From: "Wayne J. Hyde" <w...@beach.cis.ufl.edu>
Here is my review. I cannot post it due to a problem with our news
server. If you would post it for me (if it isn't already posted --
I've sent this to a few people), I'd appreciate it.
Ok! Here is the review of NT & OS2 that I promised. Keep
in mind that these results were taken from *MY* testing
of NT on *MY* machine. For those who need to know again,
my machine configuration is as follows:
Gateway 2000 486/66, 256k cache, 16 meg SIMM.
ATI VLB GUP w/ 2 megs
2, WD Caviar 340 meg IDE drives.
-- NT resides on a 160 meg NTFS partition
** Logitech Bus Mouse
** Future Domain 860M SCSI card
TeXEL 3024 CDROM
** For some reason, NT assumes that the FD8xx controllers
use IRQ 5. I have mine set to IRQ 11 or 12 with my
mouse on IRQ 5, so NT didn't recognize my mouse or SCSI
card at first. I had to manually change the IRQ setting
under the registry to get both to work. Hmph.
NOTE: I *REALLY* tried to give NT a fair shot at
winning over my system. If NT turned out to be
better than OS/2, I would have purchased NT for myself.
[I was using a friend's upgrade from the beta since he
didn't have a system to run NT on]
During my tests, I tried to run the same programs on
both operating systems to keep everything consistent.
When I used NT's Paintbrush to load the large (2 meg) BMP
that hung the system for a while, I ran the Paintbrush from
win3.1 seamlessly under OS/2.
To simulate heavy loads, I ran QRT to hog CPU time.
(I closed all applications and minimized the ProgMan)
OS/2 2.1 using 8514/A drivers from 2.1 1024x768x256
Mode Score Elapsed Time
---------- --------- ------------
FullScreen : 6,799,416 63 seconds
Seamless : 4,578,050 65 seconds
While running a single QRT trace:
FullScreen : 3,872,493 66 seconds
Seamless : 2,649,938 70 seconds
NT using ATI 1024x768x256 drivers.
"Seamless" : 4,110,144 66 seconds
While running a QRT trace:
"Seamless" : 2,953,394 75 seconds
You can make any judgement from this data that you wish.
HW_TIMER was on for OS/2.
Seamless windows under OS/2 do seem _slightly_ slower than
"normal" speed on my system.
NT ran win16 apps just as well as OS/2 from what I could tell.
Some win16 apps ran faster under OS/2 than NT. Others ran at
the same speed.
DOS Compatibility and speed
Wolf3D runs BLAZINGLY fast under OS/2. (Faster than DOS).
Wolf3D also ran very well under NT, but slightly slower (~15%).
With 1 QRT running, Wolf3D still ran well under OS/2. ~5% reduction
in speed. NT, on the other hand, was very jerky and slow. ~40% slower.
Just about the same results as Wolf3d.
During a high speed (16.8k) transfer, OS2 and NT performed well.
2000+ cps under both OS's.
Tape Backup software (TAPE.EXE)
Tape ran well under OS/2. I could restore backups while multitasking
without any problems. While using my modem I did lose characters,
but I still have a 16450 UART (getting a 16550AFN on monday).
Tape did not run under NT. NT doesn't allow it.
MOD/MIDI players for my GUS
Run fine under OS/2. Music pauses during heavy multitasking. I don't
have any native GUS/MIDI players, so this should go away once they are
Don't work under NT. (NT [homey] don't play that!) ;)
I loaded up many applications at once and compared
how responsive the two OS's stayed. I used the applets
that came with the respective OS. I also ran a full
screen WIN-OS/2 session under OS/2.
Under NT, I could load a few applications without any
slowdowns. I played with File Manager, the System
Administration Tools, and other programs. NT _DID_
start to swap immediately, but this wasn't a problem -- yet.
After I started loading more apps, swapping became a problem.
When NT isn't swapping, switching between processes is pretty
fast. When NT starts swapping, switching windows takes
a while longer. CPU intensive apps also caused NT's task
switching to slow down in a _BIG_ way. The foreground app
would run OK, but if you tried to switch to another window,
you had better go get a drink.
[I had my girlfriend use OS/2 and NT... see below]
Under OS/2, I loaded up a a few of the "Productivity" apps.
No swapping occurred until I loaded MANY apps. OS/2 really
does run great in 16 megs. Anyways, swapping does slow
down things a little, but it wasn't as bad as under NT.
A note about OS/2's swapfile. I used to have the swapfile set
initially at 2048k. My swapfile would grow up to 10 megs if
I loaded a fullscreen WIN session and a few other programs.
I recently changed the minimum size to 8192k, and my swapfile
has very RARELY grown any larger. It seems that OS/2
over compensates when it has to expand the swapfile. Even if
I load 3 fullscreen WIN-OS/2 sessions and a few OS/2 programs,
my swapfile does not grow. The same procedure would have
made the swapfile grow to about 15 megs or more. Hmmmmm.
I made the following batch file and ran it under both OS's:
unzip -t temp.zip
After I ran it, system performance was slower, but NT was still
I could notice a slight degradation of performance under OS/2, but
not as much as under NT.
I ran 4 raytraces in windowed sessions under NT and OS/2 for 5 minutes.
At the end of 5 minutes I checked how many lines had been traced.
During this time I putzed around and played with system settings
and loaded up a few (small) apps like the clipboard viewer and
a readme file. I also played with File Manager (under both OS's).
Under NT, performance became unbearable. The raytraces were bringing
the system to almost unusable levels. Window redraws were SLOW,
SLOW SLOW! I tried to switch to a different window and it took 30
seconds to redraw the window. Pretty bad performance. Again,
Program Manager did not slow down, but I've been corrected that
this is actually a "good" feature. I'll accept that point.
Under OS/2, performance was slower, but VERY *VERY* usable. The
system responded very well to anything I did. Windows were redrawn
quickly, applications loaded fast; it was nice. This is the
way I want my OS to handle.
After 5 minutes, NT's traces had completed about 100 lines each.
OS/2's had completed about 105 each in the same time. Even though
OS/2 and NT completed roughly the same number of lines, NT's performance
was very poor compared to OS/2's.
Oh, BTW, I'm running 11 QRT raytraces under OS/2 right now
while I type this post. OS/2 is performing VERY well
under stress. In fact, a while ago (under the beta), I
ran 20 raytraces at once under OS/2. Granted system
performace was much slower than normal, but the system
was still responsive unlike NT.
Ok, I just had my girlfriend use my system and judge its performance.
She switched between apps and ran a few programs under OS/2.
She has a 386/25 with 4 megs and runs Windows 3.1. She said that
it looks kinda like Windows, and "it was running fine." !! I guess
she is just used to Windows on her "slow" machine. :) I'd
definitely say that OS/2 is running slow under the circumstances,
but it is still usable IMHO (and Heather's). :)
[I'll reboot in a sec and have her judge NT while running only 4
traces instead of 11. I don't want to wait around forever
trying to close the windows.]
OK, I rebooted, had her play with NT, and here is what she had to say:
(this is her typing)
It's a lot slower than the other one, especially opening things up.
Once he loaded the four programs [QRT's] things slowed down even more.
At that point the difference between the two systems was unmistakable.
Can I go home now?
(end of Heather's editorial)
hehe. :) She doesn't like my computer--it is the "other" woman.
Anyways, when I first loaded NT for her, notice that I did not run
anything before she used it. She thought it was slower than OS/2
when OS/2 was running the 11 (ELEVEN!) raytraces. NT's tasking
was set on "foreground gets BEST priority".
Oh, don't flame ME, these were HER opinions. :)
Oh, I wish to elaborate more on the priority boosts in OS/2 and
NT when an application is in the foreground.
OS/2 does give apps a boost when they have the focus, but it is a
very small boost. When I ran the 4 traces, I gave one the focus.
The foreground trace finished about 30 lines while the others
finished about 15 lines in a given time interval.
When NT is set on the "give foreground app best priority", it really
means best! During the same test, the foreground trace finished
about 60 lines while the others finished ONE or TWO! WOW!
With the tasking set to the middle level, the boost was less drastic.
Multitasking was even when set to EQUAL.
I know, I know. Raytraces are very CPU intensive and don't
represent the every day usage of an operating system. Well, I didn't
have anything else that I could run on both systems to gauge their
performance. If someone else wishes to run some other tests,
go ahead. I'd like to hear your results. I'd also like someone
to try what I did on their NT and OS/2 system to see if they
get the same results as I have.
I think I have run enough tests on the two operating systems to
say that OS/2 handles system loads much better than NT (at least
on my system). Maybe there is something drastically wrong
with my NT setup. I don't know. It just doesn't run well.
OS/2 is ...
read more »