Path: sparky!uunet!stanford.edu!bcm!lib!oac.hsc.uth.tmc.edu!jmaynard
From: jmayn...@oac.hsc.uth.tmc.edu (Jay Maynard)
Newsgroups: comp.os.os2.misc
Subject: My OS/2 is REALLY quick now...
Message-ID: <7072@lib.tmc.edu>
Date: 4 Aug 1992 21:05:32 GMT
Sender: use...@lib.tmc.edu
Organization: UT Health Science Center Houston
Lines: 38
Nntp-Posting-Host: oac.hsc.uth.tmc.edu
I thought that OS/2 ran just fine on my previous system. It was a Mylex
80386SX-16, with 8 meg of RAM, an Everex EV-334 MFM controller, a Trident 9000
with 512K running in 800x600x16 mode, and two disk drives: an ST-251 as an
HPFS C:, and a Maxtor XT-1140 as a FAT D:. I had to use the IBMINT13 driver.
I'd get response in reasonably short order after asking the system to do
something, and didn't know what was possible...
Last night, I completed upgrading the system. The motherboard was replaced
with a no-name 386-40 (with an AMI BIOS dated 5/5/91), the Trident was
switched from 8 to 16-bit interface mode, and the disk controller was replaced
with an Adaptec ACB2372B; the Maxtor was reformatted for RLL, with a 40 MB C:
partition and a 191 MB D: partition. Both partitions were formatted as HPFS,
with all of OS/2 installed on C: per the recommendation in the redbooks. (They
recommend this approach so that an upgrade requiring a reformat of the OS/2
system partition isn't a major headache.) Finally, I picked up the
IBM1S506.ADD fix (the one dated June, not February; it's the IBM506.ZOO file)
from ftp-os2.nmsu.edu and put that on. During the reformat, I formatted the
full 1224-cylinder capacity of the Maxtor (as opposed to the DOS 1024-cylinder
limit), and had the Adaptec translate the drive to 63 virtual sectors per
virtual track. About the only thing that _didn't_ change was that I still have
only 8 MB of RAM.
I don't know which upgrade made things speed up, but it feels much snappier.
The first note was during the reinstallation of OS/2...the disk icons at the
bottom of the installation progress window appeared almost instantaneously,
instead of being drawn one at a time over the space of a few seconds.
Exploding windows are much more explosive. Even the OS/2 logo screens at boot
time don't stay around very long.
How much effect do you, the assembled gurus and near-gurus of
comp.os.os2.misc, think each of the upgrades had? What has the biggest impact
on performance, if RAM size is held constant?
--
Jay Maynard, EMT-P, K5ZC, PP-ASEL | Never ascribe to malice that which can
jmayn...@oac.hsc.uth.tmc.edu | adequately be explained by a .sig virus.
"Right now, because of you, this city is being overrun by baboons." "Don't
you think that's the responsibility of the voters?" -- _Naked Gun 2-1/2_
Path: sparky!uunet!stanford.edu!rutgers!njitgw.njit.edu!hertz.njit.edu!dic5340
From: dic5...@hertz.njit.edu (David Charlap)
Newsgroups: comp.os.os2.misc
Subject: Re: My OS/2 is REALLY quick now...
Message-ID: <1992Aug5.184824.4021@njitgw.njit.edu>
Date: 5 Aug 92 18:48:24 GMT
References: <7072@lib.tmc.edu>
Sender: n...@njit.edu
Organization: New Jersey Institute of Technology, Newark, N.J.
Lines: 26
Nntp-Posting-Host: hertz.njit.edu
In article <7...@lib.tmc.edu> jmayn...@oac.hsc.uth.tmc.edu (Jay Maynard) writes:
>
>How much effect do you, the assembled gurus and near-gurus of
>comp.os.os2.misc, think each of the upgrades had? What has the biggest impact
>on performance, if RAM size is held constant?
Probably (in this order):
1) Increased CPU speed: 40MHz over 16MHz. Also, the 32-bit memory
bandwidth of the 386DX will double the speed of memory fetches.
2) Switching the video card from 8-bit to 16-bit mode. This doubles
the speed of transfers to video memory. Some video cards (or
motherboards) will use fewer wait states transferring to 16-bit
cards, which will speed things up even more.
3) The new disk drive controller. Note that there's a 50% increase in
drive speed simply by changing the format from MFM to RLL. Of
course, your drive MUST be certified for RLL, since you may create
a lot of errors by simply reformatting an MFM drive.
4) Losing the sector/track translation. I think OS/2 can optimize its
requests based on drive characteristics. It (obviously) will
optimize incorrectly if the drive is translating its metrics.
--
|) David Charlap "I don't even represent myself
/|_ dic5...@hertz.njit.edu sometimes so NJIT is right out!.
((|,)
~|~ Hi! I am a .signature virus, copy me into your .signature file.
Path: sparky!uunet!europa.asd.contel.com!darwin.sura.net!jvnc.net!
netnews.upenn.edu!dsinc!ub!acsu.buffalo.edu!ubvmsb.cc.buffalo.edu!v053qpgh
From: v053q...@ubvmsb.cc.buffalo.edu (Eric W Sarjeant)
Newsgroups: comp.os.os2.misc
Subject: Re: My OS/2 is REALLY quick now...
Message-ID: <Bss39B.n51@acsu.buffalo.edu>
Date: 10 Aug 92 18:29:00 GMT
References: <7072@lib.tmc.edu> <1992Aug5.184824.4021@njitgw.njit.edu>
Sender: n...@acsu.buffalo.edu
Organization: University at Buffalo
Lines: 47
News-Software: VAX/VMS VNEWS 1.41
Nntp-Posting-Host: ubvmsb.cc.buffalo.edu
In article <1992Aug5.184824.4...@njitgw.njit.edu>, dic5...@hertz.njit.edu
(David Charlap) writes...
>In article <7...@lib.tmc.edu> jmayn...@oac.hsc.uth.tmc.edu (Jay Maynard) writes:
>>
>>How much effect do you, the assembled gurus and near-gurus of
>>comp.os.os2.misc, think each of the upgrades had? What has the biggest impact
>>on performance, if RAM size is held constant?
>
>Probably (in this order):
>
>1) Increased CPU speed: 40MHz over 16MHz. Also, the 32-bit memory
> bandwidth of the 386DX will double the speed of memory fetches.
I agree, for any activity enhancing the performance level of your current
CPU will enhance your system overall. This is probably where you're getting
the majority of your speed inducing qualities.
>2) Switching the video card from 8-bit to 16-bit mode. This doubles
> the speed of transfers to video memory. Some video cards (or
> motherboards) will use fewer wait states transferring to 16-bit
> cards, which will speed things up even more.
I'm going to disagree no to start a war but to point out that the OS/2
video drivers are all very limited right now. For most machines, getting
640x480x16 is about all you're going to get, and this can be accomplished
quite readily with an 8 bit adaptor. Keep in mind, however, that future
incarnations of OS/2 might support a broader scope of video boards. When a
true 32 bit graphics engine is made available, your 16 bit card will prove
itself a very worthwhile investment.
>3) The new disk drive controller. Note that there's a 50% increase in
> drive speed simply by changing the format from MFM to RLL. Of
> course, your drive MUST be certified for RLL, since you may create
> a lot of errors by simply reformatting an MFM drive.
Definitely. Taken together with your CPU upgrade, this is probably the other
half of your improvement.
>4) Losing the sector/track translation. I think OS/2 can optimize its
> requests based on drive characteristics. It (obviously) will
> optimize incorrectly if the drive is translating its metrics.
Again, definitely -- this follows from (4).
==============================================================================
Eric W. Sarjeant | "I am only too conscious that I have no
sUnyaB | claim to wisdom, great or small..."
v053q...@ubvms.bitnet |
sarje...@aol.com |
Path: sparky!uunet!wupost!bcm!lib!oac.hsc.uth.tmc.edu!jmaynard
From: jmayn...@oac.hsc.uth.tmc.edu (Jay Maynard)
Newsgroups: comp.os.os2.misc
Subject: Re: My OS/2 is REALLY quick now...
Message-ID: <7094@lib.tmc.edu>
Date: 10 Aug 1992 18:50:13 GMT
References: <7072@lib.tmc.edu> <1992Aug5.184824.4021@njitgw.njit.edu>
<Bss39B.n51@acsu.buffalo.edu>
Sender: use...@lib.tmc.edu
Organization: UT Health Science Center Houston
Lines: 68
Nntp-Posting-Host: oac.hsc.uth.tmc.edu
[The message that Eric was replying to expired before I got to it. I'll
comment on both in one shot.]
In article <Bss39B....@acsu.buffalo.edu> v053q...@ubvmsb.cc.buffalo.edu
(Eric W Sarjeant) writes:
>In article <1992Aug5.184824.4...@njitgw.njit.edu>, dic5...@hertz.njit.edu
(David Charlap) writes...
>>1) Increased CPU speed: 40MHz over 16MHz. Also, the 32-bit memory
>> bandwidth of the 386DX will double the speed of memory fetches.
...at least for motherboard memory. Just how CPU-intensive is the WPS, anyway?
>>2) Switching the video card from 8-bit to 16-bit mode. This doubles
>> the speed of transfers to video memory. Some video cards (or
>> motherboards) will use fewer wait states transferring to 16-bit
>> cards, which will speed things up even more.
I don't think my motherboard does that explicitly beyond the inherent delay in
doing two bus cycles to transfer a word of data instead of one. In any case,
this implies that a large fraction of accesses to video memory are done 16
bits at a time; is this really the case?
>I'm going to disagree no to start a war but to point out that the OS/2
>video drivers are all very limited right now. For most machines, getting
>640x480x16 is about all you're going to get, and this can be accomplished
>quite readily with an 8 bit adaptor. Keep in mind, however, that future
>incarnations of OS/2 might support a broader scope of video boards. When a
>true 32 bit graphics engine is made available, your 16 bit card will prove
>itself a very worthwhile investment.
My system is running 800x600x16 on a 512K Trident 9000 adapter.
I'm anxiously awaiting the 32-bit graphics engine, as I can see nothing but
performance benefit from it.
>>3) The new disk drive controller. Note that there's a 50% increase in
>> drive speed simply by changing the format from MFM to RLL. Of
>> course, your drive MUST be certified for RLL, since you may create
>> a lot of errors by simply reformatting an MFM drive.
>Definitely. Taken together with your CPU upgrade, this is probably the other
>half of your improvement.
Hmmm. I'd forgotten about the transfer speed improvement of the RLL
controller, but that's a really good point; this would also improve paging
performance, and on a system with virtual memory, paging I/Os are a major
performance hit. Cutting down the time one takes is going to help matters
quite a bit. By the same token, I was able to format for 1:1 interleave, and
that has to have helped as well.
BTW, I'm using a Maxtor XT1140; I don't know if it's formally RLL certified or
not, but it appears to work fine, and the Adaptec formatter reported an
acceptable number of bad sectors. I do know that the guy I got the controller
from was using an XT1140 with it with no problems. (BTW, anyone have an
AHA1542 they want to get rid of? He wants one in trade.)
>>4) Losing the sector/track translation. I think OS/2 can optimize its
>> requests based on drive characteristics. It (obviously) will
>> optimize incorrectly if the drive is translating its metrics.
>Again, definitely -- this follows from (4).
I would agree, but turning on sector translation got me 37 MB extra space. It
seems that the install process wouldn't recognize the 200 extra cylinders my
drive has (1224 total) without it, despite using the latest drivers. I'll take
a bit of a performance hit for the extra space, especially since I got quite a
bit of improvement otherwise.
--
Jay Maynard, EMT-P, K5ZC, PP-ASEL | Never ascribe to malice that which can
jmayn...@oac.hsc.uth.tmc.edu | adequately be explained by stupidity.
"This is a network here, right? How about giving a little thought to
interoperability?" -- John F. Woods, to Brian Kantor