Digital PDFs

EC-N0203-51

November 1991

57 pages

Original

0.3MB

Document:	Digital's DECstation Family Performance Summary
Order Number:	EC-N0203-51
Revision:
Pages:	57
Original Filename:	https://web-docs.gsi.de/~kraemer/COLLECTION/DEC/decstation-Dec-91.pdf

OCR Text

Digital’s DECstation Family
Performance Summary

Version 2.0
November 1991
Order Number: EC-N0203-51

Digital Equipment Corporation
Maynard, Massachusetts
TM

_______________________________________________________
First Printing, November 1991
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 ©1991 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, VT220, and ULTRIX.
The following are third-party trademarks:
AIX, IBM are trademarks of International Business Machines Corporation
COMPAQ is a registered trademark of COMPAQ Computer 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
MS-DOS is 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 1.1.

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

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

Methodology .................................................................................................................... 10

Summary of Relative Performance ............................................................................... 10

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

4.1

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

4.2

Results and Conclusions ................................................................................................ 17

Dhrystone Integer Benchmark ....................................................................................... 22

5.1

Background .................................................................................................................... 22

5.2

Results and Conclusion .................................................................................................. 22

Linpack Benchmark ........................................................................................................ 24

6.1

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

6.2

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

Whetstone 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 ............................................................................................ 30

9.1

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

9.2

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

SoftPC for ULTRIX .......................................................................................................... 32

10.1

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

10.2

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

DECstations Graphic Options ........................................................................................ 33

2D Graphics X11perf Benchmarks ................................................................................ 34

12.1

Background .................................................................................................................... 34

12.2

Results and Conclusions ................................................................................................ 34

Digital’s DECstation Family Performance Summary

iii

3D Graphics Benchmarks .............................................................................................. 40

13.1

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

13.2

Results and Conclusions ................................................................................................ 40

Picture-Level Benchmarks ............................................................................................. 42

14.1

Background .................................................................................................................... 42

14.2

Results and Conclusions ................................................................................................ 43

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 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 .......................................................................... 18
SPEC Integer Benchmark Results ....................................................................... 19
SPEC Floating Point Benchmark Results ............................................................ 20
Dhrystones Benchmark Results (MIPS) ............................................................... 22
Dhrystone Benchmark Results (Dhrystones/second) .......................................... 23
Linpack Single-Precision Benchmark Results ...................................................... 24
Linpack Double-Precision Benchmark Results .................................................... 25
Whetstones Single-Precision Benchmark Results ............................................... 26
Whetstones Double-Precision Benchmark Results .............................................. 27
DR Labs CPU2 Benchmark Results .................................................................... 29
Khornerstones Benchmark Results ...................................................................... 31
Personal DECstations’ and Competitors’ 2D Graphics X11perf Benchmark
Results ................................................................................................................. 37
DECstation 5000/133’s and Competitors’ 2D Graphics X11perf Benchmark
Results ................................................................................................................. 37
DECstation 5000/240’s and Competitors’ 2D Graphics X11perf Benchmark
Results ................................................................................................................. 37
Personal DECstations’ and Competitors 2D Fill Area Graphics X11perf
Benchmark Results .............................................................................................. 39
DECstation 5000/133’s and Competitors’ 2D Fill Area Graphics X11perf
Benchmark Results .............................................................................................. 39
DECstation 5000/240’s and Competitors 2D Fill Area Graphics X11perf
Benchmark Results .............................................................................................. 39
PLB Benchmarks Results .................................................................................... 44

Tables
Table 3-1:
Table 3-2:
Table 3-3:
Table 3-4:
Table 3-5:
Table 4-1:
Table 4-2:

DECstations and Competitive Workstations CPU and FPU Benchmark Results 11
DECstations and Competitive Workstations Graphics Benchmark Results I ....... 12
DECstations and Competitive Workstations Graphics Benchmark Results II ...... 13
DECstations and Competitive Workstations SoftPC Benchmark Results ............ 14
Key to Graphs ...................................................................................................... 16
DECstations’ SPEC Ratios (SPEC Reference Time/Elapsed Time) .................... 21
Competitors’ SPEC ratios (SPEC Reference Time/Elapsed Time) ...................... 21

Digital’s DECstation Family Performance Summary

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

SoftPC Benchmarks Results ................................................................................ 32
Key to DECstations Graphic Options .................................................................. 33
DECstations’ 2D Graphics X11perf Benchmark Results ...................................... 35
Competitors’ 2D Graphics X11perf Benchmark Results ...................................... 36
3D Graphics Benchmark Results ......................................................................... 41

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 is crucial for accurate performance prediction. Digital’s
DECstation Family Performance Summary provides performance information
on the DECstation family of systems and selected competitive computers, running standard industry benchmarks.
Digital’s DECstation Family Performance Summary introduces Digital’s newest UNIX-based RISC platforms:
•

Personal DECstation 5000 Model 20

•

Personal DECstation 5000 Model 25

•

DECstation 5000 Model 133

•

DECstation 5000 Model 240

The Personal DECstation is Digital’s first RISC-based personal workstation,
combining the power of a workstation with the productivity of a PC. The Personal DECstation Series consists of two products; the Personal DECstation
5000 Model 20 (R3000A/20MHz) and the Personal DECstation 5000 Model 25
(R3000A/25MHz). Both share the same modular design for investment protection, have built-in graphics, built-in audio, audio clips, and the ability to develop and run multimedia applications. TURBOchannel expansion slots allow
use of 3D graphics, true-color graphics, high-resolution graphics, full motion
video, specialized networking options (including FDDI networks), serial line
expansion, and numbers I/O options. The Personal DECstation 5000 Series
complements the DECstation workstation family with excellent low-end solutions. Both models are ACE compatible.
The DECstation 5000 Model 133 (33MHz) is Digital’s newest member of the
DECstation 500 Model 100 Series of systems. It has a 33MHz CPU on a 3"x 5"
daughter card that allows DECstation 5000 Model 120 and DECstation 5000
Model 125 customers to upgrade processing power easily. The removable card
will also allow an easy and economical upgrade to the R4000 generation of
processors. The DECstation 5000 Model 133 workstation with the HX (2D)
graphics option combines a low cost-of-entry with excellent performance for
any application that requires fast 2D drawing speeds. The DECstation 5000
Model 133 is ACE compatible.
The DECstation 5000 Model 240 is Digital’s newest high-performance UNIXbased RISC workstation. It provides customers with superb across-thespectrum performance in all areas of workstation computation (graphics, I/O,

Digital’s DECstation Family Performance Summary

Introduction

integer and floating point, and network throughput), and all in a compact, lowcost desktop package. The DECstation 5000 Model 240 is a workstation for demanding applications. New graphics options improve 2D performance by 200400%, and up to 33% for 3D wireframes and solids. The DECstation 5000
Model 240 is ACE compatible.
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 Benchmark

•

Dhrystone Benchmark

•

Whetstone Benchmark

•

Linpack Benchmark

•

DR Labs CPU2 Benchmark

•

Khornerstone Benchmark

•

SoftPC Benchmarks

•

2D Graphics X11perf Benchmark

•

3D Graphics Benchmark

•

3D Polygons Benchmark

•

Picture-Level Benchmarks

We ran the DECstation tests discussed in this document in October and November, 1991.
The benchmark results in this performance summary show
•

At an entry price of $3,995 and performing at 16.3 SPECmark, the Personal
DECstation 5000 Model 20 is the best price/performance workstation or PC
under $4,000.

•

The new HX 2D graphics accelerator offers leading X11 2D fill area for image applications and excellent 2D vector performance on the
DECstation 5000 Model 240.

•

The new PXG+ 3D graphics accelerator provides leading graphics performance for full configured systems priced $15,000-$20,000.

Digital’s DECstation Family Performance Summary

Summary of Relative Performance

Methodology
Wherever possible, we executed the same benchmark code on all systems.
However, due to equipment availability of competitive systems and time constraints, some 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 3-4 contain the benchmarking performance results of
the DECstation family of systems and selected competitive systems.

Digital’s DECstation Family Performance Summary

25.5
23.5
32.4
13.5
20.3
24.4
25.0
11.0
10.3
59.5
32.8

41.2

26.3
10.2

DECstation 5000/133¹

DECstation 5000/200¹

DECstation 5000/240¹

SPARCstation IPC²

SPARCstation ELC²

SPARCstation IPX²

SPARCstation 2²

HP 9000/425t

HP 9000/425e

HP 9000/720²

IBM RS/6000 320²

IBM RS/6000 320H²

SGI Personal Iris 4D/25G 12.2
31.1

DECstation 5000/125¹

SGI 4D/35²

SGI Indigo²

NeXTStation

n/a

11.9

23.6

15.9

39.5

12.2

12.3

21.7

18.0

12.8

27.9

19.5

20.9

16.1

13.8

n/a

9.2

28.4

33.4

11.1

66.8

53.1

78.5

9.3

10.3

27.4

26.5

22.0

14.0

35.8

26.7

29.1

21.7

18.4

21.7

18.4

18.97

8.40

14.11

31.27

33.00

16.45

37.10

29.64

57.00

33.09

25.87

28.50

26.68

23.07

15.70

43.0

27.27

34.42

26.80

21.69

26.70

21.60

Dhrystone
MIPS

33333

14749

24793

54945

57981

28901

65185

52083

100149

58139

45454

50075

46875

40540

27585

75557

47920

60475

47090

38116

46927

37951

Dhrystone
(Dhrystone
per second)

7541

6086

4167

22676

n/a

12903

n/a

19920

56180

10858

4417

19920

27778

23148

10204

42812

25679

33292

25627

20465

25860

20985

Whetstone
Single
(KWIPS)

7353

4710

5459

17921

n/a

9615

n/a

22173

48310

10753

4112

14641

19120

14663

6369

34457

20899

26724

20597

16560

20882

16865

1.42

1.27

1.64

4.30

n/a

2.94

11.70

8.22

22.90

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)

1.29

.76

1.58

3.15

6.00

1.42

n/a

6.91

17.20

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)

8.47

6.30

6.58

26.13

n/a

16.15

n/a

27.86

47.70

7.92

8.47

21.60

20.80

17.29

10.64

47.16

28.61

37.10

26.81

22.00

27.71

22.78

DR
Labs
CPU2
(MVUPS)

n/a = not available
Refer to Appendix A for test configurations

Whetstone
Double
(KWIPS)

¹ SPEC benchmark results recorded using new advanced compiler technology which will be available in the Spring 1992
Table current as of November 1991
² SPEC benchmark results recorded using Kuck and Associates’ pre-compilers

n/a

28.0

19.3

DECstation 5000/120¹

Compaq 486/33

14.0

16.4

DECstation 5000/25¹

n/a

20.0

19.1

DECstation 5000/20¹

Compaq 386/33

13.5

16.3

Workstation

15.7

SPEC
Integer

SPECmark

SPEC
Floating
Point

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

14614

10838

17200

92032

n/a

20727

n/a

54661

44573

42753

14512

27142

29111

23909

19030

50949

27604

39606

33731

29494

31257

26689

Total
Khornerstones

12.3

14.0
30.3
18.5
12.3

DECstation 5000/133 PXGT+¹

DECstation 5000/200 MX

DECstation 5000/200 HX

DECstation 5000/200 PXG+

DECstation 5000/200 PXGT+¹

445.0

345.0

510.0

162.0

434.0

339.0

298.0

124.0

434.0

338.0

284.0

114.0

434.0

338.0

272.0

108.0

339.0

285.0

183

153.0

2D Vectors
X11perf.
10-pixel line
(Kvectors/sec.)³

n/a=not available or not applicable

18.3

DECstation 5000/133 PXG+

Table current as of November 1991
Refer to Appendix A for test configuration

14.8

DECstation 5000/133 HX

12.3

DECstation 5000/125 PXGT+¹

8.1

18.3

DECstation 5000/125 PXG+

DECstation 5000/133 MX

14.6

12.3

DECstation 5000/120 PXGT+¹

DECstation 5000/125 HX

18.3

DECstation 5000/120 PXG+

8.0

14.6

DECstation 5000/120 HX

DECstation 5000/125 MX

7.9

18.3

Personal DECstation 5000/25 PXG+

DECstation 5000/120 MX

14.7

6.4

Personal DECstation 5000/25 board

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.)²

106

n/a

102

n/a

102

n/a

102

n/a

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

¹ not actually tested; expected results
² Mpixels=1,048,576 pixels

434

400

n/a

405

376

n/a

313

310

n/a

310

307

n/a

312

n/a

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

14.9

13.6

n/a

13.7

13.0

n/a

12.5

11.9

n/a

11.4

10.7

n/a

11.9

n/a

17.7

16.8

n/a

17.1

16.2

n/a

17.0

16.2

n/a

17.0

16.2

n/a

16.2

n/a

21.7

21.3

n/a

21.3

20.9

n/a

21.3

20.9

n/a

21.3

20.9

n/a

20.9

n/a

sys_chassis cyl_head head

³ Kvectors=1,000 vectors
* Kpolygons=1,000 polygons

12.6

11.6

n/a

12.2

11.3

n/a

12.2

11.2

n/a

12.1

11.2

n/a

11.2

n/a

pc_board

GPC Picture-Level Benchmarks

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

21.9

20.8

n/a

21.1

20.1

n/a

21.1

20.0

n/a

21.1

20.0

n/a

20.0

n/a

shuttle

8.3

Sun SPARCstation ¹

Table current as of November 1991
Refer to Appendix A for test configuration

SGI 4D/RPC Indigo¹

8.6

22.8

8.8

HP 9000/425e¹

HP 9000 Model 720 CRX¹

1.5

HP 9000/425t¹

9.7

Sun SPARCstation IPX¹

HP 9000/425t Personal VRX¹

17.9

12.3

DECstation 5000/240 PXGT+*

Sun SPARCstation ELC¹

18.5

DECstation 5000/240 PXG+

5.1

30.5

DECstation 5000/240 HX

Sun SPARCstation IPC¹

20.3

141.0

868.0

49.0

69.1

23.3

205.0

217.0

29.3

58.2

445.0

345.0

621.0

248.0

(Kvectors/sec.)³

2D Vectors
X11perf.
10-pixel line

n/a

436

401

n/a

(Kvectors/sec.)

3D Graphics
3D Vectors

not available or not applicable
¹ X11perf results from Workstation Laboratories, Inc.

(Mpixels/sec.)²

DECstation 5000/240 MX

Workstation

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

n/a

12.6

11.6

n/a

pc_board

n/a

15.3

13.8

n/a

² Mpixels=1,048,576 pixels

n/a

17.8

16.8

n/a

cyl_head

n/a

21.7

21.3

n/a

head

n/a

22.1

20.9

n/a

shuttle

³ Kvectors=1,000 vectors
~ Kpolygons=1,000 polygons

sys_chassis

GPC Picture-Level Benchmarks

* not actually tested; expected results

n/a

106

n/a

(Kpolygons/sec.)~

3D Graphics
3D Polygons

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

Table current as of November 1991

Refer to Appendix A for test configuration

n/a = not available

n/a

NeXTStation

13.1

DECstation 5000/200 CX
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

1949

1761

3479

2517

2528

2806
2750
3193

SoftPC
Dhrystone

n/a

1.0

3.3

3.1

2.5

3.2
3.1
3.2

SoftPC
PC Magazine’s
Bench V4.0

Table 3-4: DECstations and Competitive Workstations SoftPC Benchmark Results

Part II
Benchmark Results

Summary of Relative Performance

A variety of systems have been tested and compared in this document. Our
tests are primarily CPU and/or graphics-intensive and the reader is encouraged
to carefully consider the appropriateness of mapping these results into their own
environments.
Presented in this section for each benchmark is a brief description of the benchmark ran and the results and conclusions. The table appearing on this page explains the systems’ 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 3-5: 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 720
IBM AT 80286
IBM RISC System/6000 Model 320
IBM RISC System/6000 Model 320H
Silicon Graphics Personal Iris 4D/25G
Silicon Graphics 4D/35
Silicon Graphics 4D/RPC Indigo
NeXT Computer NeXTStation
Compaq Deskpro 386/33
Compaq Systempro 486/33

HP 9000/425e
HP 9000/720
IBM AT 80286
IBM RS/6000 320
IBM RS/6000 320H
SGI 4D/25G
SGI 4D/35
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

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 on the following pages are the SPEC Benchmark Suite results graphed
and various conclusions from the SPEC benchmark testing.
Note: Compaq 386/33 and Compaq 486/33 SPEC results were not available.

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

For the SPEC benchmarks
•

At an entry price of $3,995 and performing at 16.3 SPECmark, the Personal
DECstation 5000 Model 20 is the best price/performance workstation or PC
under $4,000.

•

The Personal DECstation Models 20 and 25 offer better overall SPECmark
performance than the SPARCstation IPC, the HP 9000 Models 425t and
425e, SGI’s Personal Iris 4D/25, and the NeXTStation. The Model 25 offers similar performance to the SPARCstation ELC.

•

The DECstation 5000 Model 133 offers similar performance to the
SPARCstation 2 and SPARCstation IPX at a lower entry price.

•

The DECstation 5000 Model 240 offers better overall performance than the
SPARCstation 2 and SGI’s Iris Indigo.

Figure 4-1:

SPECmark Benchmark Results

DS 5000/20

16.3

DS 5000/25

19.1
16.4

DS 5000/120
DS 5000/125

19.3

DS 5000/133

25.5
23.5

DS 5000/200

32.4

DS 5000/240
SPARCstation IPC

13.5

SPARCstation ELC

20.3
24.4

SPARCstation IPX

25.0

SPARCstation 2
HP 9000/425t

11.0
10.3

HP 9000/425e
HP 9000/720
IBM RS/6000 320

59.5
32.8

IBM RS/6000 320H

41.2
12.2

SGI 4D/25G
SGI 4D/35
SGI Indigo

31.1
26.3

NeXTStation

10.2
0.0

20.0

40.0

60.0

80.0

100.0

SPECmark

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

SPECint measures a CPU’s integer performance which is important to CASE,
electronic publishing, office, and many other commercial applications. The
DECstation 5000 Model 240 has significantly better SPEC integer performance
than Sun’s SPARCstation 2 and both the IBM RS/6000 Models 320 and 320H.
Additionally, while the DECstation 5000 Model 240 is 54% of the overall
SPEC performance of the HP 9000/720, it is 71% of the HP 9000/720’s
SPECint performance.

Figure 4-2:

SPEC Integer Benchmark Results

DS 5000/20

13.5

DS 5000/25

15.7

DS 5000/120

13.8

DS 5000/125

16.1

DS 5000/133

20.9

DS 5000/200

19.5

DS 5000/240

27.9

SPARCstation IPC

12.8

SPARCstation ELC

18.0

SPARCstation IPX

21.7

SPARCstation 2

21.7

HP 9000/425t

12.3

HP 9000/425e

12.2
39.5

HP 9000/720

15.9

IBM RS/6000 320

20.0

IBM RS/6000 320H
SGI 4D/25G

14.0
28.0

SGI 4D/35

23.6

SGI Indigo

11.9

NeXTStation
0.0

20.0

40.0

60.0

80.0

100.0

SPECint

Digital’s DECstation Family Performance Summary

SPEC Benchmark Suite

Comparing SPECfp results show
•

The Personal DECstation 5000 Models 20 and 25 offer approximately twice
the SPEC floating point performance of the HP 9000 Models 425t and 425e,
and the NeXTStation.

•

The DECstation 5000 Model 240 offers approximately 25% better performance than the SPARCstations 2 and IPX.

Figure 4-3:

SPEC Floating Point Benchmark Results

DS 5000/20

18.4

DS 5000/25
DS 5000/120
DS 5000/125

21.7
18.4
21.7
29.1
26.7
35.8

DS 5000/133
DS 5000/200
DS 5000/240
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX

14.0
22.0
26.5
27.4

SPARCstation 2
HP 9000/425t
HP 9000/425e

10.3
9.3
78.5

HP 9000/720
IBM RS/6000 320
IBM RS/6000 320H
SGI 4D/25G
SGI 4D/35
SGI Indigo
NeXTStation

53.1
66.8
11.1
33.4
28.4
9.2

0.0

20.0

40.0

60.0

80.0

100.0

SPECfp
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.

Digital’s DECstation Family Performance Summary

SPEC Benchmark 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

INT²
INT
FP³
FP
FP
INT
INT
FP
FP
FP

DS
DS
DS
5000/20¹ 5000/25¹ 5000/120¹

DS
DS
5000/125¹ 5000/133¹

DS
5000/200¹

DS
5000/240¹

11.0
14.2
9.7
15.6
17.5
15.5
13.6
47.4
18.4
17.0

13.7
17.4
11.4
18.8
20.0
17.4
16.4
57.0
21.9
19.5

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

12.2
17.4
11.4
18.7
19.9
17.5
16.5
57.4
22.1
19.5

12.1
14.2
9.7
15.4
17.5
15.6
13.6
47.5
18.3
17.1

16.8
22.7
14.8
25.2
25.4
22.4
22.3
82.7
28.4
27.1

¹ SPEC results recorded using new advanced compiler technology which will be available Spring 1992.
² Integer benchmark written in C
³ Floating-point benchmark written in FORTRAN

Table 4-2: Competitors’ 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

SPARC- SPARC- SPARC- SPARC- HP
HP
IBM
IBM
station station station station 9000/ 9000/ 6000/ 6000/
IPC¹
ELC¹
IPX¹
2¹
425e 720¹ 320¹ 320H¹

SGI
Indigo¹

INT²
INT
FP³
FP
FP
INT
INT
FP
FP
FP

11.3
13.1
10.3
8.1
15.6
12.9
14.1
42.9
11.5
11.9

23.5
22.8
19.4
22.6
30.5
25.8
22.3
86.5
19.1
23.6

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

11.8
12.5
9.4
7.5
10.6
15.1
9.8
8.3
12.3
8.2

36.0
43.4
44.5
47.6
64.5
37.7
41.2
323.2
78.5
66.4

13.5
16.9
15.9
19.9
20..6 26.0
22.9
28.8
57.7
72.6
15.8
20.0
18.8
23.6
323.2 404.0
41.8
53.0
60.9
76.6

¹ SPEC results recorded using Kuck and Associates’ pre-compilers
² Integer benchmark written in C
³ 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 developed 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).
Dhrystone V1.1 was used to include as many target competitors as possible.
5.2

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

Figure 5-1:

Dhrystones Benchmark Results (MIPS)

21.60
26.70
21.69
26.80

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 9000/720
IBM RS/6000 320
IBM RS/6000 320H
SGI 4D/25G
SGI 4D/35
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

34.42
27.27
43.00
15.70
23.07
26.68
28.50
25.87
33.09
57.00
29.64
37.10
16.45
33.00
31.27
14.11
8.40
18.97
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

Dhrystones/second are shown in the following graph.

Figure 5-2: Dhrystone Benchmark Results
(Dhrystones/second)

37951
46927
38116
47090

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 9000/720
IBM RS/6000 320
IBM RS/6000 320H
SGI 4D/25G
SGI 4D/35
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

60475
47920
75557
27585
40540
46875
50075
45454
58139
100149
52083
65185
28901
57981
54945
24793
14749
33333
0

20000

40000

60000

80000

100000

120000

Dhrystones/second

Digital’s DECstation Family Performance Summary

Linpack Benchmark

6.1

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.
6.2

Results and Conclusions
Charts containing MFLOPS results follow.
Note: SGI 4D/35 single-precision Linpack results were not available. IBM
RS/6000 320H double-precision Linpack results were also unavailable.

Figure 6-1:
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 9000/720
IBM RS/6000 320
IBM RS/6000 320H
SGI 4D/25G
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

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
22.90
8.22
11.70

2.94
4.30
1.64
1.27
1.42
0.00

5.00

10.00

15.00

20.00

25.00

MFLOPS

Digital’s DECstation Family Performance Summary

Linpack Benchmark

Figure 6-2:

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 9000/720
IBM RS/6000 320
SGI 4D/25G
SGI 4D/35
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

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
17.20
6.91
1.42
6.00
3.15
1.58
0.76
1.29
0.00

5.00

10.00

15.00

20.00

25.00

MFLOPS

Digital’s DECstation Family Performance Summary

Whetstone Benchmark

7.1

Background
The Whetstone benchmark was developed in Great Britain’s National Physical
Laboratory in Whetstone, England in 1970. This synthetic benchmark was designed to represent small engineering/scientific programs.

The Whetstone benchmark has been implemented in single-precision and
double-precision FORTRAN programs, each arranged to defeat most compiler
optimizations. The results are measured in KWIPS (thousands of Whetstone Instructions Per Second).
7.2

Results and Conclusions
Shown below are the results of the Whetstone benchmark tests.
Note: IBM RS/6000 320H and SGI 4D/35 single- and double-precision Whetstone benchmark 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 9000/720
IBM RS/6000 320
SGI 4D/25G
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

Whetstones Single-Precision Benchmark Results

20985
25860
20465
25627
33292
25679
42812
10204
23148
27778
19920
4417
10858
56180
19920
12903
22676
4167
6086
7541
0

10000

20000

30000

40000

50000

60000

70000

KWIPS

Digital’s DECstation Family Performance Summary

Whetstone Benchmark

Figure 7-2:
Results

Whetstones Double-Precision Benchmark

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 9000/720
IBM RS/6000 320
SGI 4D/25G
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

34457
6369
14663
19120
14641
4112
10753
48310
22173
9615
17921
5459
4710
7353
0

10000

20000

30000

40000

50000

60000

70000

KWIPS

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/scientific applications.

Performance is expressed as a multiple of MicroVAX II Units of Performance
(MVUPs).
8.2

Results and Conclusions
The results and conclusions of the DR Labs CPU2 benchmark tests are shown
on the following page.
Note: IBM RS/6000 320H and SGI 4D/35 MVUPs results were not available.

Digital’s DECstation Family Performance Summary

DR Labs CPU2

For DR Lab’s CPU2 benchmarks
•

All DECstations, including the Personal DECstations, recorded faster
MVUPs than all SPARCstations.

•

The DECstation 5000 Model 240 is 99% MVUP performance of HP’s
9000/720.

•

The Personal DECstations are 2.5-3.5 times faster than the Compaq 386/33
and 486/33, the HP 9000 Models 425t and 425e, and the NeXTStation.

Figure 8-1:

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 9000/720
IBM RS/6000 320
SGI 4D/25G
Sun Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

DR Labs CPU2 Benchmark Results

22.78
27.71
22.00
26.81
37.10
28.61
47.16
10.64
17.29
20.80
21.60
8.47
7.92
47.70
27.86
16.15
26.13
6.58
6.30
8.47
0.00

10.00

20.00

30.00

40.00

50.00

60.00

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
(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: IBM RS/6000 320H and SGI 4D/35 Khornerstone results were not available. HP 9000/425e Khornerstone results were recorded using large buffer
cache.

Digital’s DECstation Family Performance Summary

Khornerstone Benchmarks

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 9000/720
IBM RS/6000 320
SGI 4D/25G
SGI Indigo
NeXTStation
Compaq 386/33
Compaq 486/33

Khornerstones Benchmark Results
26689
31257
29494
33731
39606
27604

50949
19030
23909
29111
27142
14512
42753
44573
54661
20727

92032
17200
10838
14614

20000

40000

60000

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

NeXTStation

n/a

1,949

n/a

n/a = not available

Digital’s DECstation Family Performance Summary

DECstations Graphic Options

DECstations Graphic Options
Abbreviations used for the DECstation computer systems included in the graphics benchmarks 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
HX
HX
HX
HX
PXG
PXG+
PXG+
PXGT
PXGT+
PXGT+

On-board 8-Plane Frame Buffer
1-Plane Monochrome Frame Buffer
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)
(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
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 kilovectors (Kvectors/second)

•

2D fill area results from X11perf Copy 500x500 from pixmap to window
tests; shown in units of mega-pixels or Mpixels (mega-pixel=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.
Note: We believe that it is inappropriate to compare Compaq systems X11perf
results with the other systems shown here. There are several X11 server software packages available that will turn the Compaq into a X11 terminal. Each
one would probably produce different X11perf results because each has different performance characteristics (one might do lines really well and another
might do windows really well). You would not be able to directly map the results because other vendors of 486 systems’ hardware would be different and
their performance would also be different. Therefore, Compaq X11perf results
have not been included in the table nor graphic graphs.
Note: 2D X11perf results were not available for the IBM RS/6000 320, IBM
RS/6000 320H, and SGI 4D/25G.

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

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

2D Kvectors/second²

2D Mpixels/second³

DS 5000/20 Board
DS 5000/20 HX

153.0
274.0

5.7
14.6

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

183.0
115.0
285.0
260.0
339.0

6.4
8.0
14.7
13.9
18.3

DS 5000/120 MX
DS 5000/120 HX
DS 5000/120 PXG
DS 5000/120 PXG+
DS 5000/120 PXGT
DS 5000/120 PXGT+¹

108.0
272.0
259.0
338.0
434.0
434.0

7.9
14.6
13.9
18.3
12.3
12.3

DS 5000/125 MX
DS 5000/125 HX
DS 5000/125 PXG
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+¹

114.0
284.0
259.0
338.0
434.0
434.0

8.0
14.6
13.9
18.3
12.3
12.3

DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 PXG
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+¹

124.0
298.0
260.0
339.0
434.0
434.0

8.1
14.8
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+¹

162.0
510.0
263.0
345.0
445.0
445.0

14.0
30.3
13.9
18.5
12.3
12.3

DS 5000/240 MX
DS 5000/240 HX
DS 5000/240 PXG
DS 5000/240 PXG+
DS 5000/240 PXGT
DS 5000/240 PXGT+

248.0
621.0
263.0
345.0
445.0
445.0

20.3
30.5
13.9
18.5
12.3
12.3

¹ Not tested; expected results

² Kvectors=1,000 vectors

Digital’s DECstation Family Performance Summary

³ Mpixels=1,048,576 pixels

2D Graphics X11perf Benchmarks

Table 12-2: Competitors’ 2D Graphics X11perf Benchmark
Results
Workstations

2D Kvectors/second²

2D Mpixels/second³

SPARCstation IPC¹
SPARCstation ELC¹
SPARCstation IPX¹
SPARCstation 2¹

58.2
29.3
217.0
205.0

5.1
17.9
9.7
8.3

HP 9000/425t Personal VRX¹
HP 9000/425t¹
HP 9000/425e¹
HP 9000/720 CRX¹

23.3
69.1
49.0
868.0

.9
1.5
8.8
22.8

SGI 4D/RPC Indigo¹

141.0

8.6

¹ X11perf results from Workstation Laboratories benchmarking
² Kvectors=1,000 vectors
³ Mpixels=1,048,576 pixels

In 2D vector drawing

•

The Personal DECstation Models 20 and 25 with no graphics accelerations
outperform all competitors in their class. They are more than twice as fast
as the SPARCstation IPC and the HP 9000 Models 425t and 425e; they are
5 times faster than the SPARCstation ELC.

•

With HX, the 2D graphics accelerator, the Personal DECstation 5000
Model 25 is faster than the SPARCstation 2 and SPARCstation IPX at a
much lower entry price.

•

Both the Personal DECstation 5000 Model 25 and DECstation 5000
Model 133 with PXG+ 3D graphics acceleration are 14 times faster than the
HP 9000 Model 425t with Personal VRX, also a 3D graphics accelerator.

•

With HX, the 2D graphics accelerator, the DECstation 5000 Model 133 is
faster than the SPARCstation IPX, SPARCstation 2, and SGI’s Iris Indigo.

•

The DECstation 5000 Model 240 offers over 2.8 times the performance of
the SPARCstation IPX and SPARCstation 2, and over 4.4 times the performance of SGI’s Iris Indigo.

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Figure 12-1: Personal DECstations’ and Competitors’ 2D
Graphics X11perf Benchmark Results
153.0
183.0

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

58.2
29.3

285.0
339.0

217.0
205.0
23.3
69.1
49.0
141.0

200.0

0.0

400.0

600.0

800.0

1000.0

2D Kvectors/second

Figure 12-2: DECstation 5000/133’s and Competitors’ 2D
Graphics X11perf Benchmark Results
124.0

DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 PXG+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/720 CRX
SGI Indigo

298.0
339.0
58.2
29.3
217.0
205.0
23.3
69.1

868.0
141.0
0.0

200.0

400.0

600.0

800.0

1000.0

2D Kvectors/second

Figure 12-3: DECstation 5000/240’s and Competitors’ 2D
Graphics X11perf Benchmark Results
DS 5000/240 HX
DS 5000/240 PXG+
DS 5000/240 PXGT+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/720 CRX
SGI Indigo

621.0

345.0
445.0

58.2
29.3
23.3
69.1
0.0

217.0
205.0

141.0
200.0

868.0
400.0

600.0

2D Kvectors/second

Digital’s DECstation Family Performance Summary

800.0

1000.0

2D Graphics X11perf Benchmarks

2D fill area measures the copy of a pixmap to a window which is very important for
image applications and double-buffer animation techniques. All DECstations listed
in the following charts use full 8-bit planes.
The results of 2D fill area tests indicate

•

The Personal DECstation 5000 Models 20 and 25 with no graphics acceleration
are over five times faster than the HP 9000 Model 425t. Note that the
SPARCstation ELC is displaying a single-bit monochrome image, which makes it
faster than most other systems shown in Figure 12-4.

•

With HX, the 2D graphics accelerator, the Personal DECstation 5000 Model 25 is
faster than the SPARCstations 2 and IPX at a much lower entry price.

•

Both the Personal DECstation 5000 Model 25 and DECstation 5000 Model 133
with PXG+ 3D graphics acceleration are 14 times faster than the HP 9000
Model 425t with Personal VRX, also a 3D graphics accelerator.

•

The DECstation 5000 Model 240 MX is over twice as fast as the SPARCstation 2.

•

The DECstation 5000 Model 240 with HX graphics accelerator outperforms the
HP 9000 Model 720. At 30.5 Mpixels/second, the DECstation 5000 Model 240
can achieve real-time image animation.

Digital’s DECstation Family Performance Summary

2D Graphics X11perf Benchmarks

Figure 12-4: Personal DECstations’ and Competitors 2D Fill
Area Graphics X11perf Benchmark Results
5.7
6.4

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

14.7
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

Figure 12-5: DECstation 5000/133’s and Competitors’ 2D Fill
Area Graphics X11perf Benchmark Results
8.1

DS 5000/133 MX
DS 5000/133 HX
DS 5000/133 PXG+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/720
SGI Indigo

14.8
18.3
5.1
17.9

9.7
8.3
0.9
1.5
22.8
8.6
0.0

5.0

10.0

15.0

20.0

25.0

30.0

35.0

Mpixels/second

Figure 12-6: DECstation 5000/240’s and Competitors 2D Fill
Area Graphics X11perf Benchmark Results
20.3

DS 5000/240 MX
DS 5000/240 HX
DS 5000/240 PXG+
DS 5000/240 PXGT+
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
HP 9000/425t PVRX
HP 9000/425t
HP 9000/720
SGI Indigo

30.5
18.5
12.3

5.1
17.9
9.7
8.3
0.9
1.5
22.8
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

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, Z-buffered, 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. One 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).

Digital’s DECstation Family Performance Summary

3D Graphics Benchmarks

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 commonlyquoted 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. (And 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/second²

3D Kpolygons/second³

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+¹

288
307
310
310

51
68
102
102

DS 5000/125 PXG
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+¹

288
310
313
313

51
68
102
102

DS 5000/133 PXG
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+¹

288
376
405
405

51
68
102
102

DS 5000/200 PXG
DS 5000/200 PXG+
DS 5000/200 PXGT
DS 5000/200 PXGT+¹

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 Not tested; expected results
² Kvectors=1,000 vectors
³ 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:
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:

•

"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

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

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.
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 approximates 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 PLBlit (PLB Literal) results for the
DECstation family of systems. The GPC Quarterly Report, coming out in December 1991, will contain the competitors’ and some of these DECstation PLB
results.

Digital’s DECstation Family Performance Summary

Picture-Level Benchmarks

Figure 14-1:

PLB Benchmarks Results

System

PLB Benchmark PLBlit:PLBopt
pc_board sys_chassis cyl_head head

shuttle

DS 5000/20 PXG
DS 5000/20 PXG+

9.8:nr¹
11.2:nr

10.2:nr
10.8:nr

14.4:nr
16.2:nr

18.8:nr
20.9:nr

17.6:nr
20.0:nr

DS 5000/25 PXG
DS 5000/25 PXG+

9.8:nr
11.2:nr

10.5:nr
11.9:nr

14.4:nr
16.2:nr

18.8:nr
20.9:nr

17.6:nr
20.0:nr

DS 5000/120 PXG
DS 5000/120 PXG+
DS 5000/120 PXGT
DS 5000/120 PXGT+³

9.8:nr
11.2:nr
12.1:nr
12.1:nr

10.2:nr
10.7:nr
11.4:nr
11.4:nr

14.4:nr
16.2:nr
17.0:nr
17.0:nr

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

17.6:nr
20.0:nr
21.1:nr
21.1:nr

DS 5000/125 PXG8
DS 5000/125 PXG
DS 5000/125 PXG+
DS 5000/125 PXGT
DS 5000/125 PXGT+³

9.8:nr
9.8:nr
11.2:nr
12.2:nr
12.2:nr

10.5:nr
10.5:nr
11.9:nr
12.5:nr
12.5:nr

np ²
14.4:nr
16.2:nr
17.0:nr
17.0:nr

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

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

DS 5000/133 PXG
DS 5000/133 PXG+
DS 5000/133 PXGT
DS 5000/133 PXGT+³

9.8:nr
11.3:nr
12.2:nr
12.2:nr

11.3:nr
13.0:nr
13.7:nr
13.7:nr

14.4:nr
16.2:nr
17.1:nr
17.1:nr

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

17.8:nr
20.1:nr
21.2:nr
21.2:nr

DS 5000/200 PXG8
DS 5000/200 PXG
DS 5000/200 PXG+
DS 5000/200 PXGT
DS 5000/200 PXGT+³

10.0:nr
10.0:nr
11.6:nr
12.6:nr
12.6:nr

11.7:nr
11.7:nr
13.6:nr
14.9:nr
14.9:nr

np
14.9:nr
16.8:nr
17.7:nr
17.7:nr

np
19.2:nr
21.3:nr
21.7:nr
21.7:nr

np
18.3:nr
20.8:nr
21.9:nr
21.9:nr

DS 5000/240 PXG
DS 5000/240 PXG+
DS 5000/240 PXGT
DS 5000/240 PXGT+³

10.0:nr
11.6:nr
12.6:nr
12.6:nr

11.7:nr
13.8:nr
15.3:nr
15.3:nr

14.9:nr
16.8:nr
17.8:nr
17.8:nr

19.2:nr
21.3:nr
21.7:nr
21.7:nr

18.3:nr
20.9:nr
22.1:nr
22.1:nr

¹ not reported
² not possible
³ not actually tested; expected results

PXG8 = 8-Plane
PXG = 24-Plane with optional 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 were run on systems with the following configurations. Source of benchmark results follow
configuration listing.
Compaq Deskpro 386/33:
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 Size(s) (Unformatted/For.)
Operating System Name & Level
Compilers and Switches

80386 - 33 MHz
80387 - 33 MHz & Weitek 3167 - 33 MHz
64 Kbytes - ? ns
8 MB
Compaq Color VGA
640 - 480
5.25" half-high
380 / 320 MB
ISC Unix Version 2.0
Microway Fortran 1.4e -OLM -n4, Microway C 1.4e -OLM -n4

Dhrystone, Whetstone, Linpack, Khornerstone, and DR Labs CPU2 benchmark results from Workstation
Laboratories, 9/1/89, Volume 8, Chapter 9, page V8-9-Config.

Compaq SystemPro 486-840:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Display Size & Type
Hardfile Brands & Model Numbers
Hardfile Quantity/Interface
Hardfile Size(s) (Unformatted/For.)
Operating System
Compilers and Switches

80486 - 33 MHz (one processor)
Built into 80486
8 Kb on 80486 / 512 Kb external cache
10 MB
14" Color VGA Monitor
Conners 210 MB
4 drives / IDE on each drive
840 MB total / 240 MB per disk drive
SCO Unix 3.2v2.0s
Microway Fortran 2.0.6 -n2 -OLM, Microway C 2.0.6 -n2 -OLM

Dhrystone, Whetstone, Khornerstone, DR Labs CPU2, and Linpack benchmark results from Workstation
Laboratories, 6/1/91, Volume 13, Chapter 5, page V13-5-Config.

Personal DECstation 5000 Model 20 MX, CS, 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
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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992. Khornerstone2 tests used 10% buffer cache.
All benchmark testing was performed by Digital Equipment Corporation.

A-2

Digital’s DECstation Family Performance Summary

Test Configurations

Personal DECstation 5000 Model 25:
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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992. Khornerstone 2 tests used 10% buffer cache.
All benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 120 MX, HX, PXG+,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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992.
All benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 125 MX, HX, PXG+, and 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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992.
All benchmark testing was performed by Digital Equipment Corporation.

Digital’s DECstation Family Performance Summary

A-3

Test Configurations

DECstation 5000 Model 133 MX, HX, PXG+, and 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

Khornerstone2 results used 10% buffer cache.
All benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 200 MX, HX, PXG+, and 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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992. Khornerstone2 tests results shown.
All benchmark testing was performed by Digital Equipment Corporation.

DECstation 5000 Model 240 MX, HX, PXG+, and 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

The SPEC benchmarks were recorded using advanced compiler technology which will be available in the
Spring 1992. Khornerstone2 test used 10% buffer cache.
All benchmark testing was performed by Digital Equipment Corporation.

A-4

Digital’s DECstation Family Performance Summary

Test Configurations

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
Disk Buffer Sizes
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 numbers from SPEC Newsletter, Volume 3, Issue 1 Winter 1991 ,page 18. Dhrystone, 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 HPUX Version 8.0.
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, Whetstone, Linpack, Khornerstone (with large buffer cache) , DR Labs CPU2, and
X11perf benchmark results from Workstation Laboratories, 6/1/91, Volume 14, Chapter 5.

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.
Dhrystones, Whetstones , Khornerstone 2 , and Linpack benchmark numbers from HP Apollo Series 700
Workstations Performance Overview, March 1991. DR Labs CPU2 and X11perf benchmarks numbers from
Workstation Laboratories, 4/1/91, Volume 13, Chapter 10. Configuration used by WSL was HP 9000 Model
720 CRX.
IBM RS/6000 Model 320 Workstation:
Processor Type & Frequency
Floating Point Unix & Frequency
Cache Memory Size & Speed
RAM Memory Size & Speed
Disk Buffer Sizes
Hardfile Brands & Model Numbers
Network Type
Operating System
Compilers & Switches
Graphic Libraries Used

Custom RISC Processor - 20 MHz
Integrated in CPU
8 Kb Instruction + 32 Kb Data
16 MB
Mapped Files
IBM 320 MB SCSI 3.5"
Ethernet NFS
AIX 3.1
IBM XLF Fortran -O, AIX C Version 3.1 -O
X11, graPHIGS, DPS & GSL

SPEC benchmark numbers as announced by IBM on 5/7/91 using Fortran compiler V2 R2 (available in September 1991). Dhrystones, Whetstones, Khornerstones , and Linpack benchmark numbers from Workstation Laboratories, 8/1/90, Volume 11, page V11-10-Config. DR Labs CPU2 MVUPS number from Digital Review, 3/11/91, page 19, running AIX 3.0.
IBM RISC System/6000 POWERstation 320H Computer:
CPU
FPU
Memory
Disk Controller
Disk
Cache Size
Network Interface
Operating System
Compilers
File System

25MHz IBM 2564
Integrated
16MB
DBA/SCSI
320MB
64KB data/8 KB instruction
Ethernet
AIX
C, FORTRAN
AIX

SPEC results from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 14.
Dhrystone and Linpack ratings from IBM information dated March 1991.

A-6

Digital’s DECstation Family Performance Summary

Test Configurations

NeXTStation:
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"
For Network Tests: Server Type
For Network Tests: Network Type
For Network Tests: Network Speed
Operating System Name & Level
Compilers & Switches

68040 - 25 MHz
builtin to 68040
8 Kb Total / 4 Kb in I & D 68040 chip
16 MB
17 " Monochrome
1120 x 832 pixels
? Manufacturer / 3.5"
1/SCSI
480 / 400 MB each
Local and Remote
Mips RC3240 running NFS
Ethernet
10 MB/second
NeXT OS 2.0
Oasys GF68kB Fortran 1.8.5 -OLM, NeXT (Gnu C) C Compiler 2.0
-O

Dhrystone. Whetstone, Linpack, Khornerstones, and DR Labs CPU2 results from Workstation Laboratories,
6/1/91, Volume 13, Chapter 19, page V13-19-Config.
SoftPC Dhrystone results from Workstation Laboratories, 6/191, Volume 13, Chapter 20, page V13-20Config. Operating system was SoftPC MS/DOS 3.30 under NestOS. Compilers were Microsoft Fortran 5.0
/G2 /AL /FPi87 /OX, Microsoft C 6.0 /G2 /AL /FPi87 /Ox.
SPEC benchmark results from NeXT Computer, Inc., 10/6/91. Configuration used 330 MB internal SCSI
disk, NeXTstep 2.1 OS operating system, GNU C 1.36 C compiler, Absoft FORTRAN 77 3.1 Greenhills
Fortran-68000 1.8.5 (Grh) fortran compilers, default tuning parameters, multi-user, normal background load,
and no source code changes to benchmarks.
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. Dhrystones,
Whetstones, Khornerstones, 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:
SPEC benchmark numbers from Silicon Graphics.
Dhrystone and Linpack rating from UNIX WORLD.

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, Whetstone, Linpack, Khornerstone, and X11perf benchmark results from Workstation Laboratories, 9/1/91, Volume 15, Chapter 23, page V15-23-Config.
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, singleuser 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. Whetstones, DR Labs COU2 MVUPS, and X11perf benchmark numbers from Workstation Laboratories, 2/1/91,
Volume 12, page V12-23-Config. Linpack and Dhrystone numbers form SPARCstation 2 Performance Brief,
Sun Microsystems, Inc., November 1990, page 17. Volume 12, page V12-23-Config. Linpack and
Dhrystone numbers from SPARCstation 2 Performance Brief, Sun Microsystems, Inc., November 1990,
page 17.
Khornerstones benchmarks were performed by Digital Equipment Corporation. System tested disk configuration was 1 Quantum 210s and 1 Conners CP3200F, 207MB each formatted capacity.
AIM benchmarks were performed by Digital Equipment Corporation. A CPU 4/75 and 1 203MB SUN0207
disk were used. File system was F.S. Type 4.2. System state was full multi-user and no windows. Compile
flags were -fast, -O4.

A-8

Digital’s DECstation Family Performance Summary

Test Configurations

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.
Dhrystones and Linpack numbers from SPARCstation 2 Performance Brief, Sun Microsystems, Inc., November 1990, page 17. Whetstones 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 .
Khornerstones benchmarks were performed by Digital Equipment Corporation with 24MB of memory, Sun
OS 4.1.1, 1 Quantum 210s, 207 MB formatted, and Sun C compiler and Sun Fortran Compiler.
Aim benchmarks were performed by Digital Equipment Corporation. CPU was a 4/40 and FPU was WEITEK
3172. Memory was 24MB and the disk configuration was 1 203MB SUN0207 and 1 653MB SUN 0669. The
operating system was SUN O.S. 4.1.1 and File system F.S. type 4.2. System state was full multi-user, no
windows. Compile flags were -fast, -O4.

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 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.
Dhrystones, Whetstones, Khornerstone, Linpack, DR Labs CPU2, and X11perf benchmark numbers from
Workstation Laboratories, 9/1/1991, Volume 15, Chapter 21.

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.
Dhrystones, Whetstones, Khornerstone, Linpack, DR Labs CPU2, and X11perf benchmark numbers from
Workstation Laboratories, 9/1/91, Volume 15, Chapter 22.

A-10

Digital’s DECstation Family Performance Summary

B
References

Digital’s DECstation Family Performance Summary

B-1

References

Digital Equipment Corporation, DECstation 5000/100 Series Technical Overview, Version 2.0, Order number: EC-N0580-51, 1991.
Digital Equipment Corporation, Digital RISC Family Performance Summary,
April 2, 1990.
Digital Review, Digital Review, Product Review/Workstations, March 4, 1991.
Digital Review, Digital Review, Product Review/Workstations, March 11,
1991.
Digital Review, Digital Review, Product Review/Workstations, April 15, 1991.
Hewlett-Packard, Hewlett-Packard Series 700 Performance Brief, March 1991.
Norcott, Bill. Review of Industry Standard UNIX Benchmarks, Digital Equipment Corporation, May 1, 1990.
Silicon Graphics Computer Systems, INTRODUCING IRIS INDIGO Competitive Analysis, July 22, 1991.
SPEC, SPEC Newsletter, Volume 3, Issue 1, Winter 1991.
SPEC, SPEC Newsletter, Volume 3, Issue 2, Spring 1991.
SPEC, SPEC Newsletter, Volume 3, Issue 3, September 1991.
Sun Microsystems, Inc., SPARCstation 2 Performance Brief, November 1990.
Sun Microsystems, Inc., SunFLASH, Volume 31, Issue 14, July 1991.
Workstation Laboratories, Workstation Laboratories Benchmark Test Directories, Version of 4/1/90.
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.

B-2

Digital’s DECstation Family Performance Summary

References

Workstation Laboratories, Workstation Laboratories, Volume 23, Chapter 23.

Digital’s DECstation Family Performance Summary

B-3