United States. Congress. Senate. Committee on the.

The Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) online

. (page 36 of 140)
Online LibraryUnited States. Congress. Senate. Committee on theThe Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) → online text (page 36 of 140)
Font size
QR-code for this ebook


Data- handling operations - time spent moving data from one
storage location to another and in rearranging and modifying
the data.

Housekeeping operations - counting of iterations, modifying and
constructing addresses, and determining what is to be done next.

Storage accesses - time spent in transferring the instructions
from storage to the control unit and the time spent transferring
operands and partial results to and from storage.



The relative capacity and efficiency of each of the computer systems
considered to be competitive to the G-20 are examined in the follow-
ing points.



5243



APPENDIX D (7)

4. THE INTERNAL OPERATIONS OF THE G-20 ARE RELATIVELY

POWERFUL AND EFFICIENT FOR A COMPUTER IN ITS PRICE
CLASS

Each of the four key internal operating characteristics of the G-20
and its competitors has been measured and reduced to an index value
relative to the G-20. These data are presented in graphical form on
the following pages, along with a discussion pertaining to the method of
measurement employed in each case. Exhibit D-I, following this page,
is typical of the other exhibits to follow. A few moments of study will
aid in the interpretation of the remainder of this evaluation.



(1) Capacity and Efficiency Are a Function of the Machine's
Ability To Perform Simple Operations



In engineering and scientific applications, the most important
operations in terms of time consumed are, in general, the arithmetic
operations, housekeeping operations, and storage accesses. In addi-
tion, some problems require extensive input-output operations, but
these are not considered typical.

Based upon the work of John Von Newmaim and H. H. Goldstein, ^^'
it has been assumed that the speed of internal computation is
directly proportional to storage access time, addition time, and



W H. H. Goldstein: "Systematics of Automatic Electric Computers. '
Proceedings of the Darmstadt Collojuiu.Ti, October 1955.



5244



' . 1

. i


i


i




i

i
1


I
1


C-IO.U.


— - -J


1 ;

!


i

1

i


1


1

i

j
i


i


i
1


i
1 „


' I


i —
1
i

i




i 1


iffir.m


1 1 i

1 1

j i
111-


j
1

1
i i


i '








1 1




"iTT'^


!


1




i


1 i
i i


■ ;




V


j .
1






<0




a**


\\


!
i


i


H












ifl


8


j


1
1

1


;


"M"






1








\_ ,>






1


1


1 1

I i

1 i i






j










!

1


1




1

I

1


i


ill

1 i i.






i
1 -












1


. i




1 : 1






1








'<o


1
\





I


L .. . 1 ■ -


i


1 ' i






1
1


1 —
*

1


L ■


it






!






^ 1


! j 1

ill






i
1




' "


^


1










1


1 i ■

1 ;






1






3


S"

^^^














%i


1 !

i i


i
i


,


1








" \


'm&


m.


fS.i\


KJS


rr':[email protected]


Hi.


CE

210

AS COMPARED TO THE G-20 THE G
HAS 81% L ESS CAPACITY; IS 77S L
EFFICIENT; AT A COST ONLY 20«
THAN THE BENDIX SYSTEM


si

1 i " !












1


■B





1






I'M

i 1 j

i 1

!


!


1























5245



APPENDIX D (8)



(9)

multiplication time. ' Each of these operations is shown in

Exhibits D-II, -III and -IV, following this page.



(2) Housekeeping Operations Were Evaluated on the Basis of a
Computer's Ability To Perform an Iterative Routine



The internal arithmetic and memory access operations con-
sidered above are common to all engineering and scientific problems,
and many data processing-type problems. However, the above
method does not show the relative ability of the various systems in
terms of their power and efficiency in handling internal "housekeeping
operations. This capability can best be evaluated by considering a
problem that requires extensive use of this nonarithmetic task.



(2)consider that the times required for arithmetic, "housekeeping" and storage access operations are
additive. Then, as pointed out in reference 1, the time required for solving a problem may be
written as:

T = KMN

where,

N = number of multiplications

M = multiplication time (exclusive of access time)

K = constant for each machine and problem

This formula is based upon the assumption that, on the average, each multiplication is imbedded
in a sequence of (A + 1) instructions consisting of one multiplication and A nonmultiplicative
instructions, each of which takes an average time L to execute. The time to execute these
instructions is then:

t = M + AL + (Atl)(aj + a^)

where,

a. = Instruction access time

ap = operand access time
then,

T = Nt = KMN

M
K , 1 ^. AL ^ (A+l)(ai t ap)

M M

If we assume that the average time to execute an instruction (L) is directly proportional to addition

time, then the speed of internal computation is directly proportional to storage access times, addition

times, and multiplication times.



40-927 O - pt. 7-27



5246



2 K

■■i s



! I











ieiii^Hiiti



Si



5247



i
1

































































i


M


isi^m's-m^mn


ii


1


Si


-


L








: i






















Pi'ifr'>!iK«^


;^^d sg






v.;-:j-' -




t!o_
























1




1


iSS




















"° J


1


Sg^


|^S3«;5


^?*«!4'


is




















1


^fe


i5


(?¥


SCO
























,^ :


«';•




























































































i"










imii




^.>


1


^^




ii^^vT^^Si






1








































1

1




1 i
i








^


it'*


aa


^- -S


1

1


1 =s!




3


i 1 1






*


r '


<
>_








^ '


1

SPER UNIT OF

S PER DOLLAR
RENCE: TADL

R B DECIMAL
EATER.
SYSTEMS
PRECISION
KS '.










•Ml*




fj^'


m£mi<


- , , ■*


1

ADDITION

ADDITION
REFE

SHOWN ARE FO
ECISION OR Gfi
FOR THE TWO
lONAL 6 DIGIT
WN AS ASTERIC


.






■Ml.


aJ-,


.^


*


1


^ ;■& A^l




Hi^Mii -




i


\


1


1


1


z

1


1


1








mk


1 1
1 1 .





5248















2g|


_'


'm


ss


Wif,










!


1










^


Sgg


i


m


^g


S^


^gjg


j^^


gas-s^-


11




*J


sm


a




iiSsS^


















9


a__


m


g


f-'te:!


^s^Jssd


o

_loo

Q.




















1


I






is


















1


il


.^ia&


tasaagi


^yv-^;;


E^^gj








t&^


x>ii


ft^S?






1
















i

i


^ Ma


il


1


gs


^


^jgj


f^3ftB«




1




















i
1








1


ii


t


SiS


K^


^ii


































i






!
























oi


1





-














ii


Si


Hi


^


^3S


fe»i<a


^gftSg


asa^agj^iss:






























































i°-


ii^


g


^


|a


^


ffi^ja.^-


j^j^^^^






















s










ffl"


ii


i


ii


^


|l;








° 2






f








ATIONS PER UN
ATIONS PER DO

E FOR 8 DECIM
R GREATER.
IT PRECISION
E SHOWN AS






T.?.


s s


i


i


n


^


^^


^^g




1

multiplic

multiplic

:e: table

s shown ar
precision c

S FOR 6 DIG
NIVAC Ml AR








f.M


m


ijav


WTl


E"';**


jg[^






B «i^ 1 Mi^






i^






<
<


i




t-












m






MM




^^■-


s*



5249



APPENDIX D (9)

In the engineering and scientific application areas, problems
that fit these requirements are:

Matrix inversions

Solution of sets of linear equations

Integration of partial differential equations

The number of instructions contained in the iterative procedures
or "inner loops" involved in each of these is highly variable. How-
ever, all of them require the "housekeeping" operations of counting
iterations, constructing and modifying addresses, and determining
what to do next (testing and control).

The following problem was selected as representative of
this particular machine requirement.



IfiOO

Swan. = ^ (A^ X B^ + Ci + D^)



It has been programmed and timed in both floating point and fixed
point form. A minimum level of precision of eight decimal digits
was assumed to be necessary for most engineering and scientific
problenis. However, the six-digit floating point form possible
with the G-20 was also evaluated and is shown in the supporting
tables at the end of this appendix.



5250



APPENDIX D (10)

Exhibits D-Vand VI, following this page, show the compara-
tive results of this analysis for the fixed and floating point form.
Four machines are not shown on the floating point exhibit; three
must be programmed and are, thus, not competitive in this instance,
and the fourth, or Univac III, could not be evaluated for lack of the
necessary data.

5. THE INPUT-OUTPUT CAPABILITIES OF THE G-20 - THOUGH

GOOD - ARE NOT AS OUTSTANDING AS ITS INTERNAL SPEEDS

The G-20 is primarily directed toward the engineering-scientific
field, and, as such, requires superior internal operating capacity and
efficiency. However, data processing and E&S applications are becoming
increasingly intermixed; dual usage is becoming more the rule than the
exception. In an effort to appraise the G-20's relative competitive posi-
tion as a data processor, two capabilities have been exam.ined: internal
data handling and merge-sort operations.



(1) Internal Data Handling Is Somewhat Slower than the Better
Competitive Machines



In data processing operations, both input-output and internal
data-handling capabilities are important. In general, arithmetic
operations are not as important as they are in engineering- scientific
work. As an indication of a given computer's internal data-handling



5251



I i I I !



I I



i I i J

i . : ;




d'j-^gjis.gss'-^kg.ak'&^^sm'a^saag



e2itS?Si9is



^



lintiriifii



^ssmi^iM



U-



.lv^.^,...^...>^^....w^^^



i I i



jSjsa&i^



■■ft^mv.mZ'Vi.ifirttfUi






111-



5252



o
o o

Sw

z <; Id

-° 5 3

<s

z5
u

H

z



rr




iHiiii



;-t?K'^^jid'?:r5v-Stt?SSfe®>:



mm






I I i



^^iavv: ?




WS?-¥S S:



^ S*



! I



©^•Ss



iiiiB^eaiiii^^H



mmamm



5253



APPENDIX D (11)

abilities, an evaluation of the number of characters available per
memory access was performed and is presented as Exhibit D-VII,
following this page. This characteristic and the ability to transfer
blocks of data (greater than one word) are indicative of a machine's
data processing capabilities. These operations are very important
in sorting, file maintenance, and report preparation routines where
large quantities of data are consistently being moved and modified.



(2) The Pending Change in Tape Specifications Reduces the

G-20's Sorting Capabilities from "Outstanding" to "Good"



Merge-type sorting on magnetic tape is one of the best indica-
tions available of a machine's input-output capabilities. In this type
of sorting, extensive use is made of the system's magnetic tape
facilities by continually moving data from tape through the central
processor and back to tape. Because of the buffering associated
with present-day computers, it is possible to perform these three
operations simultaneously. Normally, the reading and writing of
naagnetic tape is the most time-consuming element of this process.
In an effort to appraise the G-20 as a data processor, the following
merge-sort problems were employed:



5254



gx s











i

i
i






£f










1
1

I

i












1

<
z

„ 1*




i
1 !
1 i




1












!




i ^S 1










1 1










I Sop






1 ! 1
1 1 i

1 i ! 1

! 1 1
1 1 1






^


i


! ! i ' '


a^


moE!










1 , i






SJ>iS!£W* S


1 ;

j


5






'Mi
i 1 1 1




I "^








-


j


j


1


^s


«Jl'*;^W


1 , ■ , i


i




! 1




*


3» n^E


"~~'








1 I 1 ! i 1




i i


''*'*"'"-"""•




Mi!: 1






1 i
' i i




! 1 !










ill i i
i i ! i ^








1 i








ill! ^




! ;
j 1




!


1
















i












= 2

5










i '

j 1
1 i










I

1
i












ix


i




1


:

!






5S


i : 1 ' :


^^'2s








i


l


'






2 =


gg^g






















i
i


z

■D

<

<
I


-J










i


i


i


iii


m


,.,,„




X -1








ii*''






■ 1














>-

<
*




Z q:

;










m


■<


s*






















uj -*
S UJ



5255



APPENDIX D (12)



1. 10,000 records, 200 characters per record,
20 character key

2. 40, 000 records, 400 characters per record,
20 character key



Assumptions:



6 magnetic tape units

60, 000 character internal memory capacity



The results of this evaluation are presented as Exhibits D-VIII
and -IX. The indexes shown are based on G-20's originally
published tape specifications. The improved relative perform-
ance of the competitive equipment, based on the pending change
in G-20 tape performance, is also shown.

It can readily be seen that the tremendous increase in start
time for the writing operation greatly reduces the effective speed
of the G-20 magnetic tape units. Under the original 5 millisecond
start-stop time, the G-20 was better in merge-sort operations
than all competitive systemis except the RCA 601 on a capacity
and efficiency basis. The pending change from 5 millisecond to
25 millisecond start-write time will place the G-2 at a competi-
tive disadvantage with respect to the Univac III, Honeywell 800,
Philco 2000 (Models 210 and 211), and the RCA 601. This is
particularly important, in that it gives the Univac III a 2. 5 to 1
advantage over the G-20, with the former being a lower priced
system.



5256



2 a




ho



2 Z



y w



•^ o

SS

CO

m • -

« H
O K

O o

a w
<^

H K
< W

3s



k?*Jf».Baf-«iSJk





I "i:Ka"?Ps??3y'.^T?



rJ ,



i i ij^ii i iV iii u i iAji i i




' mmik



5257




5258



APPENDIX D (13)

6. THE G-20 IS AMONG THE BEST COMPUTERS ON THE MARKET

TODAY

A number of individual machine characteristics and capabilities
have been evaluated on the preceding pages. Of these, two are reason-
ably indicative of a specific computer's capabilities in engineering and
scientific applications and in data processing applications. For scientific,
the floating point summation problem shown in Exhibit D-VI and for data
processing, the merge-sort problem shown in Exhibit D-VIII.

The relative indexes for each of these problems are shown on
Exhibit D-X, following this page. This chart differs slightly from the
preceding nine, in that system capacity is plotted against the system
price index. The efficiency index (capacity per dollar of rental), is not
shown as a bar for each computer, but as a range of equivalent efficiency
to the G-20. In effect, this range shows what the capacity should be
relative to system price for a unit equally efficient as the G-20. A
range of plus or minus 20% has been used, since we do not believe the
average consumer would go into as detailed an analysis as this one and
thus would not place much emphasis on relatively small degrees of
difference in efficiency.

The second major difference between this and the preceding
exhibits is that each system is represented by two bars, one showing
its capacity in the engineering-scientific area (as reflected by its



5259



APPENDIX D (14)

capabilities in the solution of the floating point summation problem) and
the other in the data processing area (as reflected by its capabilities in
the 10,000 record mcrge-sort problem).

The significance of this exhibit lies in the relative positions of the
various equipments. First, consider the four quadrants:



The Upper Left : System.s that fall within this quadrant have
capabilities equal to or better than the G-20 at a price equal
to or less than the Bendix system.

The Upper Right: These systems have capacity equal to or
greater than the G-20 and an equivalent or higher price.

The Lower Right : Here, we find systems with capacity equal
to or less than the G-20, but at a higher price.

The Lower Left : This quadrant contains systems with lower
capacity and at a lower price than the Bendix system.



Next, consider the range of equivalent efficiency. Systems that
fall below the lower limit of this range provide more than 20% less
capacity per dollar of rental than the Bendix system. Those above the
upper limit perform 20% or more operations per dollar than the G-20.
The systems within this range are approximately equal (+ 20%) to the
G-20 in efficiency, though capacity may be higher or lower. Examina-
tion of this exhibit brings forth several significant conclusions relative
to the competitive position of the Bendix G-20 in terms of its price and
capabilities.



5260



23 Q

Z DC

9o

H <->
UW
Z K
D o
fc. o

is"

CO ,

to

W H
CJ K
OO

K M

a w

H K
<W
Q S



""




















^














VX^i*




















JSS&'^SSsg 1






Bl




1




















i.

:


aa


^^m^^^B












.stfas-


ii


11








B


'SSS«


f.fe


m^k


■■bJ


BBBHi






!


-.1








!


1








, _|S^?J'-!i^«« 1












IBB^H










■ISs




-J»'


<^'
























ii^^^i^rtt


T^^sa
















Sj
























'


Li-i' ■''






1




■■.::*.■..'. us ]


"


5

< G




iii


S^iJf'


£: SB






















..


i


>^-»

r*


iPfS


t^S


































































i


i


i


i^,;


Si


i








































^














1












B9!^^^g 1












1




^




S"




















5- '

3


i'■X■i^„v4^


.„.%






















' ■ f?


i


i


i


g t^


5S










































«S*


1
1
i








^■^■naan












1










......I






iHHMMmiMBS

























5261



APPENDIX D (15)



(1) Of the Sixteen Systems with Which It Competes, the G-20
Faces Strong Hardware Competition from Only Five in
Engineering- scientific Applications



Only five competing systems offer equivalent value to the
G-20. These are the Honeywell 800, CDC 1604, RCA 601,
Philco 2000 (Model 211), and the IBM 7090. Each of these
systems offers 50% or more capacity at a relatively higher price.
The RCA 601 stands out as the best value in the group, with the
Philco, Honeywell, and Bendix systems running second.

Within a price range of 50% either way, the Bendix system
is the best value and offers the highest capacity of any computer
on the market. The possible exception to this may be the Univac III,
which is not shown for lack of data.



(2) In Data Processing Applications, the G-20 Faces Strong
Competition from Seven Systems



The most serious threat to the G-20 in this area appears
to be the Univac III, for it offers almost two and a half times the
capacity of the G-20 at a price 5% lower. The GE 210 would be
the second contender, in that it offers the same capacity at a 20%
lower price, while the RCA 501, though of about the same capacity,
is approximately 5% more expensive.



40-927 O - pt. 7-28



5262



APPENDIX D (16)

Among the units with higher data processing capacity than
the G-20, the Honeywell 800, Philco 2000 (Models 210 and 211),
and the RCA 601 stand out as equal or better values than the
Bendix system.



(3) In Dual Applications, Serious Competition May Be Expected
from About Three Systems, All Higher Priced than the G-20



There appear to be only three other systems that offer
equivalent value to the G-20 in both types of applications, those
requiring fast internal computation and those requiring extensive
data handling and input-output. These systems are the Honeywell
800. RCA 601, and Philco 2000 (211), and probably the Univac 111.



(4) These Conclusions - Though Indicative of Relative Hardware
Capabilities - Should Be Kept in Their Proper Perspective



This evaluation has only taken into consideration each system's
ability to perform a series of arbitrarily selected operations and
problems. Those selected may not be in any respect similar to
the requirements of a specific prospect or user. Every effort has
been made to select "typical" computer requirements, but in truth,
relatively few "typical" situations exist.

This evaluation also does not reflect simplicity of program-
ming, availability of programs and routines, or the manufacturer's



5263



APPENDIX D (17)

service. Each of these has a bearing on final equipment selection.
Partial proof of this fact is evident from the acceptance of IBM
products in the market place, in the face of their poor showing
throughout this "hardware" comparison.

The tables following this page present the detailed support-
ing data from which this analysis and Exhibits D-I through D-X
were developed.



This appendix has presented a comparative technical evaluation of
the Bendix G-20 against the 16 computers with which it competes. In a
sense, this evaluation was conducted in a vacuum, in that the capabilities
of each of the units were considered purely on the basis of performance
and price - the criteria that the sophisticated user would select as a
measure of the most economical unit for his needs.

Appendix E, which follows, presents a brief summary of the
responses obtained from a selected group of computer users in various
industries and areas of the United States.



5264



S 3 S 3



3 73 3



5265



fi



H



if



5266



CO CO r-t Tf



CO iH CT>



3 s



lOCMCOOiCOOrHOO



W OS Oi



tH lO lO O rH CO



CI CD (N



Oi N CO O^ 00



i-f o> to



i-l(NCT>CT>01CO OOlO
■^^COCOCOOOOl COiO






^


s






XI


X










CL.


oj



CO oa 05



(M CO to i-l CO



W 05 CD OJ lO



W (N OO 00
00 CO CO 00



CQ ca f-t iO 00



oi CD CO as CO

<N tH CO



O > CJ

u g t>



rH>-ICDCDC<10O00CO0Or-l



■-I ■* ^



e 3

E 2



CO CO OS CO CO cs



CM CO oo ca



CO i-l <N



^ IH CO



CO CO CO



U X, ^ ^ rl






CO lO <N



O O Tj« m
OS O O CO



N CJ d N CJ



i f




>-


CD


s


o


en
o

r-


O


O

o


z

o

X


u

Q
U


2

03




2

CQ


<


X
a.



2 2



5267



,—^. CO CD



lO O CM CO



a> C] CD iH



CD Tj« 05



cn CO 00



CD »H Cl



E 2



CD CO 00 CO Tt*
O C- lO CO c^



C- UO CD OO CJi CO CO
liO OJ CJ O^ C4 c^l o



lo r- CO



CD 00 CO CO CO CD
f-H -^ CQ CD lO C-^



E 2
I 2



OOOCJOOOO^

CDCOC>irHC001COCOo6
CNCO-^i-(C4-«;rC - -^CD
lO tH CJ r-t lO f-H r^



1-1 rH Tj«



CO LO D) O



CM 04 03 W N



^ S*



o


o


U

<






o




o


>•


o

to


s


o

o


Ol


o


o
O




W
O


>


ri


n




p~


u




c-




r-




(.J


03


5

D


z

u

CO


z

m


<


2

CO


s

CQ


O

X


U
O

u


s

CO


X
a.




u


X
o.



c 2 S



5268



CO o3 00 as Oi CO



C^ CO <Ji



in (N (N <^



CO (N <N



1-1 CD CO to



00 lO CO (N



o ^ D

CO 0) fc-



00 CD CO



Oi OS (N C^ CO



i-H CO 00



CO lO f-H CD O
-^ CO C^ <N



lO lO CD 00 CO lO OS



<D CD N O



cn i-H CO



CO as o> CO



CO O) a>



S! -



lo CN ca



CO CO CO CO 03



"? s



CO «D «D «
CO 00 00 CO
CO CQ C3 CO



i-H t-H CO CJ> CJ



0> IN CS)



UO C3 C- '-I



1-3



00 CD CD CO



CO CD CD



CJ N CO in



CO U3 N CO <M IN



■-1 rH r-l It



CO lO C3



CQ <N C^ <N C^



4) "O WW

a ^ a '=;' ■=?

5 ^ S 6 o

2 S o ^ 2 2

f- f- J3 > Q D

2 2 M g g g

2 S O P m m



_, o



g 2 2
^ na to



o


CT>


J-j


o


()


^


CO


u




2


^


-1
5


p.


i-t







2 2



5269



SI



»-( CTi N cr> as
00 ai 00 c<» CO



ai 00 uo CO



^-^1 CSI CO CO 00



lO lO C£) <N IN yD



00 en ^







^




>>


^


tf


Q


o


o








i>






a.


D.



^tCSJ r-tCOlOCOCD 00



CO Ol rH tr-



Ol rH CO



ei



00 CN <N



(N t-l ,H



i-i o d



OJ 05 <N
CD CO (M



i-H CO CSl CO CO OS
rH W C) OI rH CD






^25

w fo c^
-■So
m a: g
< t^ o
H O o






P 151



UO O O O 00
(N CO <N O CM



CD CD -^



lO CO 00



COCDCO(NOOiOCMCNOSC<)COCD
FHCOCOr-lrHCQWCOfH CO



£. &



Online LibraryUnited States. Congress. Senate. Committee on theThe Industrial reorganization act. Hearings, Ninety-third Congress, first session [-Ninety-fourth Congress, first session], on S. 1167 (Volume pt. 7) → online text (page 36 of 140)