X consortium at the Massachusetts Institute of Technology, X11perf tests various aspects of X server performance including
simple 2D graphics, window management functions, and X-specific operations.
Other non-traditional graphics included are CopyPlane, and various stipples and
tiles.

X11perf employs an accurate client-server synchronization technique to measure
graphics operations completion time. Both graphics primitive drawing speeds and
window environment manipulation are tested.
Measurements reported in this section are:

12.2

•

2D vector results from X11perf 10-pixel line tests; shown in units of kilo-vectors
(Kvectors/second)

•

2D fill area results from X11perf Copy 500x500 from pixmap to window tests;
shown in units of megapixels or Mpixels (megapixel = 1,048,576 pixels)

Results and Conclusions
The following table and graphs contain excerpts from the X11perf results of the two
most commonly requested performance metrics for 2D graphics systems, 10-pixel
lines and Copy 500x500 from pixmap to window. Results are shown in units of 2D
Kvectors/second drawing rate and Mpixels/second fill rate.
All DECstation graphic measurements reported in this section with the following
default visual types:
•

StaticGray: MX

•

PseudoColor: HX, TX

•

TrueColor: PXG, PXG+, PXGT+

Note: 2D X11perf results were not available for the HP 9000/705, IBM RS/6000
220, IBM RS/6000 350, and SGI 4D/25G.

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Table 12-1: DECstations 2D Graphics X11perf Benchmark
Results
Workstation

2D Kvectors/second¹

2D Mpixels/second²

DS 5000/20 Board
DS 5000/20 HX
DS 5000 /20 TX³

153.0
274.0
137.0

5.7
14.6
4.1

DS 5000/25 Board
DS 5000/25 MX
DS 5000/25 HX
DS 5000/25 TX³
DS 5000/25 PXG
DS 5000/25 PXG+

183.0
115.0
285.0
156.0
260.0
339.0

6.4
8.0
14.7
4.3
13.9
18.3

DS 5000/120 MX
DS 5000/120 HX
DS 5000/120 TX³
DS 5000/120 PXG
DS 5000/120 PXG+
DS 5000/120 PXGT
DS 5000/120 PXGT+

108.0
272.0
137.0
259.0
338.0
434.0
434.0

7.9
14.6
4.1
13.9
18.3
12.3
12.3

DS 5000/125 MX
DS 5000/125 HX
DS 5000/125 TX³
DS 5000/125 PXG
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+

114.0
284.0
156.0
259.0
338.0
434.0
434.0

8.0
14.6
4.3
13.9
18.3
12.3
12.3

DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 TX³
DS 5000/133 PXG
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+

124.0
298.0
158.0
260.0
339.0
434.0
434.0

8.1
14.8
5.1
13.9
18.3
12.3
12.3

DS 5000/200 MX
DS 5000/200 HX
DS 5000/200 PXG
DS 5000/200 PXG+
DS 5000/200 PXGT
DS 5000/200 PXGT+

162.0
510.0
263.0
345.0
445.0
445.0

¹ Kvectors=1,000 vectors

14.0
30.3
13.9
18.5
12.3
12.3
(continued on next page)

² Mpixels=1,048,576 pixels ³ PseudoColor

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Table 12-1: DECstations 2D Graphics X11perf Benchmark
Results (continued)
Workstation
DS 5000/240 MX
DS 5000/240 HX
DS 5000/240 TX³
DS 5000/240 PXG
DS 5000/240 PXG+
DS 5000/240 PXGT
DS 5000/240 PXGT+
¹ Kvectors=1,000 vectors

2D Kvectors/second¹

2D Mpixels/second²

248.0
621.0
278.0
263.0
345.0
445.0
445.0

20.3
30.5
8.1
13.9
18.5
12.3
12.3

² Mpixels=1,048,576 pixels ³ PseudoColor

Table 12-2: Competitors 2D Graphics X11perf Benchmark
Results
Workstation

2D Kvectors/second¹

2D Mpixels/second²

SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2

58.2
29.3
217.0
205.0

5.1
17.9
9.7
8.3

HP 9000/425t Personal VRX
HP 9000/425t
HP 9000/425e
HP 9000/710
HP 9000/720 CRX

23.3
69.1
49.0
492.0
868.0

.9
1.5
8.8
3.4
22.8

IBM RS/6000 320H

77.4

3.4

SGI 4D/35
SGI 4D/RPC Indigo

11.3
141.0

.9
8.6

All X11perf results from Workstation Laboratories benchmarking
¹ Kvectors=1,000 vectors
² Mpixels=1,048,576 pixels

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Figure 12-1: Personal DECstations and Competitors 2D
Graphics X11perf Benchmark Results
153.0

DS 5000/20 board
DS 5000/20 HX
DS 5000/20 TX
DS 5000/25 board
DS 5000/25 HX
DS 5000/25 TX
DS 5000/25 PXG+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/425e
SGI Indigo

274.0
137.0
183.0
285.0
156.0
339.0
58.2
29.3
217.0
205.0
23.3
69.1
49.0
141.0
0.0

200.0

400.0

600.0

800.0

1000.0

2D Kvectors/second

Figure 12-2: DECstation 5000/120, 125, & 133 and Competitors
2D Graphics X11perf Benchmark Results
DS 5000/120 MX
DS 5000/120 HX
DS 5000/120 TX
DS 5000/120 PXG+
DS 5000/125 MX
DS 5000/125 HX
DS 5000/125 TX
DS 5000/125 PXG+
DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 TX
DS 5000/133 PXG+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/710
HP 9000/720 CRX
IBM RS/6000 320H
SGI 4D/35
SGI Indigo

108.0

272.0
137.0

338.0

114.0
284.0

156.0
338.0
124.0

298.0
158.0
339.0
58.2
29.3
217.0
205.0
23.3
69.1
492.0
868.0
77.4

11.3
141.0

0.0

200.0

400.0

600.0

800.0

1000.0

2D Kvectors/second

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Figure 12-3: DECstation 5000/200 & 240 and Competitors 2D
Graphics X11perf Benchmark Results
DS 5000/200 MX
DS 5000/200 HX
DS 5000/200 PXG+
DS 5000/240 MX
DS 5000/240 HX
DS 5000/240 TX
DS 5000/240 PXG+
DS 5000/240 PXGT+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/710
HP 9000/720
IBM RS/6000 320H
SGI 4D/35
SGI Indigo

162.0
510.0
345.0
248.0
621.0
278.0
345.0
445.0
58.2
29.3
217.0
205.0
23.3
69.1
492.0
868.0

77.4
11.3
141.0
200.0

0.0

400.0

600.0

800.0

1000.0

2D Kvectors/second

Figure 12-4: Personal DECstations and Competitors 2D Fill Area
Graphics X11perf Benchmark Results
5.7

14.6
4.1
6.4
14.7

4.3
18.3
5.1

17.9
9.7
8.3

0.9
1.5
8.8
8.6

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

Mpixels/second

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Figure 12-5: DECstation 5000/120, 125, & 133 and Competitors
2D Fill Area Graphics X11perf Benchmark Results
7.9

DS 5000/120 MX
DS 5000/120 HX
DS 5000/120 TX
DS 5000/120 PXG+
DS 5000/125 MX
DS 5000/125 HX
DS 5000/125 TX
DS 5000/125 PXG+
DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 TX
DS 5000/133 PXG+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/710
HP 9000/720
IBM RS/6000 320H
SGI 4D/35
SGI Indigo

14.6
4.1
18.3
8.0
14.6
4.3
18.3
8.1
14.8
5.1
18.3
5.1
17.9

9.7
8.3
0.9
1.5
3.4
22.8
3.4
0.9
8.6

0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

Mpixels/second

Figure 12-6: DECstation 5000/200 & 240 and Competitors 2D Fill
Area Graphics X11perf Benchmark Results
14.0

DS 5000/200 MX
DS 5000/200 HX
DS 5000/200 PXG+
DS 5000/240 MX
DS 5000/240 HX
DS 5000/240 TX
DS 5000/240 PXG+
DS 5000/240 PXGT+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/710
HP 9000/720
IBM RS/6000 320H
SGI 4D/35
SGI Indigo

30.3
18.5
20.3
30.5
8.1
18.5
12.3
5.1
17.9
9.7
8.3
0.9
1.5
3.4
22.8
3.4
0.9
8.6
0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

Mpixels/second

Digital’s DECstation Family Performance Summary

3D Graphics Benchmarks

13.1

Background
Digital has developed several proprietary benchmarks to test the 3D primitive level
performance of our PEX servers. This consists of building the appropriate 3D structures on the server, and then measuring the time to complete drawing these structures a number of times. We use the same accurate client-server synchronization
technique to insure graphics operation completion time that is used in X11perf. The
transport mechanism reported in this report is local:0.0 (shared memory transport).

These benchmarks measure 3D vector and 3D polygon performance. 3D vectors are
10-pixel, 10-segment polylines, random orientation, and clip checked. 3D vector results are reported in kilo-vectors (Kvectors)/second or 1,000 vectors/second. 3D
polygons are 100-pixel, 10-triangle strip, shaded, default, and directional lighted, Zbuffered, and clip checked. 3D polygon results are shown in kilo-polygon
(Kpolygons)/second or 1,000 polygons/second.

13.2

Results and Conclusions
In general, when evaluating graphics performance, it is important to fully understand
the benchmarking being quoted. You should refrain from comparing one primitive
performance number to another without first understanding how the entities are defined and measured. In addition, primitive-level benchmarks are good for measuring drawing rates for particular entities, but do not take into account other operations
that an application may perform (such as picking or structure editing), or characteristics of the entire system (such as typical background load or disk I/O). Therefore,
the best way to evaluate a system is to run the actual application itself. Because the
DECstation 5000 family exhibits excellent balanced performance, graphics applications may see greater throughput on this workstation type than on others with the
same graphics primitive performance.
Characterizing graphics performance is a complex task. How the application is written, the characteristics of the system the application is running on, and the nature of
the graphics data itself are all factors that affect graphics performance. In addition,
it is possible to measure graphics performance at several different levels. For example, you could measure how long it takes to draw an individual primitive at peak
hardware rates, or you could measure how long it takes to set up and draw an entire
picture using a high-level Application Programming Interface (API).
Because of the complexity of the problem, there are widely varying approaches
within the industry for generating graphics metrics. While X11perf offers some
standardization in the realm of 2D benchmarks, examination of commonly-quoted

Digital’s DECstation Family Performance Summary

3D Graphics Benchmarks

3D numbers show that there is little uniformity regarding what is being measured,
and at what level it is measured. As a result, comparing 3D graphics performance
numbers quoted by one vendor to those of another is rarely a meaningful comparison and can be misleading. It is for this reason that we are not offering comparisons
with competitive systems in this section.
The following table and graphs present the 3D vectors and 3D polygons results for
the DECstation systems.

Table 13-1: 3D Graphics Benchmark Results

Workstations

3D Kvectors/second¹

3D Kpolygons/second²

DS 5000/25 PXG
DS 5000/25 PXG+

288
312

51
68

DS 5000/120 PXG
DS 5000/120 PXG+
DS 5000/120 PXGT
DS 5000/120 PXGT+

288
307
310
310

51
68
102
102

DS 5000/125 PXG
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+

288
310
313
313

51
68
102
102

DS 5000/133 PXG
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+

288
376
405
405

51
68
102
102

DS 5000/200 PXG
DS 5000/200 PXG+
DS 5000/200 PXGT
DS 5000/200 PXGT+

302
400
434
434

52
70
106
106

DS 5000/240 PXG
DS 5000/240 PXG+
DS 5000/240 PXGT
DS 5000/240 PXGT+

302
401
436
436

52
70
106
106

¹ Kvectors=1,000 vectors
² Kpolygons=1,000 polygons

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

14.1

Background
Picture-Level Benchmark (PLB) is software that allows comparisons to be made of
graphics display performance for different hardware platforms. It is the first product from the Graphics Performance Characterization (GPC) committee, a volunteer
group of vendors, users, and consultants that provide and support standardized
benchmarks for measuring graphics performance as related to specific applications.
The National Computer Graphics Association (NCGA) is administrator for the committee.

PLB is designed to measure the performance of CRT-based display systems such as
engineering workstations, personal computers, and special-purpose attached display
systems. Two requirements exist for the PLB to work. The geometry must be presented to the system in a specified format and the PLB code must have been ported
to the device under test.
The five major components of the PLB are as follows:
1. Benchmark Interchange Format (BIF), the file format for specifying the geometry.
2. Benchmark Timing Methodology (BTM) which provides a standardized performance measurement.
3. Benchmark Reporting Format (BRF), for standardized reporting of test results.
4. Picture-Level Benchmark (PLB) program which implements BIF file processing
and runs the test.
5. A suite of standard tests and a report summary sheet.
In order to run BIF files, the PLB code must be customized for each hardware configuration.
To date, five application files have been approved by the GPC committee for use.
They are as follows:
•

"pc_board" - a typical 2-D electrical CAD application

•

"sys_chassis" - a 3-D wire frame model of a computer chassis

•

"cyl_head" -a 3-D solid model of an automobile engine’s cylinder head

•

"head" - depicts a 3-D human head modeled using data generated by a laser
scanner

•

"shuttle" - an example of low-end 3-D simulation

Note: Although the PLB allows buyers to compare performance, it does not address
the issue of display quality. It is the user’s responsibility to look at the image on the
screen and determine superiority.

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

14.2

Results and Conclusions
PLB performance results are reported using a measure called the "GPCmarks". The
GPCmarks is a ratio determined by dividing a normalizing constant by the elapsed
time in seconds required to perform the test. The higher the number, the better the
performance.
Each benchmark generates two GPCmarks; the "PLBlit" (PLB Literal) and the
"PLBopt" (PLB Optimized). The PLBlit results of the GPC are most useful for users who know how their applications draw pictures. They select the benchmarks
which most closely approximate the software they use, or they develop BIF files for
benchmarks. They want to know what the performance of the workstation will be if
the picture is drawn "as is".
PLBopt results are for the users who may make whatever changes necessary to their
applications to get the best possible performance for the workstation. The picture
will not be drawn "as is". Instead, the drawing may be re-ordered or it might use
different primitives or additional information such as surface normals may be provided.
The GPCmarks are reported in the format:
PLBlit: PLBopt
The following table contains the GPCmarks for the DECstation family of systems
and comparable, competitive workstations.

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

Figure 14-1:

PLB Benchmarks Results
PLB Benchmark PLBlit:PLBopt

System

pc_board sys_chassis cyl_head

head

shuttle

DS 5000/20 PXG
DS 5000/20 PXG+

9.8:na¹
11.2:na

10.2:na
10.8:na

14.4:na
16.2:na

18.8:na
20.9:na

17.6:na
20.0:na

DS 5000/25 PXG8
DS 5000/25 PXG
DS 5000/25 PXG+8
DS 5000/25 PXG+

9.7:na
9.8:nr
11.2:na
11.2:nr

10.8:na
10.5:na
12.0:na
11.9:na

nr²
14.4:na
nr
16.2:na

nr
18.8:na
nr
20.9:na

nr
17.6:na
nr
20.0:na

DS 5000/120 PXG
DS 5000/120 PXG+
DS 5000/120 PXGT
DS 5000/120 PXGT+

9.8:na
11.2:na
12.1:na
12.1:na

10.2:na
10.7:na
11.4:na
11.4:na

14.4:na
16.2:na
17.0:na
17.0:na

18.8:na
20.9:na
21.3:na
21.3:na

17.6:na
20.0:na
21.1:na
21.1:na

DS 5000/125 PXG8
DS 5000/125 PXG
DS 5000/125 PXG+8
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+

9.7:na
9.8:na
11.2:na
11.2:na
12.2:na
12.2:na

10.8:na
10.5:na
12.0:na
11.9:na
12.5:na
12.5:na

np³
14.4:na
nr
16.2:na
17.0:na
17.0:na

np
18.8:na
nr
20.9:na
21.3:na
21.3:na

np
17.6:na
nr
20.0:na
21.1:na
21.1:na

DS 5000/133 PXG8
DS 5000/133 PXG
DS 5000/133 PXG+8
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+

9.8:na
9.9:na
11.2:na
11.3:na
12.2:na
12.2:na

11.3:na
11.4:na
13.2:na
13.2:na
13.7:na
13.7:na

nr
14.3:na
nr
16.1:na
17.1:na
17.1:na

nr
18.8:na
nr
20.9:na
21.3:na
21.3:na

nr
17.6:na
nr
19.9:na
21.1:na
21.1:na

DS 5000/200 PXG8
DS 5000/200 PXG
DS 5000/200 PXG+
DS 5000/200 PXGT
DS 5000/200 PXGT+

10.0:na
10.0:na
11.6:na
12.6:na
12.6:na

11.7:na
11.7:na
13.6:na
14.9:na
14.9:na

np
14.9:na
16.8:na
17.7:na
17.7:na

np
19.2:na
21.3:na
21.7:na
21.7:na

np
18.3:na
20.8:na
21.9:na
21.9:na

DS 5000/240 PXG8
11.5:na
13.7:na
nr
nr
nr
DS 5000/240 PXG
10.0:na
11.7:na
14.9:na
19.2:na
18.3:na
DS 5000/240 PXG+
11.6:na
13.8:na
16.7:na
21.3:na
20.7:na
DS 5000/240 PXGT
12.6:na
15.3:na
17.8:na
21.7:na
22.1:na
DS 5000/240 PXGT+
12.7:na
15.3:na
17.6:na
21.7:na
21.9:na
------------------------------------------------------------------------------------------------------------(continued on next page)

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

PLB Benchmark PLBlit:PLBopt
System

pc_board sys_chassis cyl_head head

shuttle

HP 9000/710
24.2:26.8 20.1:34.3
nr
nr
nr
HP 9000/720 CRX
27.6:31.6 21.1:37.1
11.4:na
14.3:15.8 15.6:na
HP 9000/720 CRX24
29.7:34.3 21.5:40.3
18.9:na
14.4:16.0 15.7:na
HP 9000/720 CRX24Z 28.4:33.0 20.9:38.0
31.9:na
32.6:42.5 21.1:na
-------------------------------------------------------------------------------------------------------------IBM RS/6000 220 GTO 18.5:32.4 29.5:na
11.3:17.4 27.0:31.8 np:22.4
IBM RS/6000 320H GTO 18.8:35.9 31.5:na
11.5:17.4 27.2:34.2 np:24.2
IBM RS/6000 350 GTO 18.8:41.0 31.5:na
11.3:17.4 27.4:35.5 np:25.3
-------------------------------------------------------------------------------------------------------------SPARCstation IPX
14.3:14.3
SPARCstation 2 GX
14.6:14.6
SPARCstation 2 GXplus 14.0:14.0
SPARCstation 2 GS
3.6:3.6
SPARCstation 2 GT
9.7:9.7

¹ not available
² not reported
³ not possible
Digital used DEC PHIGS
HP used HP-PHIGS
IBM used graPHIGS
SPARCstations used SunPHIGS

10.7:10.7
10.8:10.8
10.7:10.7
4.0:4.0
18.6:18.6

nr
nr
nr
7.2:7.2
15.1:15.1

nr
nr
nr
nr
nr
nr
6.2:6.2 9.2:9.2
29.3:29.3 17.2:17.2

PXG8 = 8-Plane
PXG = 24-Plane with optional Z-buffer
PXG+8 = 8-Plane with Z-buffer
PXG+ = 24-Plane with Z-buffer
PXGT = 24-Plane with Z-buffer
PXGT+ = 24-Plane with Z-buffer

Digital’s DECstation Family Performance Summary

A
Test Configurations

Digital’s DECstation Family Performance Summary

A-1

Test Configurations

The benchmarks ran on systems with the following configurations. Sources of benchmark results follow configuration listing.

Personal DECstation 5000 Model 20 Board, TX, HX Workstations:
CPU chipset
CPU MHz
FPU chipset
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters

R3000A
20 MHz
R3010A
20 MHz
16 MB
SCSI 426MB RZ25, 665 MB RZ56
64KB data/64KB instruction
Ethernet
ULTRIX X4.2A-4 (Rev. 20)
DEC Fortran T3.1 (ft2), DEC C V1.0
Berkeley FFS
10% bufcache, delay_wbuffers=1, cache_bufcache=1

Version 1.1 and Version 2.1 Dhrystones/second reported. Khornerstone2 tests used 10% buffer cache. All
benchmark testing was performed by Digital Equipment Corporation.

Personal DECstation 5000 Model 25 Board, MX, TX, HX, PXG, PXG+ Workstations:
CPU chipset
CPU MHz
FPU chipset
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters

R3000A
25 MHz
R3010A
25 MHz
16 MB
SCSI 426MB RZ25. 665 MB RZ56
64KB data/64KB instruction
Ethernet
ULTRIX X4.2A-4 (Rev. 20)
DEC Fortran T3.1 (ft2), DEC C V1.0
Berkeley FFS
0% bufcache, delay_wbuffers=1, cache_bufcache=1

Version 1.1 and Version 2.1 Dhrystones/second reported. Khornerstone 2 tests used 10% buffer cache. All
benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 120 MX, TX, HX, PXG, PXG+,PXG Turbo, PXG Turbo+ Workstations:
CPU chipset
CPU MHz
FPU chipset
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters
Background Load
System State

R3000A
20
R3010
20
16
665 MB RZ56
64KB data/64KB instruction
Ethernet
ULTRIX V4.2 (Rev. 85)
DEC FORTRAN, DEC C
Berkeley FFS
Unlimited Stack Size , cache_bufcache=1 , delay_wbuffers=1
none
single-user

Version 1.1 and V2.1 Dhrystones/second reported. All benchmark testing was performed by Digital Equipment Corporation.

A-2

Digital’s DECstation Family Performance Summary

Test Configurations

DECstation 5000 Model 125 MX, TX, HX, PXG, PXG+, PXG Turbo, PXG Turbo+ Workstations:
CPU chipset:
CPU MHz:
FPU chipset:
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters
Background Load
System State

R3000A
25
R3010
25
16
665 MB RZ56
64KB data/64KB instruction
Ethernet
ULTRIX V4.2 (Rev. 85)
DEC FORTRAN EFT5, DEC C
Berkeley FFS
Unlimited Stack Size, cache_bufcache=1, delay_wbuffers=1
none
single-user

Version 1.1 and V2.1 Dhrystones/second reported. All benchmark testing was performed by Digital Equipment Corporation.
DECstation 5000 Model 133 MX, TX, HX, PXG, PXG+, PXG Turbo, PXG Turbo+ Workstations:
CPU Chipset
CPU MHz
FPU chipset
FPU MHz
Memory
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters
Background Load
System State

R3000A
33
R3010A
33
64 MB
2 209 MB RZ24, 1 332 MB RZ55
128KB data/64KB instruction
Ethernet
ULTRIX T4.2A-5
DEC FORTRAN T3.1, DEC C V1.0
Berkeley FFS
10% bufcache, cache_bufcache=1, delay_wbuffers=1
none
single-user

Version 1.1 and Version 2.1 Dhrystones/second reported. Khornerstone2 results used 10% buffer cache. All
benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 200 MX, HX, PXG, PXG+, PXG Turbo, PXG Turbo+ Workstations:
CPU chipset
CPU MHz
FPU chipset
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters
Background Load
System State

R3000
25
R3010
25
16
665 MB RZ56
64KB data/64KB instruction
Ethernet
ULTRIX T4.2 (Rev. 54)
DEC FORTRAN EFT5, DEC C
Berkeley FFS
Unlimited Stack Size , cache_bufcache=1, delay_wbuffers=1
none
single-user

Version 1.1 and Version 2.0 Dhrystones/second reported. Khornerstone2 tests results shown. All benchmark testing was performed by Digital Equipment Corporation.

Digital’s DECstation Family Performance Summary

A-3

Test Configurations

DECstation 5000 Model 240 MX, TX, HX, PXG, PXG+, PXG Turbo, PXG Turbo+ Workstations:
CPU chipset
CPU MHz
FPU chipset
FPU MHz
Memory (MB)
Disk
Cache Size
Network Interface
Operating System
Compilers
File System
Tuning Parameters

R3000A
40 MHz
R3010A
40 MHz
64 MB
SCSI 1.0 GB RZ57
64KB data/64 KB instruction
Ethernet
ULTRIX X4.2A-1 (Rev. 25)
DEC Fortran T3.1 (ft2), DEC C V1.0
10% bufcache, delay_wbuffers=1, cache_bufcache=1

Version 1.1 and Version 2.1 Dhrystones/second reported. Khornerstone2 test used 10% buffer cache. All
benchmark testing was performed by Digital Equipment Corporation.

Hewlett-Packard 9000/425t with Personal VRX Graphics Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory and Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Operating System
Compilers & Switches
Graphics Libraries Used

68040 - 25 MHz
built-in
8 Kb Total ( 4 Kb each I & D)
32 MB
HP 200 Mb 3.5" internal disk/Rodime
2/SCSI
?/200 MB
HP-UX Version 7.03
HP Fortran -O3, HP C -O3
X11

X11perf benchmarking numbers from Workstation Laboratories, 2/1/91, Volume 12, page V12-28-Config.

Hewlett-Packard 9000/425t Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Operating System
Compilers
Graphic Libraries Used

68040 - 25 MHz
built-in
8 Kb Total (4 Kb each I & D)
32 MB
HP 200 Mb 3.5" internal disk/Rodime
2/SCSI
?/200 MB
HP-UX Version 7.03
HP Fortran -O3, HP C -O3
X11

SPEC results from SPEC Newsletter, Volume 3, Issue 1 Winter 1991, page 18. Dhrystone Versions 1.1
and 2.0, Whetstone, Linpack, Khornerstone, and X11perf benchmarking numbers from Workstations Laboratories, 2/1/91, Volume 12, page V12-20-Config. DR Labs CPU2 MVUPS from Digital Review, 3/4/91, page
19, running HP-UX Version 8.0.

A-4

Digital’s DECstation Family Performance Summary

Test Configurations

Hewlett-Packard 9000 Model 425e:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Disk Buffer Sizes
Display Size & Type
Display Resolution
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
For Network Tests: "Remote" or "Local"
For Network Tests: Server Type
For Network Tests: Network Type
For Network Tests: Network Speed
Operating System Name & Level
Fortran Supplier & Version
Fortran Compiler Switches Used
C Compiler & Version
Compiler Switches Used
Graphics Libraries Used

68040 - 25MHz
built-in
8 Kb Total (4 Kb each I & D)
32 MB
9.3 MB except where indicated (small=2.3 MB)
HP 16" Color
1280 x 1024 pixels
HP 200 MB 3.5" internal disk/Rodime
1/SCSI
?/200 MB
Both
MIPS RC3240
Ethernet NFS
10 MB/second
HP-UX Version 8.05
HP Fortran
+O3
HP C
+O3
X11

SPEC, Dhrystone Versions 1.1 and 2.0, Whetstone, Linpack, Khornerstone (with large buffer cache), DR
Labs CPU2, and X11perf benchmark results from Workstation Laboratories, 6/1/91, Volume 14, Chapter 5.
Hewlett-Packard 9000 Model 705 Workstation:
Processor Type & Frequency
Floating Point Unit
Cache Memory Size & Speed
RAM Memory Size & Speed
Display Size & Type
Display Resolution
Disk Subsystem
Operating System
Compilers & Versions
Network Type

PA-RISC 1.1 - 35 MHz
35 MHz Custom
32 KB Instruction / 64 Kb Data
16 MB
19" Grayscale
1280 x 1024, 72 MHz refresh
2-SCSI HP C2235A 420 MB
HP-UX 8.05
HP C A.08.53, HP Fortran A.08.05
Ethernet

SPEC benchmark results from SPEC Newsletter, Volume 4, Issue 1, March 1992, page 90. Dhrystone Version 1.1,and Linpack ratings from HP January 1992 announcement.

Hewlett-Packard 9000 Model 710 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Disk Buffer Sizes
Display Size & Type
Display Resolution
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Network Type
Operating System
Compilers and Switches

HP/PA 1.1-50 MHz
built-in
32 Kb Instruction/64 Kb Data
16 MB - 80 ns DRAM
10% of memory
19" Color
1280 x 1024 pixels
1/SCSI
?/425 MB
Ethernet
HP/UX 8.07
HP Fortran 8.05 +O3 (+OP4 used for some tests), HP C 8.07 +O3

Dhrystone Version 1.1 and Version 2.0, Whetstone, Linpack, Khornerstone, and X11perf ratings from Workstation Laboratories, 3/1/92, Volume 16, page V16-17-Config. SPEC benchmark numbers from SPEC
Newsletter, Volume 4, Issue 1, March 1992, page 90, running HP-UX 8.05, HP C A.08.53, and HP Fortran
A.08.05 and using 2-SCSI HP C2235A 420 MB. DR Labs CPU2 from Digital Review, 2/3/92, page 26.
GPCmarks from The GPC Quarterly Report, Vol. 2, No. 1, 1st Qtr 1992, page 12.

Digital’s DECstation Family Performance Summary

A-5

Test Configurations

Hewlett-Packard 9000 Model 720 Workstation:
SPU
Memory
Disk Size and Type
Operating System
Compilers & Switches

HP720
16 MB
210 MB internal
HP-UX 8.05M (available in June 1991)
HP-UX FORTRAN/9000s 700 optimizing prep-processor (avail
able in June 1991); HP C Compiler HP92453-01 a.08.53
SPEC - optimized with -O; linked with archive libraries by setting
the environment variable LDOPTS to -a archive; floating point
compiled with +OP or +OP4; Dhrystone LDOPTS set to -a and
+O3; Linpack LDOPTS set to -a and compiled with +OP3; Whet
stone LDOPTS to -a and optimization level -O; Khornerstone
10% default buffer cache, LDOPTS set to -a and optimization +O.

SPEC benchmarking numbers from SPEC Newsletter, Volume 3, Issue 2, Spring 1991, page 20.
Dhrystone, Whetstone, Khornerstone 2 , and Linpack benchmark numbers from HP Apollo Series 700 Workstations Performance Overview, March 1991. X11perf benchmarks numbers from Workstation Laboratories,
4/1/91, Volume 13, Chapter 10. Configuration used by WSL was HP 9000 Model 720 CRX. DR Labs CPU2
rating from Digital Review, 1/20/92, page 47. GPCmarks from The GPC Quarterly Report, Vol. 2, No. 1, 1st
Qtr 1992, page 12.
IBM RISC System/6000 POWERstation 220 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Operating System
Compilers & Switches
Graphic Libraries Used

IBM RISC - 33MHz
built into CPU chip set
8 Kb data + instruction
32 MB
IBM 400 MB SCSI 3.5"
AIX 3.2
AIX XL FORTRAN Ver. 2.2, AIX XL C Ver. 1.2
X11 and graPHIGS

SPEC benchmark numbers from SPEC Newsletter, Volume 4, Issue 1, March 1992, page 91. Dhrystone
Versions 1.1 and 2.0, Whetstone, Khornerstone , DR Labs CPU2, and Linpack benchmark numbers from
Workstation Laboratories, 3/1/92, Volume 16, page V16-19-Config. GPCmarks from The GPC Quarterly
Report, Vol. 2, No. 1, 1st Qtr 1992, page 13.
IBM RISC System/6000 POWERstation 320H Computer:
CPU
FPU
Memory
Disk Controller
Disk
Cache Size
Network Interface
Operating System
Compilers
File System

IBM Power RISC - 25MHz
Integrated
16MB
DBA/SCSI
400 MB
32KB data/8 KB instruction
Ethernet
AIX 3.2
AIX XL Fortran Ver. 2.2, AIX XL C Ver. 1.2
AIX

SPEC results from SPEC Newsletter, Volume 4, Issue 1, March 1992, page 92. Dhrystone Versions 1.1 and
2.0, Linpack, Whetstone, Khornerstone, and X11perf ratings from Workstation Laboratories, 3/1/92, Volume
16, page V16-9-Config. GPCmarks from The GPC Quarterly Report, Vol. 2, No. 1, 1st Qtr 1992, page 13.

A-6

Digital’s DECstation Family Performance Summary

Test Configurations

IBM RISC System/6000 POWERstation 350 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Operating System
Compilers & Switches
Graphic Libraries Used

IBM RISC- 42 MHz
built into CPU chip set
32 Kb Data / 8 Kb Instruction
32 MB
IBM 400 MB SCSI 3.5"
AIX 3.2
IBM Fortran 3.2 -O, IBM C 3.2 -O
X11 and graPHIGS

SPEC benchmark numbers from SPEC Newsletter, Volume 4, Issue 1, March 1992, page 94, using AIX XL
C/6000 Ver. 1.2 and AIX XL FORTRAN Ver. 2.2. Dhrystone V1.1 and V2.0, Whetstone, Khornerstone , DR
Labs CPU2, and Linpack benchmark numbers from Workstation Laboratories, 3/1/92, Volume 16, page V1620-Config. GPCmarks from The GPC Quarterly Report, Vol. 2, No. 1, 1st Qtr 1992, page 13.
Silicon Graphics Personal Iris (4D/25G) Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands
Hardfile Size(s) (Unformatted/For.)
Operating System Name & Level
Compilers & Switches
Graphics Libraries Used

R3000 Mips RISC CPU - 20 MHz
R3000 Mips RISC FPU - 20 MHz
96 Kbytes Total - 32 Kb Data/64 Kb Instruction
16 MB
Control Data SCSI
380/170 MB
Silicon Graphics Unix Operating System - 4D1-3.2
Silicon Graphics Fortran (Mips) version 4D1-3.2-O, Silicon
Graphics C (Mips) version 4D1-3.2 -O
X11 and GL Graphics Libraries

SPEC benchmark numbers from SPEC Newsletter,Volume 3, Issue 1 Winter 1991, page 4. Dhrystone V1.1,
Whetstone, Khornerstone, and Linpack benchmark numbers from Workstation Laboratories, 2/1/90, Volume
9, page V9-19-Config. DR Labs CPU2 MVUPS from Digital Review, 3/11/91, page 19, Silicon Graphics Personal Iris 4D/25 running IRIX 3.3.1.
Silicon Graphics 4D/35 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Operating System Name & Level
Compilers & Switches

R3000 - 36 MHz
R3010 - 36 MHz
128 Kb total / 64 Kb Instruction / 64 Kb Data
32 MB - 80 ns DRAM
Seagate (CDC) 5.25" SCSI
1/SCSI
?/320 MB
IRIS 3.31
Silicon Graphics Fortran -O3 (-O2 on some), Silicon Graphics C
-O3 (-O2 on some)

SPEC benchmark numbers from Silicon Graphics. Dhrystone V1.1 and V2.0, Linpack , Whetstone, DR Labs
CPU2, Khornerstone, and X11perf ratings from Workstation Laboratories, 3/1/92, Volume 16, page V16-13Config.

Digital’s DECstation Family Performance Summary

A-7

Test Configurations

Silicon Graphics 4D/RPC Indigo Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Display Size & Type
Display Resolution
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
For Network Tests: "Remote" or "Local"
Operating System Name & Level
Fortran Compiler, Version, & Switches
C Compiler, Version, & Switches
Graphics Libraries Used

R3000 Mips RISC CPU - 33 MHz
R3010 Mips RISC FPU - 33 MHz
64 Kbytes Total - 32 Kbytes each I & D
56 MB
16" Color
1024 x 768 Pixels
Seagate 3.5" ST1480N / SCSI
1/SCSI
480/400 MB
Remote, using Mips RC3240 server, Ethernet, 10 MB/second
Silicon Graphics Unix OS - 4D1-4.0
Silicon Graphics (Mips) version 4D1-4.0 -O3 or -O2
Silicon Graphics (Mips) version 4D1-4.0 -O3 or -O2
X11 and GL Graphics Libraries

Dhrystone V1.1 and V2.0, Whetstone, Linpack, Khornerstone, and X11perf benchmark results from Workstation Laboratories, 9/1/91, Volume 15, Chapter 23, page V15-23-Config. DR Labs CPU from Digital Review,
1/20/92, page 47. SPEC benchmark results from Silicon Graphics Computer Systems, INTRODUCING IRIS
INDIGO Competitive Analysis, July 22, 1991, page 27. Configuration used was 33 MHz MIPS R3000A CPU,
32 MB memory, Ethernet network, IRIX 4.0 version 240, Beta software operating system, IRIX system
daemons, xdm background load, network daemons for remotely-run-clients test case, and system state was
multi-user, single-user login.
Sun SPARCstation 2 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Disk Buffer Sizes
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Network Interface
Operating System
Compilers & Switches
Graphics Libraries Used

SPARC - 40 MHz
SPARC (TI) - 40 MHz
65 Kb
16 MB
(14,901 available)
Quantum 210S & Conners CP3200F
2/SCSI
Ethernet
Sun OS 4.1.1
Optional Sun Fortran Compiler -O3, Optional Sun C Compiler -O3
X11 (xnews)

SPEC benchmark numbers from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 21. Whetstone, DR Labs CPU2 MVUPS, Dhrystone V2.0, and X11perf benchmark numbers from Workstation Laboratories, 2/1/91, Volume 12, page V12-23-Config. Linpack and Dhrystone V1.1 numbers from SPARCstation
2 Performance Brief, Sun Microsystems, Inc., November 1990, page 17. GPCmarks from The GPC Quarterly Report, Vol. 2, No. 1, 1st Qtr 1992, page 13.

A-8

Digital’s DECstation Family Performance Summary

Test Configurations

Sun SPARCstation ELC Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/for.)
Operating System Name & Level
Fortran Supplier & Version
Fortran Compiler Switches Used
C supplier & Version
Compiler Switches Used
Graphics Libraries Used

SPARC (LSI) - 33MHz
SPARC (Fujitsu) - 33 MHz
64 Kilobytes
8 Megabytes / 80 ns (except khorner1 @ 16Mb)
CDC (Imprimis) 94191
1 / SCSI
760 / 680 MB
Sun OS 4.1.1
Sun Fortran 1.4
-O4
Sun C 1.1
-O4
X11

SPEC benchmark numbers from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 24. Memory
configured at 16MB, disk subsystem 207 MB SCSI, other software KAP/SUN pre-processor. System state
single user, no tuning parameters in use, and no background load. Dhrystone V1.1 and V2.0, Whetstone,
Khornerstone, Linpack, and X11perf benchmark numbers from Workstation Laboratories, 9/1/1991, Volume
15, Chapter 21. DR Labs CPU2 rating from Digital Review, 2/3/92, page 26.
Sun SPARCstation (4/40) IPC Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Network Interface
Operating System Name & Level
Compilers & Switches
Graphics Libraries Used

SPARC- 25 MHz
SPARC (TI) - 25 MHz
64 Kb
8MB
Maxtor 3.5"
1/SCSI
?/206
Ethernet
Sun OS 4.1
Sun Fortran 4.1 -O3
X11

SPEC benchmark numbers from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 26.
Dhrystone V1.1 and Linpack numbers from SPARCstation 2 Performance Brief, Sun Microsystems, Inc., November 1990, page 17. Dhrystone V2.0 and Whetstone benchmark numbers from Workstation Laboratories,
11/1/90, Volume 11, page V11-21-Config. DR Labs CPU2 MVUPS number from Digital Review, April 15,
1991 page 24 .

Digital’s DECstation Family Performance Summary

A-9

Test Configurations

Sun SPARCstation IPX Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/for.)
Operating System Name & Level
Fortran Supplier & Version
Fortran Compiler Switches Used
C Supplier & Version
Compiler Switches Used
Graphics Libraries Used

SPARC (LSI) - 40MHz
SPARC (Fujitsu) - 40MHz
64 Kilobytes
16 Megabytes / 80 ns
Maxtor 3.5" SCSI
1 SCSI
? / 207 Mb
Sun OS 4.1.1
Sun Fortran 1.4
-O4
Sun C 1.1
-O4
X11

SPEC benchmark numbers from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 22. Other
software used was KAP/SUN pre-processor. Disk Subsystem was 424 MB SCSI. No tuning parameters in
use, no background load, and system state was single user. Dhrystone V1.1 and V2.0, Whetstone,
Khornerstone, Linpack, and X11perf benchmark numbers from Workstation Laboratories, 9/1/91, Volume 15,
Chapter 22. DR Labs CPU2 from Digital Review, 2/3/92, page 26. GPCmarks from The GPC Quarterly
Report, Vol. 2, No. 1, 1st Qtr 1992, page 13.

A-10

Digital’s DECstation Family Performance Summary

B
References

Digital’s DECstation Family Performance Summary

B-1

References

Digital Review, Digital Review, February 3, 1992.
Digital Review, Digital Review, January 20, 1992.
Digital Review, Digital Review, April 15, 1991.
Digital Review, Digital Review, March 11, 1991.
Digital Review, Digital Review, March 4, 1991.
Hewlett-Packard, HP Apollo 9000 Series 700 Performance Brief, revised
January 1992.
Hewlett-Packard, Hewlett-Packard Series 700 Performance Brief, March 1991.
Silicon Graphics Computer Systems, INTRODUCING IRIS INDIGO
Competitive Analysis, July 22, 1991.
SPEC, SPEC Newsletter, Volume 4, Issue 4, March 1992.
SPEC, SPEC Newsletter, Volume 3, Issue 3, September 1991.
SPEC, SPEC Newsletter, Volume 3, Issue 1, Winter 1991.
SPEC, SPEC Newsletter, Volume 3, Issue 2, Spring 1991.
Sun Microsystems, Inc., SunFLASH, Volume 31, Issue 14, July 1991.
Sun Microsystems, Inc., SPARCstation 2 Performance Brief, November 1990.
Workstation Laboratories, Workstation Laboratories, Volume 8, Chapter 9.
Workstation Laboratories, Workstation Laboratories, Volume 9, Chapter 19.
Workstation Laboratories, Workstation Laboratories, Volume 11, Chapters 10
and 11.
Workstation Laboratories, Workstation Laboratories, Volume 11, Chapters 20
and 21.
Workstation Laboratories, Workstation Laboratories, Volume 12, Chapters 2, 3,
11, 20, 23, and 28.
Workstation Laboratories, Workstation Laboratories, Volume 13, Chapters 5,
10 and 19.
Workstation Laboratories, Workstation Laboratories, Volume 14, Chapter 5.
Workstation Laboratories, Workstation Laboratories, Volume 15, Chapters 21
and 22.
Workstation Laboratories, Workstation Laboratories, Volume 16, Chapters 9,
13, 17, 19, and 20.

B-2

Digital’s DECstation Family Performance Summary