Документ взят из кэша поисковой машины. Адрес оригинального документа : http://crydee.sai.msu.ru/ftproot/pub/misc/doc/humor/microsoft/ms.response Дата изменения: Tue May 30 19:42:16 1995 Дата индексирования: Tue Apr 12 17:26:17 2016 Кодировка: Поисковые слова: п п п п р п р п р п р п

Newsgroups: comp.os.os2.advocacy
From: gordon@lab.lwpi.com (Gordon Letwin)
Subject: Re: IBM Whitepaper, FUD, 4M, etc.
Date: Tue, 11 Oct 1994 15:23:26 GMT
Lines: 198

Every few months I sample this group to gauge the level of sanity; usually
it's nearly nonexistant. This past year, however, sanity has been
gradually creeping up. I guess the stalwarts of this group are getting
their meds better adjusted. :-)

One thing I particularly admire is the way that the true faithful are able
to completely ignore that which they wish weren't true. A facility that
comes in handy for religious fanatics everywhere.

For example, a couple of years ago the writer for the Wall Street Journal
that specialized in the computer trade wrote a book about IBM. It
had some great comments about OS/2 development from folks such as Ed
Iacabucchi (sp?) - who was the head of the IBM design team and my
"opposite number" there during the design stage. He's quoted as saying
that Microsoft's approach was to have a small number of talented programmers,
whereas the IBM approach was to hire "masses of asses". But gee, I
never saw this book mentioned once in this group, nor, needless to say,
did I ever see any discussion of these evaluations of IBM's efforts made
by IBM's own senior people (at that time).

This brings me to the reason I decided to waste 10 minutes reaping abuse
from crazies. I read with interest IBM's "white paper" on Warp vs
WIndows 95, because I'd just finished reading a Microsoft paper on
performance benchmarks between the two systems. I was looking to see
what IBM had to say about real life issues, such as performance.
And what they had to say was nothing. I read lots of "blind 'em with
science" talk about the heartbreak of not having a foobarish distributarian
interprise model in your software, but nothing about reality - how well
does it work? (IBM, in my experience, is the absolute master of the bullshit
memo. There's an amazing story about this that I may post one day. Remember
that IBM lawyers threatened me to try to make me stop posting.)

So I excerpt the performance parts of the Microsoft paper here for your
benefit, since I haven't seen any posts which show that anyone in
this group is even aware of this work. Note that most of the benchmarks
were run with the Windows 95 beta running in 4 megs.

Don't bother to send me abuse; I'm not interested.

Gordon Letwin

- - - - - - - -

Microsoft WinNews Electronic Newsletter,
Vol. 1, #3, October 10, 1994
****************************************

Windows 95 Performance

In tests, performed on two common PC systems, Windows 95 and
Windows for Workgroups performed similarly, and proved to be
significantly faster than OS/2 Warp (beta 2). The test systems
were 486/66 machines from IBM and Dell, each configured with 4,
8 and 16 MBytes of memory.

Here is a summary of the significant findings.

* Disk throughput: Windows 95 and Windows for Workgroups
have 300% - 400% more disk throughput of OS/2 Warp.
* Graphics performance: Windows 95 and Windows for Workgroups
have over 200% more graphics performance than OS/2 Warp.
* Application startup: Windows 95 and Windows for Workgroups
load applications 600% - 800% faster than OS/2 Warp in
a 4 MB configuration.
* Application execution: Windows 95 and Windows for
Workgroups ran Windows applications 150% - 210% faster than
OS/2 Warp.
* Multitasking: Windows 95 and Windows for Workgroups ran
multiple applications and tasks 130% - 180% faster than
OS/2 Warp.

In general, Windows 95's and Windows for Workgroups' performance
lead over OS/2 increased as the amount of RAM in the PC decreased.
The lead was largest on 4MB systems.

Here are some of specific results:

Disk I/O Performance

Windows 95 and Windows for Workgroups deliver 300% to 400%
more disk I/O performance than OS/2 Warp.

IBM ValuePoint 466DX2/D

Disk WinMark (KB/sec - 4 MB 8 MB 16 MB
higher is better)
Windows 95 1,030 1,180 1,190
Windows for Workgroups 1,010 1,120 1,220
OS/2 Warp 297 366 394

Windows 95 % faster than 347% 322% 302%
Warp

Graphics Performance

Windows 95 and Windows for Workgroups deliver over 200%
more graphics performance than OS/2 Warp.

IBM ValuePoint 466DX2/D

Graphics WinMark (M 4 MB 8 MB 16 MB
Pixels/sec - higher is
better)
Windows 95 12.5 12.6 12.7
Windows for Workgroups 10.8 10.8 10.9
OS/2 Warp 5.2 5.3 5.3
Windows 95 % faster than 240% 239% 240%
Warp

Application Load Time Performance

Windows 95 and Windows for Workgroups outperformed OS/2 Warp
by over 250% when starting applications. Application load
time is the amount of time between launching Word for Windows
6.0a and when the application is ready to accept input. The
difference was largest in the 4 MB RAM configuration where
OS/2, because of its large size, exhibits long application
load times due to excessive disk paging.

Time required to load first application

IBM ValuePoint 466DX2/D

Time to Load Word 6.0c 4 MB 8 MB 16 MB
(seconds, lower is better)
Windows 95 12.9 6.4 5.6
Windows for Workgroups 11.6 5.7 5.5
OS/2 Warp 101.3 21.8 16.2
Windows 95 % faster than 785% 341% 289%
Warp

Application Execution Performance

To test the speed of application execution, Windows Magazine
created benchmark macros for Word and Excel that execute a
\wide variety of application functions. The run time of these
macros is influenced by graphics performance, disk performance
and available memory. Application load time is not included.
When running a single macro in Excel 5, Windows 95 and Windows
for Workgroups proved to be 150% - 210% times faster than OS/2 Warp.

Single task application performance

IBM ValuePoint 466DX2/D

Excel Macro (seconds - 4 MB 8 MB 16 MB
lower is better)
Windows 95 193 106 102
Windows for Workgroups 181 118 112
OS/2 Warp 384 159 152
Windows 95 % faster than 199% 150% 149%
Warp

Multi-task application performance

To test multitasking performance, both the Excel and Word
macros were executed in side-by-side windows with the times
for each added to give an overall score. OS/2 has the option
of running each Windows application in a separate session - which
ostensibly provides better multitasking because application execution
is preempted by OS/2 instead of waiting for each to yield. However,
test results show that this option does not improve performance
running multiple tasks. This separate session option was used for
the OS/2 16 MB test, but caused the 8 MB test to fail with insufficient
memory, in which case a single shared session was used.

Windows 95 and Windows for Workgroups ran multiple applications 130% -
180% faster than OS/2 Warp in either shared or separate Windows
sessions.

IBM ValuePoint 466DX2/D

Simultaneous Excel & Word
macros 8 MB 16 MB
(total time score - lower
is better)
Windows 95 295 247
Windows for Workgroups 349 269
OS/2 Warp 468 319
Windows 95 % faster than 159% 129%
Warp

OS/2 Warp failed simultaneous macro test w/8MB RAM
w/Insufficient memory error in Word approx. 5:15 into the
tests when the applications were run in separate sessions.
Time shown is in a single shared Windows session

Component benchmarks (disk throughput and graphics performance)
were performed using the Ziff-Davis Winbench 4.0 suite. Appli-
cation benchmarks were performed using the Windows Magazine
WinTune Excel and Word Macro benchmarks.

More detail on these benchmarks is available in the file
DPERF4.ZIP (APERF4.ZIP or PPERF4.ZIP for TXT or PostScript)
on the WinNews servers on Compuserve, the Internet, AOL, Genie
and Prodigy.