Digital PDFs

EC-N0626-51

March 1992

42 pages

Original

0.2MB

Document:	Digital's VAXstation 4000 Family Performance Summary
Order Number:	EC-N0626-51
Revision:
Pages:	42
Original Filename:	https://web-docs.gsi.de/~kraemer/COLLECTION/DEC/vaxstation-family.pdf

OCR Text

Digital’s VAXstation 4000 Family
Performance Summary

Version 2.0
March 1992
Order Number: EC-N0626-51

Digital Equipment Corporation
Maynard MA
TM

________________________________________________________________________________________________
First Printing, March 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 ©1991 Digital Equipment Corporation
All right reserved.
Printed in U.S.A.
The following are trademarks of Digital Equipment Corporation: DEC, DECsystem, DECwrite , the DIGITAL Logo, ULTRIX, VAX, VAXstation 3100, VMS
The following are third-party trademarks:
dBASE IV is a registered trademark of Ashton-Tate Corporation.
HP is a trademark of Hewlett-Packard Company.
IBM is a trademark of International Business Machines Corporation.
Lotus and 1-2-3 are registered trademarks of the Lotus Development Corporation.
SPEC is a trademark of the Standard Performance Evaluation Corporation.
SPARCserver is a trademark of Sun Microsystems, Incorporated.
Sun is a trademark of Sun Microsystems, Incorporated.
UNIX is a registered trademark of UNIX System Laboratories, Inc.
WordPerfect is a registered trademark of the WordPerfect Corporation.
X11 is a trademark of The Massachusetts Institute of Technology.
This document was prepared using DECwrite, Version 2.0

Contents
Part I Executive Summary ................................................................................................... 5
1

Introduction ......................................................................................................................... 6

New Product Highlights ....................................................................................................... 6

New Product Descriptions ................................................................................................... 7

Summary of Relative Performance and System Attributes ................................................. 8

Part II Benchmark Results ............................................................................................... 13
3D Graphics Benchmarks .................................................................................................. 15
4.1

Background ................................................................................................................... 15

4.2

Results and Conclusions ................................................................................................ 15

Picture-Level Benchmarks ................................................................................................. 18
4.3

Background .................................................................................................................... 18

4.4

Results and Conclusions ................................................................................................ 19

2D Graphics X11perf Benchmarks ................................................................................. 21
4.5

X11perf Benchmark Results ........................................................................................... 21

SoftPC for VMS ....................................................................................................................... 23
4.6

SoftPC Benchmark Results ............................................................................................ 23

The SPEC Benchmark Suite .............................................................................................. 24
4.7

SPEC Results ................................................................................................................. 24

Dhrystone Integer Benchmark ......................................................................................... 27
4.8

Dhrystone Results .......................................................................................................... 27

Linpack Benchmark .............................................................................................................. 30
4.9

Linpack Results .............................................................................................................. 30

Whetstone Benchmark ........................................................................................................ 32
4.10

Whetstone Results ......................................................................................................... 32

A Test Configurations ......................................................................................................... A-1

iii

B References .......................................................................................................................... B-1
Figures
Figure 4-1:
Figure 4-2:
Figure 4-3:
Figure 4-4:
Figure 4-5:
Figure 4-6:
Figure 4-7:
Figure 4-8:
Figure 4-9:
Figure 4-10:
Figure 4-11:
Figure 4-12:
Figure 4-13:
Figure 4-14:

3D Graphics Kvectors/second Results ................................................................. 16
3D Graphics Kpolygons/second Results .............................................................. 17
PLB Benchmarks Results .................................................................................... 20
X11perf 2D Kvectors/second Benchmark Results ............................................... 22
X11perf Mpixels/second Benchmark Results ....................................................... 22
SPECmark Ratings .............................................................................................. 24
SPECint Ratings .................................................................................................. 25
SPECfp Ratings ................................................................................................... 25
Dhrystones MIPS Ratings .................................................................................... 28
Dhrystones/second Benchmark Results .............................................................. 29
Linpack Single Precision MFLOPS Ratings ........................................................ 30
Linpack Double Precision MFLOPS Ratings ........................................................ 31
Whetstone Single Precision KWIPS Ratings ....................................................... 32
Whetstone Double Precision KWIPS Ratings ...................................................... 33

Tables
Table 3-1:
Table 4-1:
Table 4-2:
Table 4-3:
Table 4-4:
Table 4-5:
Table 4-6:
Table 4-7:

VAXstation 4000 VLC and VAXstation 4000 Model 60 System Features .............. 7
Digital VAXstations and Competitive Systems Graphics & SoftPC Benchmark
Results ................................................................................................................... 9
Digital VAXstations and Competitive Systems CPU and Floating Point Benchmark
Results ................................................................................................................. 10
Digital VAXstations and Competitive Systems PLB Results ................................ 11
Key to Graphs ...................................................................................................... 14
Comparison of SoftPC Software Benchmarking ................................................. 23
SPEC Ratios for VAXstations .............................................................................. 26
SPEC Ratios for RISC-based Systems ................................................................ 26

Part I
Executive Summary

Introduction
This revision of Digital’s VAXstation 4000 Family Performance Summary introduces the SPXg and SPXgt, the new generation of graphics accelerators for the
VAXstation 4000 Model 60. The SPXg and SPXgt provide high-resolution
(1280 x 1024) 3D capability for the VAXstation 4000 Model 60.
The 8-plane SPXg is intended for design automation (CAD/CAE/CAM), molecular
modeling, architectural design, and true-color imaging applications. The SPXg option provides support for full double buffering and 16-bit Z buffering. The higher
performance 24-plane SPXgt is designed for 3D modeling and visualization applications. Again, full double buffering is supported along with 24-bit Z buffering. Both
the SPXg and SPXgt contain hardware acceleration to support major industry standard graphics protocols including PHIGS (Programmers Hierarchical Interactive
Graphical System) and GKS (Graphics Kernel System). PEX support provides the
PHIGS link to the X Window System.
Digital’s VAXstation 4000 Family Performance Summary is a technical reference
document for Digital sales support personnel, customers, and other individuals who
need to understand the performance characteristics of the VAXstation family.
This document provides information about the workstation and the system-level performance of the VAXstation family running standard industry benchmarks and
workloads. The metrics for recording system performance vary according to the
workload and focus on useful work done by the system.
The following sections briefly describe the workloads and graphically depict the results. Test configurations and sources of test results are shown in Appendix A. References are listed in Appendix B.
The VAXstation 4000 Model 60 SPXg and SPXgt tests were performed in February
1992.

New Product Highlights
•

The SPXg and SPXgt provide much better graphics than a SPARCstation 2 GS,
proving that raw CPU power doesn’t necessarily provide the graphics performance needed for 3D.

•

The VAXstation 4000 Model 60 with SPXgt (24-plane) offers better performance and costs less than the Hewlett Packard 9000/425t VRX (CISC, 16-plane).

•

The SPXg and SPXgt offer pricing and graphics performance comparable to the
DECstation PXG graphics adapter.

Digital’s VAXstation 4000 Family Performance Summary

New Product Descriptions
VAXstation 4000 VLC and VAXstation 4000 Model 60 system features are shown
in the following table.

Table 3-1: VAXstation 4000 VLC and VAXstation 4000 Model 60
System Features
VAXstation 4000
VLC

VAXstation 4000
Model 60

Packaging
CPU

Desktop
SOC (System on a chip)

Clock Rate
Read or Write Cycle
Time to Cache
Memory
Disks - Internal fixed
External fixed
Bus

25 MHz

Desktop
VAX CPU and FPU
chipset
55 MHz

Internal Options
Ethernet

None
Thick (adapter to Thin)

System

40 ns
8-24 MB
121 MB maximum
6.1 GB maximum
Sync SCSI

18 ns
8 - 104 MB
852 MB maximum
6.4 GB maximum
Sync SCSI/
Turbochannel
Sync Communications
Thick/Thin

Graphics
1024 x 768
1280 x 1024
8-plane color / 8-plane
greyscale

1280 x 1024
Dual 1280 x 1024
8-plane color / 8-plane
greyscale
8-plane 1280 x1024
3D SPXg graphics
24-plane 1280 x 1024
3D SPXgt graphics

17"
13"
19"
16" / 19"

N/A
N/A
17" / 19"
16" / 19"

Monitors
1024 x 768 monochrome
1024 x 768 color
1280 x 1024 monochrome
1280 x 1024 color

Digital’s VAXstation 4000 Family Performance Summary

Summary of Relative Performance and System Attributes
The following tables show relative application performance positioning only. It is
important to consider other factors when selecting a system to solve a specific problem. Other factors include operating environment, expandability (such as expanded
memory and storage), and price. Using this data, the most appropriate system can
be selected based on system attributes, as well as performance characterization.

Digital’s VAXstation 4000 Family Performance Summary

7.8
14.2
13.4
14.6
24.8
10.4
13.9

VAXstation 3100 Model 76 GPX

VAXstation 3100 Model 76 SPX

VAXstation 4000 VLC

VAXstation 4000 Model 60

VAXstation 4000 Model 60 SPXg

VAXstation 4000 Model 60 SPXgt

DECstation 5000 Model 125 PXG

17.9
9.7
8.3

SPARCstation ELC

SPARCstation IPX

SPARCstation 2 GS

¹n/a = not available or not applicable
Refer to Appendix A for test configurations
Table current as of March 1992

26.3

5.1

SPARCstation IPC

SGI Indigo

22.8

HP 9000/720 CRX

1.5

14.2

VAXstation 3100 Model 38 SPX

HP 9000/425t

7.7

VAXstation 3100 Model 38 GPX

Workstation

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

141.0

205.0

217.0

29.3

58.2

868.0

69.1

259.0

371.0

365.0

216.0

156.0

183.0

23.7

214.0

24.0

250

150

n/a

820

206

288

300

295

n/a

n/a¹

n/a

14.0

6.9

n/a

9.2

n/a

4.3

n/a

3079

1149

n/a

1915

n/a

940

n/a

3.3

1.6

n/a

2.2

n/a

1.1

2D Vectors
3D Graphics 3D Graphics
X11perf
3D Vectors 3D Polygons
SoftPC PC
10-pixel line
(Kvectors/ (Kpolygons/ SoftPC
SoftPC
Magazine’s
(Kvectors/sec.)
sec.)
sec.)
Norton SI Dhrystone Bench V4.0

Table 4-1: Digital VAXstations and Competitive Systems Graphics & SoftPC Benchmark Results

n/a¹
11.0
59.5
13.5
20.3
24.4
25.0
26.3

Compaq SystemPro 486-840

HP 9000/425t

HP 9000/720

SPARCstation IPC

SPARCstation ELC

SPARCstation IPX

SPARCstation 2

SGI Indigo

¹ n/a = not available
Refer to Appendix A for test configurations
Table current as of March 1992

19.3

DECstation 5000 Model 125

6.2

VAXstation 4000 VLC
12.0

6.8

VAXstation 3100 Model 76

VAXstation 4000 Model 60

3.7

23.6

21.7

18.0

12.8

39.5

12.3

n/a

16.1

11.1

5.8

7.1

3.5

SPECmark SPECint

VAXstation 3100 Model 38

Workstation

28.4

27.4

26.5

22.0

14.0

78.5

10.3

n/a

21.7

12.6

6.3

6.6

3.8

SPECfp

31.27

28.50

26.68

23.07

15.70

57.00

25.87

18.97

26.80

17.14

10.87

12.65

6.14

Dhrystone
MIPS

54945

50075

46875

40540

27585

100149

45454

33333

47090

30120

19105

22222

10794

Dhrystone
per second

4.30

6.10

4.34

3.61

3.20

22.90

1.69

1.42

6.67

2.66

1.20

1.92

1.01

Linpack
Single
MFLOPS

3.15

4.20

2.65

2.20

1.70

17.20

1.62

1.29

3.01

1.72

.75

1.12

.49

Linpack
Double
MFLOPS

22676

19920

27778

23148

10204

56180

4417

7541

25627

12774

6357

8216

3919

Whetstone
Single
KWIPS

17921

14641

19120

14663

6369

48310

4112

7353

20597

8628

4070

5784

2514

Whetstone
Double
KWIPS

Table 4-2: Digital VAXstations and Competitive Systems CPU and Floating Point Benchmark Results

34.0 : NR
17.7 : 23.7
14.3 : 14.3
17.8 : 23.8
14.6 : 14.6

HP 9000/720 CRX (HP-PHIGS)

SPARCstation IPX (XGL)

SPARCstation IPX (SunPHIGS)

SPARCstation 2 GX (XGL)

SPARCstation 2 GX (SunPHIGS)

¹ NR= Not Reported
Refer to Appendix A for test configurations
Table current as of March 1992

3.6 : 3.6

9.8 : NR

DECstation 5000 Model 125 PXG

SPARCstation 2 GS (SunPHIGS)

12.3 : NR

VAXstation 4000 Model 60 SPXgt

3.6 : 3.6

11.9 : NR¹

VAXstation 4000 Model 60 SPXg

SPARCstation 2 GS (XGL)

pc_board

Workstation

4.0 : 4.0

10.8 : 10.8

9.5 : 13.6

10.7 : 10.7

9.7 : 13.7

22.1 : NR

10.5 : NR

11.1 : NR

11.0: NR

sys_chassis

7.2 : 7.2

7.0 : 7.0

14.3 : NR

14.4 : NR

8.4 : NR

cyl_head

PLBlit : PLBopt

head

6.2 : 6.2

13.8 : NR

18.8 : NR

8.5 : NR

Table 4-3: Digital VAXstations and Competitive Systems PLB Results

9.2 : 9.2

8.9 : 9.2

16.4 : NR

17.6 : NR

12.5 : NR

shuttle

Part II
Benchmark Results

Results of individual benchmarks can be changed dramatically by the choice of operating system version, compiler version, level of optimization used, memory size,
configuration, cache size, process scheduling, buffer management, and so on. Because some systems have more than one compiler available from the vendor, using
different compilers can have a significant impact on benchmark performance.
The following table explains the abbreviations used in the graphs and tables contained in this section.

Table 4-4: Key to Graphs

Abbreviation

Full Product Description

VS 3100/38
VS 3100/76
VS4000 VLC
VS4000/60
Compaq 486-840
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCserver 2
SGI Indigo

Digital VAXstation 3100 Model 38
Digital VAXstation 3100 Model 76
Digital VAXstation 4000 VLC
Digital VAXstation 4000 Model 60
Compaq SystemPro 486-840
Hewlett-Packard 9000 Model 425t
Hewlett-Packard 9000 Model 720
Sun Microsystems SPARCstation (4/40) IPC
Sun Microsystems SPARCstation ELC
Sun Microsystems SPARCstation IPX
Sun Microsystems SPARCserver 2
Silicon Graphics, Inc. Indigo

Digital’s VAXstation 4000 Family Performance Summary

3D Graphics Benchmarks
4.1

Background
Digital’s 3D graphics benchmarks measure 3D line and 3D polygon performance using a PEX server benchmark. For Digital systems, 3D vectors are 10pixel, 10 lines/polyline, structure mode. 3D vector results are reported in kilovectors (Kvectors)/second or 1,000 lines/second. For Digital systems, 3D
polygons are 100-pixel triangles, 10 triangles/triangle strip, Z-buffered, default+directional lighting. 3D polygon results are shown in kilo-polygon
(Kpolygons)/second or 1,000 polygons/second. Refer to Appendix A for descriptions of competitive systems’ 3D graphics benchmark measurements.

4.2

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.
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 commonlyquoted 3D numbers show that there is little uniformity regarding what is being
measured, and at what level it is measured. Although comparing 3D graphics
performance numbers quoted by one vendor to those of another is rarely a
meaningful comparison and can be misleading, we are showing 3D graphic performances reported by our competitors. The text appearing below each chart
explains what each vendor is measuring.

Digital’s VAXstation 4000 Family Performance Summary

The following graph presents the 3D graphics kvectors/second.
57

VS3100/76 SPX
VS4000/60 SPXg

295

VS4000/60 SPXgt

300

DS5000/125 PXG

288
206

HP 9000/425t VRX

820

HP 9000/720 CRX
SPARCstation 2 GS

150
250

SGI Indigo
0

100

200

300

400

500

600

700

800

900

Kvectors/second

Figure 4-1:

3D Graphics Kvectors/second Results

Digital:
HP:
Sun:
SGI:

10 pixel lines, 10 lines/polyline, structure mode. From PEX server
benchmark.
10 pixel, random orientation, clipping on perspective projection, and
constant color. Through Starbase API.
10 pixel vectors, transformed and clipped. API was not specified.
10 pixel, connected, arbitrarily rotated.

Digital’s VAXstation 4000 Family Performance Summary

The graph below shows the the 3D kpolygons/second results for the VAXstation
4000 Model 60 SPXgt and competitive systems.

VS3100/76 SPX
VS4000/60 SPXg

30
33

VS4000/60 SPXgt

DS5000/125 PXG
HP 9000/425t VRX

29
23

HP 9000/720 CRX
SPARCstation 2 GS

20
60

SGI Indigo

3D Kpolygons/second

Figure 4-2:

3D Graphics Kpolygons/second Results

Digital:
HP:
Sun:
SGI:

100-pixel triangles, 10 triangles/triangle strip, Z-buffered,
default+directional lighting. From PEX server benchmark.
10x10, 50 pixel triangles in triangular strips, random orientation,
transformed, clipped checked, and Z-buffered.
100 pixel triangles, gourand-shaded, clipped, through. Through
SunPHIGS.
Filled 10x10 pixels, 4-sided, independent quadrilaterals, arbitrarily
rotated. Gourand Z, T-mesh.

Digital’s VAXstation 4000 Family Performance Summary

Picture-Level Benchmarks
4.3

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 VAXstation 4000 Family Performance Summary

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.
4.4

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 PLB results for the VAXstation 4000 Model 60 and
competitive systems. The competitors’ results shown in the table were taken from
The GPC Quarterly Report, Volume 1, Number 3, 4th Quarter 1991, except for HP
9000/720 CRX which came from Hewlett-Packard’s 700 Series performance report.

Digital’s VAXstation 4000 Family Performance Summary

Figure 4-3:

PLB Benchmarks Results

PLB Benchmark PLBlit:PLBopt
System

pc_board sys_chassis cyl_head

head shuttle

VS 4000/60 SPXg

11.9 : NR

11.0 : NR

VS 4000/60 SPXgt

12.3 : NR

11.1 : NR

8.4 : NR

8.5 : NR

12:5 : NR

DS 5000/125 PXG

9.8 : NR

10.5 : NR

14.4 : NR

18.8 : NR

17.6 : NR

HP 9000/720 CRX (HP-PHIGS)

34.0 : NR

22.1 : NR

14.3 : NR

13.8 : NR

16.4 : NR

SPARCstation IPX (XGL)

17.7 : 23.7

9.7 : 13.7

SPARCstation IPX (SunPHIGS)

14.3 : 14.3

10.7 : 10.7

SPARCstation 2 GX (XGL)

17.8 : 23.8

9.5 : 13.6

SPARCstation 2 GX (SunPHIGS)

14.6 : 14.6

10.8 : 10.8

SPARCstation 2 GS (XGL)

3.6 : 3.6

4.0 : 4.0

7.0 : 7.0

6.2 : 6.2

8.9 : 9.2

SPARCstation 2 GS (SunPHIGS)

3.6 : 3.6

4.0 : 4.0

7.2 : 7.2

6.2 : 6.2

9.2 : 9.2

¹Not Reported
PXG = 24-Plane with optional Z-buffer

Digital’s VAXstation 4000 Family Performance Summary

2D Graphics X11perf Benchmarks
Developed by Digital and 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.
X11perf is a suite of X-window system display server tests. X11perf determines the number of operations/second that a X server can deliver for 222 client
requests, plus 121 client requests with XOR writing mode, for a total of 343 different client requests. To test the server, there is a single client communicating
exclusively with the server by one transport mechanism.
Measurements reported here are:

4.5

•

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

•

2D fill area results from the X11perf Copy 500X500 from pixmap
to window tests shown in units of mega-pixels or Mpixels (megapixel=1,048,576 pixels).

X11perf Benchmark Results
X11perf results and conclusions for the VAXstation family and its comparable
competitors are presented in the following pages and graphs.
Note: We believe that it is inappropriate to compare Compaq SystemPro
486-840 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 hardware performance would also be different. Therefore, Compaq X11perf results have not been included in the graphic graphs.

Digital’s VAXstation 4000 Family Performance Summary

2D X11perf kvectors/second benchmark results are shown below.
VS3100/38 GPX
24.0
VS3100/38 SPX
214.0
VS3100/76 GPX
23.7
183.0
VS3100/76 SPX
VS4000 VLC
156.0
VS4000/60
216.0
VS4000/60 SPXg
365.0
VS4000/60 SPXgt
371.0
DS5000/125 PXG
259.0
HP 9000/425t
69.1
HP 9000/720 CRX
58.2
SPARCstation IPC
SPARCstation ELC 29.3
SPARCstation IPX
217.0
SPARCstation 2
205.0
SGI Indigo
141.0
0.0

Figure 4-4:

200.0

868.0

400.0

600.0

Kvectors/second

800.0

1000.0

X11perf 2D Kvectors/second Benchmark Results

The following graph shows the 2D X11perf graphics Mpixels/second benchmark results.
7.7

VS3100/38 GPX
VS3100/38 SPX
VS3100/76 GPX
VS3100/76 SPX
VS4000 VLC
VS4000/60
VS4000/60 SPXg
VS4000/60 SPXgt
DS 5000/125 PXG
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

14.2
7.8
14.2
13.4
14.6
24.8
10.4
13.9
1.5
22.8
5.1
17.9
9.7
8.3
26.3
0.0

5.0

10.0

15.0

20.0

25.0

30.0

Mpixels/second
Figure 4-5:

X11perf Mpixels/second Benchmark Results

Digital’s VAXstation 4000 Family Performance Summary

SoftPC for VMS
The SoftPC for VMS software product provides the VAXstation user the ability
to operate DOS based software. The SoftPC product emulates an IBM PC AT
system (real mode only) using standard VAX VMS hardware and software.
The SoftPC for VMS product is a VMS layered product. It supports both
DECwindows as well as the VT220 terminal. As such, it can be executed on
any VAX VMS system and can be displayed on a VT220, or either a local or
remote DECwindows X windows display station.
The performance of the SoftPC facility is dependent upon a number of factors
as well as the speed of the processor upon which it executes.
4.6

SoftPC Benchmark Results
The benchmark results shown below are the averages of ten runs of each of the
following: Norton SI V3.1.0, Dhrystone V1.1, and PC Magazine’s BENCH
4.0.1.
Peter Norton SI (System Information) benchmark measures overall system performance. Dhrystone is an integer benchmark that measures the efficiency of
both the CPU and the C compiler. The CPU test partition of PC Magazine’s
BENCH test was run. The BENCH program gives an index number relative to
a 8MHz IBM AT.
Digital performed the SoftPC benchmarking. SoftPC was run emulating the
286 instruction set (8 MHz).
As shown in the results below, the VAXstation 4000 VLC running SoftPC offers similar performance to an IBM AT. The VAXstation 4000 Model 60 running SoftPC offers twice the performance of the IBM AT.

Table 4-5: Comparison of SoftPC Software Benchmarking

Norton SI
Dhrystone
BENCH Average

VS
3100/38

VS
3100/76

VS
4000 VLC

VS
4000/60

IBM AT
80286

4.3
940
1.1

9.2
1915
2.2

6.9
1149
1.6

14.0
3079
3.3

6.9
1716
1.0

Digital’s VAXstation 4000 Family Performance Summary

The SPEC Benchmark Suite
This section presents the results of System Performance Evaluation Cooperative (SPEC) Benchmark Suite. SPEC is a nonprofit organization formed to develop a standard suite of benchmark programs that characterize system performance.
The release suite consists of ten codes/programs, four of which are written in C
and considered to be compute intensive. The geometric mean of these make up
the SPEC metric called SPECint and are classified as integer benchmarks. The
remaining six programs are FORTRAN based and floating point intensive.
They make up the SPECfp rating. The SPECmark is the geometric mean of the
ten programs’ elapsed times normalized to the VAX 11/780.
4.7

SPEC Results

SPEC ratings for the VAXstation 4000 VLC and VAXstation 4000 Model 60
are shown in the following three charts.
Note: The SPECmark, SPECint, and SPECfp ratings for the Compaq
SystemPro 386-840 were unavailable.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

3.7
6.8
6.2
12.0
19.3
11.0
59.5
13.5
20.3
24.4
25.0
26.3

0.0

10.0

20.0

30.0

40.0

50.0

60.0

70.0

SPECmark

Figure 4-6:

SPECmark Ratings

Digital’s VAXstation 4000 Family Performance Summary

The SPECint ratings for the comparable systems are shown below
VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

3.5
7.1
5.8
11.1
16.1
12.3
39.5
12.8
18.0
21.7
21.7
23.6

0.0

10.0

20.0

30.0

40.0

50.0

SPECint

Figure 4-7:

SPECint Ratings

The following graph presents the SPECfp measurements.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

3.8
6.6
6.3
12.6
21.7
10.3
78.5
14.0
22.0
26.5
27.4
28.4

0.0

Figure 4-8:

10.0

20.0

30.0

40.0 50.0 60.0
SPECfp

70.0

80.0

90.0

SPECfp Ratings

Digital’s VAXstation 4000 Family Performance Summary

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 4-6: SPEC Ratios for VAXstations

Benchmark
No. & Name

Type

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

VS
3100/
38

VS
3100/
76

VS
4000/
VLC¹

VS
4000/
60

3.3
3.3
3.8
3.3
4.3
3.5
3.5
4.6
3.2
3.8

7.5
6.7
8.2
7.0
5.6
7.6
6.5
6.6
7.3
5.4

6.3
5.3
5.7
5.2
7.4
6.2
6.7
11.7
3.9
6.1

11.2
9.9
10.6
10.5
12.7
10.5
12.9
23.6
11.0
11.1

Table 4-7: SPEC Ratios for RISC-based Systems

Benchmark
No. & Name

Type

DS
5000/
125

HP
HP
SPARC SPARC SPARC SPARC
9000/ 9000/ station station station station SGI
425t 720
IPC
ELC
IPX
2
Indigo

13.7
17.4
11.4
18.8
20.0
17.4
16.4
57.0
21.9
19.5

12.4 36.0
12.8 43.4
11.8 44.5
7.7 47.6
11.5 64.5
14.9 37.7
9.7 41.2
10.8 323.2
12.1 78.5
8.4 66.4

11.3
13.1
10.3
8.1
15.6
12.9
14.1
42.9
11.5
11.9

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

23.5
22.8
19.4
22.6
30.5
25.8
22.3
86.5
19.1
23.6

Digital’s VAXstation 4000 Family Performance Summary

Dhrystone Integer Benchmark
The Dhrystone benchmark was introduced in 1984 as an ADA program by
Reinhold P. Weicker. It has since been translated into C and TURBO PASCAL.
This synthetic benchmark measures processor and compiler efficiency with emphasis on the type of data and operations encountered in a system rather than
numerical programming. It is CPU-intensive.
Dhrystones are most commonly expressed in Integer MIPS (Millions of Instructions Per Second) where 1 MIP is the number Dhrystones per second that can be
performed by a VAX 11/780 (1757 Dhrystones/second).
4.8

Dhrystone Results
The following graph shows the MIPS ratings for the VAXstation family and
comparable, competitive workstations.
Note: Both "Inlined" and "Standard" Dhrystone results are provided. "Inlined"
results can be one of the following:
•

The compiler takes modules that are called and placed them "inline", removing the overhead of the call to the module and the overhead with the
register set-up in the Dhrystone code.

•

The loader uses "inline" string functions that are called in Dhrystone code.

We have presented the competitors’ Dhrystone results without labels because
that is how the vendor or benchmarking service has reported them. Some
manufacturers may include inline string libraries in the standard "c" library or
some string manipulation instructions in their instruction sets. We were unable
to verify or determine if the competitive results shown here were "Standard" or
"Inlined".

Digital’s VAXstation 4000 Family Performance Summary

The chart below shows Dhrystone MIPS ratings for the VAXstation family and
comparable, competitive systems.

VS3100/38 (Standard)
VS3100/38 (Inlined)
VS3100/76 (Standard)
VS3100/76 (Inlined)
VS4000 VLC (Standard)
VS4000 VLC (Inlined)
VS4000/60 (Standard)
VS4000/60 (Inlined)
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo
0.00

6.14
6.93

12.65
14.49

10.87
12.02
17.14
18.98
26.80
18.97

25.87
57.00
15.70
23.07
26.68
28.50
31.27

10.00

20.00

30.00

40.00

50.00

60.00

70.00

MIPS

Figure 4-9:

Dhrystones MIPS Ratings

Digital’s VAXstation 4000 Family Performance Summary

Dhrystones per second are charted below.

VS3100/38 (Standard)
VS3100/38 (Inlined)
VS3100/76 (Standard)
VS3100/76 (Inlined)
VS4000 VLC (Standard)
VS4000 VLC (Inlined)

10794
12168
22222
25484
19105
21114

VS4000/60 (Standard)
VS4000/60 (Inlined)
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC

30120
33355

47090
33333
45454
100149
27585

SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

40540
46875

50075
54945

20000

40000

60000

80000

100000 120000

Dhrystones/second
Figure 4-10:

Dhrystones/second Benchmark Results

Digital’s VAXstation 4000 Family Performance Summary

Linpack Benchmark
Developed at Argonne National Laboratories, Linpack is a FORTRAN benchmark that solves a 100x100 system of linear equations. The benchmark is used
to compare the performance of mathematical and scientific applications where
floating point computations are prevalent. When running, the benchmark gives
little weight to I/O.
The results are measured in millions of floating point operations per second
(MFLOPS). Both single and double precision operations are reported here.
4.9

Linpack Results
Single precision MFLOPS ratings are shown in the following graph.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

1.01
1.92
1.20
2.66

0.00

Figure 4-11:

6.67
1.42
1.69
22.90
3.20
3.61
4.34
6.10
4.30

5.00

10.00
15.00
MFLOPS

20.00

25.00

30.00

Linpack Single Precision MFLOPS Ratings

Digital’s VAXstation 4000 Family Performance Summary

The following chart contains the double precision MFLOPS results.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

0.49
1.12
0.75
1.72
3.01
1.29
1.62

0.00

17.20
1.70
2.20
2.65
4.20
3.15

10.00

5.00

15.00

20.00

MFLOPS

Figure 4-12:

Linpack Double Precision MFLOPS Ratings

Digital’s VAXstation 4000 Family Performance Summary

Whetstone Benchmark
The Whetstone benchmark was developed at Great Britain’s National Physical
Laboratory in Whetstone, England in 1970. It is a synthetic benchmark designed to represent small engineering and 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).
4.10

Whetstone Results
Charted below are the Whetstone single and double precision benchmark tests
results.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

3919
8216
6357
12774
25627
7451
4417
56180
10204
23148
27778
19920
22676

Figure 4-13:

10000 20000 30000 40000 50000 60000 70000
KWIPS

Whetstone Single Precision KWIPS Ratings

Digital’s VAXstation 4000 Family Performance Summary

Whetstone double precision KWIPS ratings appear in the following chart.

VS3100/38
VS3100/76
VS4000 VLC
VS4000/60
DS5000/125
Compaq 486-480
HP 9000/425t
HP 9000/720
SPARCstation IPC
SPARCstation ELC
SPARCstation IPX
SPARCstation 2
SGI Indigo

2514
5784
4070
8628
20597
7353
4112
48310
6369
14663
19120
14641
17921

10000

20000

30000

40000

50000

60000

KWIPS

Figure 4-14:

Whetstone Double Precision KWIPS Ratings

Digital’s VAXstation 4000 Family Performance Summary

A
Test Configurations

Digital’s VAXstation 4000 Family Performance Summary

A-1

Test Configurations

The benchmarks were run on the following system configurations:
VAXstation 3100 Model 38 Workstation:
Memory
Disk Controller
Disks
Display Device
Network Interface
Operating System
Compilers

16 MB
SCSI
2 104MB RZ23
VRT19 (1280x1024)
Ethernet
VAX/VMS V5.4
VAX FORTRAN, VAX C

Linpack, and Whetstones tests used 1 665MB RZ56 disk. X11perf benchmarks used VR299 (1024x864) display device. All tests were performed by Digital.
VAXstation 3100 Model 76 Workstation:
CPU
FPU
Number of CPUs
Cache Per CPU
Memory
Disk Controller
Disk
Network Interface
Operating System
Compilers
Tuning Parameters
Background Load
System State

REX520
DC523
1
128KB
16 MB
SCSI
2 332 MB RZ55
Ethernet
VAX/VMS V5.4
VAX FORTRAN V5.2, VAX C V3.1
WSEXTENT=16384, WSQUO=8192, WSDEFAULT=1024
None
single user

SPEC benchmark results from Spec Newsletter, Volume 3, Issue 1, Winter 1991, page 34. Linpack and
Whetstone benchmarks used 1 665MB RZ56 disk. X11perf benchmarks used VR299 (1024x864) display
device and 2 104MB RZ23 MB disks. All tests were performed by Digital.
VAXstation 4000 VLC Workstation:
VAX CPU/FPU
Clock Speed
Memory
Read or write cycle to cache
Display Size and Type
Disk Controller
Disk
Network Interface
Operating System
Compilers
Tuning Parameters
Background Load
System State

SOC (System on a chip)
25 MHz
24 MB / 80 nanosecond RAM
40 nanosecond
Digital VRT13
Sync SCSI
1 121 MB RZ23, 1 332 MB RZ55
Ethernet
VAX/VMS T5.5-4x6
VAX FORTRAN 5.6, VAX C 3.2
WSMAX=16400, WSQUO=16400, WSEXTENT=16400
None
Standalone

System was booted minimum for all benchmarks except for X11perf. DECwindows was running for X11perf
tests. All tests were performed by Digital.

A-2

Digital’s VAXstation 4000 Family Performance Summary

Test Configurations

VAXstation 4000 Model 60 Workstation:
VAX CPU/FPU
Clock Speed
Memory
Read or write cycle time to cache
Display Size and Type
Disk Controller
Disk
Network Interface
Operating System
Compilers
Tuning Parameters
Background Load
System State

VAX CPU/FPU
222MHz
104 MB / 80 nanosecond RAM
18 nanosecond
Digital VRT19
Sync SCSI
2 209MB RZ24
Ethernet
VAX/VMS T5.5-4x6
VAX FORTRAN V5.6, VAX C V3.2
WSMAX=53200, WSQUO=25000, WSEXTENT=25000
None
Standalone

System was booted minimum for all benchmarks except for X11perf. DECwindows was running for X11perf
tests. All tests were performed by Digital.
DECstation 5000 Model 125 PXG Workstation:
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/65KB 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

All benchmark testing was performed by Digital Equipment Corporation.
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
Fortran Supplier & Version
Fortran Compiler Switches Used
C Supplier & Version
C Compiler Switches Used

80486 - 33 MHz (one processor)
Built into 80486
8 Kb on 80486 / 512 Kbytes 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, and Linpack benchmark results from Workstation Laboratories, Volume 13, Chapter
5, page V13-5-Config., June 1, 1991.

Digital’s VAXstation 4000 Family Performance Summary

A-3

Test Configurations

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 numbers from SPEC Newsletter, Volume 3, Issue 1 Winter 1991, page 18. Dhrystone, Whetstone,
and Linpack benchmark results from Workstation Laboratories, 2/1/21, Volume 12, page V12-20-Config.
Hewlett Packard 9000/425t with Personal VRX Graphics 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

X11perf benchmarking numbers from Workstation Laboratories, February 1, 1991, Volume 12, Chapter 28,
page V12-28-Config. 3D graphics tests run on HP 9000/425t P VRX/
Hewlett Packard 9000/720 CRX
Processor Type & Frequency
Number of CPUs
Floating Point
Memory
Network Interface
Operating System/Rel.
Windows System/Rel.
Graphics Library

PA-RISC/50 MHz
1
Yes
16 MB
Ethernet
HP/UX 8.07
X11R4
HP-PHIGS 2.2

SPEC benchmarking numbers from SPEC Newsletter, Volume 3, Issue 2, Spring 1991, page 20.
Dhrystones, Whetstones, and Linpack numbers from HP Apollo Series 700 Workstations Performance Overview, March 1991. 2D graphics results (X11perf) from Workstation Laboratories, 4/1/91, Volume 13, Chapter
10. Model used by Workstation Laboratories was HP 9000 Model 720 CRX. 3D graphic results from HP
Apollo 9000 Series 700 System Performance, 2nd Edition, November 1991, pages 24-25. Configuration
used 420 MB disk and shared memory transport. HP nominal vectors: 10 pixel, random orientation, clipping on perspective projection, and constant color. Through Starbase API. 3D Triangles/second benchmark
conditions were 10x10, 50 pixel triangles in trianglular strips, random orientation, transformed, clipped
checked, and Z-buffered. PLB results from HP Apollo 9000 Series 700 System Performance, 2nd Edition,
November 1991, page 2.

A-4

Digital’s VAXstation 4000 Family Performance Summary

Test Configurations

Silicon Graphics (SGI) 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/Formatted)
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" ST140N/SCSI
1/SCSI
480/400 MB
Remote, using Mips RC3240 server, Ethernet, 10MB/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 or 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 as 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-runclients test case, and system state was multi-user, single-user login. 3D graphics tests run on SGI Indigo
XS24, CPU R3000, Geom. Engines 1, Screen Ref. 60/72 Hz, Z buffer optional, Color planes 24-bit, planes
4/4, Screen resolution 1280 x 1024. 3D lines were 10-pixel, connected, arbitrarily rotated. 3D polygons were
50-pixel triangles, arbitrarily rotated. Per Silicon Graphics, Inc., 1/31/92.
Sun SPARCserver 2 Computer:
CPU Processor
CPU MHz
FPU
FPU MHz
Memory
Disk Buffer Sizes
Disk Controller
Disk
Cache Size
Network Interface
Operating System
Compilers
Compiler Switches

SPARC
40 MHz
SPARC (TI)
40 MHz
16 MB
(14,901 available)
2/SCSI
Quantum 210S & Conners CP3200F
64 KB data/64 KB instruction
Ethernet
Sun OS 4.1.1
Optional Sun Fortran, Optional Sun C
-O3

SPEC benchmark ratings from SPEC Newsletter, Volume 3, Issue 1, Winter 1991, page 10. Linpack and
Dhrystone benchmark ratings from SPARCstation 2 Performance Brief, Sun Microsystems, Inc., November
1990. Tested configuration was SPARCstation 2 running SunOS 4.1.1beta, Sun FORTRAN 1.4beta and
Sun C1.1beta. Whetstone and X11perf results from Workstation Laboratories, Inc., Volume 12, Chapter 23,
page V12-23-Config. 3D graphics results from HP Apollo 9000 Series 700 System Performance, 2nd Edition, November 1991, pages 24-25. 3D vector benchmark conditions were 10 pixel vectors, transformed and
clipped. API was not specified. 3D triangles/second benchmark conditions were 100 pixel triangles,
gourand-shaded, clipped, through SunPHIGS. Tests run on SPARCstation 2GS. PLB results from The GPC
Quarterly Report, Volume 1, Number 1, 4th Quarter 1991, pages 37-38, 41-42. Floating point SPARC IEEE
754, 424 MB disk, 32 MB of memory, operating system SunOS4.1.1revB GFXrev2, windows system
OpenWindows Version 3, and graphics library XGL2.0. Graphics subsystem was 24 or 12+12 DB, 6 window
planes, 16 Z-buffer, resolution 1152x900, 19" CRT, and 76 Hz CRT refresh rate.

Digital’s VAXstation 4000 Family Performance Summary

A-5

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 and Level
Fortran Supplier, Version & Switches
C Supplier, Version & Switches
Graphics Libraries Used

SPARC (LSI) - 33MHz
SPARC (Fujitsu) - 33MHz
64Kb
8 Mb/80 ns
CDC (Imprimis) 94191
1/SCSI
760/680MB
Sun OS 4.1.1
Sun Fortran 1.4 -O4
Sun C1.1 -O4
X11

SPEC benchmark numbers from SPEC Newsletter, Volume 3, Issue 3, September 1991, page 24. Memory
configured was 16MB, disk subsystem 207 MB SCSI, other software KAP/SUN preprocessor. System state:
single user, no tuning parameters, and no background load. Dhrystones, Whetstones, and X11perf benchmarks results from Workstation Laboratories, 9/1/91, Volume 15, Chapter 21.
SPARCstation 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
8 MB
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 7. Whetstones and X11perf benchmark numbers from Workstation Laboratories, 11/1/90,
Volume 11, page V11-21-Config.
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 & Switches
C Supplier, Version & Switches
Graphics Libraries Used

SPARC (LSI) - 40MHz
SPARC (Fujitsu) - 40 MHz
64KB
16MB/80 ns
Maxtor 3.5" SCSI
1 SCSI
? / 207MB
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. Memory
configured was 16MB, disk subsystem 424 MB SCSI, other software KAP/SUN preprocessor. System state:
single user, no tuning parameters, and no background load. Dhrystones, Whetstones, and X11perf benchmarks results from Workstation Laboratories, 9/1/91, Volume 15, Chapter 22. PLB results from The GPC
Quarterly Report, Volume 1, Number 3, 4th Quarter 1991, pages 35-36. Floating point SPARC IEEE 754, 16
MB of memory, operating system SunOS4.1.1revB GFX rev2, windows system Open Windows Version 3,
and graphics library XGL2.0. Graphics subsystem was 8 total bit planes, 8 image no overlay/underlay, no
window planes, 1152 x 900 resolution, 16" CRT size, and 76 Hz CRT refresh rate.

A-6

Digital’s VAXstation 4000 Family Performance Summary

B
References

Digital’s VAXstation 4000 Family Performance Summary

B-1

References

Digital Equipment Corporation, Digital’s DECstation Family Performance
Summary, Version 2.0, November 1991.
Hewlett-Packard, HP Apollo Series 700 Workstations Performance Overview,
March 1991.
Hewlett-Packard, HP Apollo Series 700 System Performance, 2nd Edition,
November 1991.
Standard Performance Evaluation Corporation, SPEC Newsletter, Volume 3,
Issue 3, September 1991.
Standard Performance Evaluation Corporation, SPEC Newsletter, Volume 3,
Issue 2, Spring 1991.
Standard Performance Evaluation Corporation, SPEC Newsletter, Volume 3,
Issue 1, Winter 1991.
Sun Microsystems, Inc., SPARCstation 2 Performance Brief, November 1990.
Workstation Laboratories, Volume 13, Chapter 5.
Workstation Laboratories, Volume 12, Chapters 20, 23 & 28.
Workstation Laboratories, Volume 15, Chapter 21 & 22.

B-2

Digital’s VAXstation 4000 Family Performance Summary