[comp.sys.ibm.pc.hardware] 286 w/co-processor vs. 386 price/performance ???

gchin@ssf.Eng.Sun.COM (Gary Chin) (10/31/90)

I am looking into a system for home/small business applications.
I was given some advice claiming that a 286 w/co-processor
is a better buy today than getting a 386SX or a 386.  If anyone
has any benchmark data or references, please post it.

Thanks,
Gary Chin

ergo@netcom.UUCP (Isaac Rabinovitch) (10/31/90)

In <144365@sun.Eng.Sun.COM> gchin@ssf.Eng.Sun.COM (Gary Chin) writes:

>I am looking into a system for home/small business applications.
>I was given some advice claiming that a 286 w/co-processor
>is a better buy today than getting a 386SX or a 386.  If anyone
>has any benchmark data or references, please post it.

Please.  Surely you've heard that there are lies, damn lies, and
benchmarks!

Everybody who sells a system can "prove" to you that it's the best
bargain around.  You should evaluate a system on whether it can do
what you want for a price you can afford rather than looking for the absolute
"best buy".

Nowadays, the fastest 286 board seems to be about 12 Mhz, the slowest
386 about 16 Mhz.  That's probably just because nobody's found it
worthwhile to design a 16 Mhz 286 motherboard, now that 386's are all
the rage.  If you need that extra speed, you buy the 386.  But how
many people need that extra speed?  Most probably don't even need
12Mhz.

The 386 knows how to make extended memory look like expanded memory.
If you need a lot of expanded memory, it makes more sense to buy a 386
than a 286 and an expanded memory board.  If you don't have any memory
intensive programs, or if you can make do with extended memory (and,
now that we don't need proper support for extended memory, we're
finally getting it!), you don't need the 386.

The 386 knows how to create "virtual machines".  Very handy if you're
going to run Windows or DesqView, especially if you want to multitask
programs that don't want to multitask.  But are you really going to
run such fancy programs?  Or are you just going to run WordPerfect and
Zelda Returns?

-- 

ergo@netcom.uucp			Isaac Rabinovitch
netcom!ergo@apple.com			Silicon Valley, CA
{apple,amdahl,claris}!netcom!ergo

WISE SAYING NEEDED.  Must reflect positive human values.  Gentle humor a
plus.  Cuties, pseudo-quotations, and jingoistic proverbs need not apply.

thoger@solan8.solan.unit.no (Terje Th|gersen) (11/01/90)

In article <144365@sun.Eng.Sun.COM>, gchin@ssf.Eng.Sun.COM (Gary Chin) writes:
|> I am looking into a system for home/small business applications.
|> I was given some advice claiming that a 286 w/co-processor
|> is a better buy today than getting a 386SX or a 386.  If anyone
|> has any benchmark data or references, please post it.
|> 
|> Thanks,
|> Gary Chin

I don't have any data-references for you, but here are some things I figured
out when trying to decide whether to go with a 16-20 MHz 286 or a 16MHz 386sx.
As you'll probably guess from the following, I got a 386sx.. :-)

A 16MHz 286 is marginally faster than a 16 MHz 386sx, but the '286
was born brain-dead, and hasn't gotten any smarter over the years.. :-)

A 386sx is only slightly more expensive than a similarly equipped '286, and 
you get all the memory-managment advantages a 386 offers, like the ability
do 'loadhi' TSR's and device drivers out of the 640k DOS base RAM area. This 
gives me 570K free base RAM now, vs. 510K when I had the same TSR's / device
drivers installed on a 286.

The 386sx or dx are also 5-10 times faster in getting in and out of protected
mode. This becomes an issue if you use exTended memory on your computer.
(using Windows 3.0, for example)

A 386sx/dx can also remap it's exTended memory into real EMS 4.0 for programs
who use this, without speed or compatibility penalties. Although some newer
286's support EMS 4.0 on the motherboard, most 286's need a dedicated EMS 4.0 
expansion board on the bus to do this. On my 386sx, motherboard-based RAM is 72% 
faster than the RAM I have on the expansion bus.


	 -Terje

____________________________________________________________________________
thoger@solan.unit.no       |                 Institute of Physical Chemistry
THOGER AT NORUNIT.BITNET   | Div. of Computer Assisted Instrumental Analysis
                           |               Norwegian Institute of Technology

ted@helios.ucsc.edu (Ted Cantrall) (11/01/90)

>I was given some advice claiming that a 286 w/co-processor
>is a better buy today than getting a 386SX or a 386.  If anyone
>has any benchmark data or references, please post it.
-------------------
If you just look at "benchmarks" (which can prove anything, depending on
the benchmark), the fast 286 may be a better deal. Perhaps that assumes
that you're only going to use your computer for running benchmarks.;-}
You should also be considering the longer term factors. With the far
superior instruction set and memory management of the 386, more and more
(don't ask me for numbers) programs are being written that won't work on
a 286. The prices of the 2 are so close, that I think you should buy for
the future!			-ted-
-------------------------------------------------------------------------------
ted@helios.ucsc.edu |"He has showed you, O man, what is good; and what does the
W (408)459-2110     |Lord require of you but to do justice and to love kindness
H (408)423-2444     |and to walk humbly with your God?" Micah 6:8 (RSV)

elliot@xenna.encore.com (Elliot Mednick) (11/02/90)

In article <15914@netcom.UUCP>, ergo@netcom.UUCP (Isaac Rabinovitch) writes:
> In <144365@sun.Eng.Sun.COM> gchin@ssf.Eng.Sun.COM (Gary Chin) writes:
> 
> >I am looking into a system for home/small business applications.
> >I was given some advice claiming that a 286 w/co-processor
> >is a better buy today than getting a 386SX or a 386.  If anyone
> >has any benchmark data or references, please post it.
> 
> [...]
> 
> Nowadays, the fastest 286 board seems to be about 12 Mhz, the slowest
> 386 about 16 Mhz.  That's probably just because nobody's found it
> worthwhile to design a 16 Mhz 286 motherboard, now that 386's are all
> the rage.  If you need that extra speed, you buy the 386.  But how
> many people need that extra speed?  Most probably don't even need
> 12Mhz.
> 
>[...]


In article <1990Oct31.185158.27459@idt.unit.no>,
thoger@solan8.solan.unit.no (Terje Th|gersen) writes:
> In article <144365@sun.Eng.Sun.COM>, gchin@ssf.Eng.Sun.COM (Gary Chin)
writes:
> [...]
> 
> A 386sx is only slightly more expensive than a similarly equipped '286, and 
> you get all the memory-managment advantages a 386 offers, like the ability
> do 'loadhi' TSR's and device drivers out of the 640k DOS base RAM area. This 
> gives me 570K free base RAM now, vs. 510K when I had the same TSR's / device
> drivers installed on a 286.
> 
> [...]
> 
> A 386sx/dx can also remap it's exTended memory into real EMS 4.0 for programs
> who use this, without speed or compatibility penalties. Although some newer
> 286's support EMS 4.0 on the motherboard, most 286's need a dedicated
EMS 4.0 
> expansion board on the bus to do this. On my 386sx, motherboard-based
RAM is 72% 
> faster than the RAM I have on the expansion bus.
> 
STOP!!  I dislike reading responses from people who speculate rather than who
know the facts.  All *new* 286 boards have LIM 4.0. A 286-16 board WITH
LIM 4.0 costs $125.  A 286-12 board is $89.  And, there is a 286-20 
available.

A 386sx-16 still costs more that $300 (ok, so I saw it once in some
small ad in Computer Shopper for $295, but...)  To me, the $175
savings was significant.  And I could do everything a 386 could at
about the same speed (did you ever notice that the SI and Landmark
ratings for the 286-16 were actally slightly GREATER than
the 386sx-16?  Of course, there is that BUG in the SI program...) just
not at the SAME TIME.  Of course, ther exists some programs (NOT
Windows) that multitask on a 286: Double DOS, vmix (shareware: see SIMTEL),
etc.  I have not tried these yet, but I will try vmix.

Naturally, I reserve the right to upgrade my motherboard at a future date
when my multiprocessing needs become significant.  Staying in an editor
while a C program is compiling or with a terminal window on doesn't count.

Oh, if you really NEED the multitasking of the 386 (I'm going to offend
a lot of people, now) you should forgo the sx and get a 20-25-33MHz 'dx.
With cache.  Then you have a REAL computer.

So there.
__
Elliot Mednick (elliot@encore.com) |  This .signature file is undergoing
Encore Computer Corp.              |  remodeling for your convenience.
Marlborough, MA. 01752             |  Please pardon our appearance.

poffen@sj.ate.slb.com (Russ Poffenberger) (11/02/90)

In article <144365@sun.Eng.Sun.COM> gchin@ssf.Eng.Sun.COM (Gary Chin) writes:
>I am looking into a system for home/small business applications.
>I was given some advice claiming that a 286 w/co-processor
>is a better buy today than getting a 386SX or a 386.  If anyone
>has any benchmark data or references, please post it.
>

A 286 processor running at the same speed as a 386SX will give comparable
results for the most part. A co-processor will only help for those math
intensive applications that know how to use it.

The big win with a 386SX is the memory management that the 286 can't even
come close to. The 386SX is functionally compatible with the 386(DX) and can
run the same protected and multitasking software (like windows in 386 mode)
that the 386 can.

Go with the 386SX, you will be happier in the long run.

Russ

baer@uwovax.uwo.ca (11/02/90)

A quick response to the conversation regarding 386's vs. 286's


> STOP!!  I dislike reading responses from people who speculate rather than who
> know the facts.  All *new* 286 boards have LIM 4.0. A 286-16 board WITH
> LIM 4.0 costs $125.  A 286-12 board is $89.  And, there is a 286-20 
> available.
> 

Agreed.  My older 286 requires dip switch setting, though, so I cannot
switch back and forward from expanded to extended memory "on the fly"
as I can with my 386.
> A 386sx-16 still costs more that $300 (ok, so I saw it once in some
> small ad in Computer Shopper for $295, but...)  To me, the $175
> savings was significant.  And I could do everything a 386 could at
> about the same speed (did you ever notice that the SI and Landmark
> ratings for the 286-16 were actally slightly GREATER than
> the 386sx-16?  Of course, there is that BUG in the SI program...) just
> not at the SAM
Agreed, one pays a slight performance price on normal DOS applications
for a 386sx vs. a 286 at the same speed (16mHz.).

BUT:   1) As a previous posting noted, switching into and out of protected
       mode takes longer with a 286 (important when using DOS extenders,
       some of which don't even bother to support 286's, OS/2, etc.).
       2) On math applications, a 386sx/387sx combination is CONSIDERABLY
       faster than a 286/287 application at the same clock speed.  One 
       needs a really computationally intensive program to see this, though
       (even stats packages performing low-level stats on large databases
      may not see this  difference since much of the "work" is I/O, etc.)
      I run a stats application that takes 4 hours on some models with
      my old 286 and I get by in less than 1 1/2 hours with my 386sx
     (both with a math chip) -- big difference.  

E TIME.  Of course, ther exists some programs (NOT
> Windows) that multitask on a 286: Double DOS, vmix (shareware: see SIMTEL),
> etc.  I have not tried these yet, but I will try vmix.
> 
> Naturally, I reserve the right to upgrade my motherboard at a future date
> when my multiprocessing needs become significant.  Staying in an editor
> while a C program is compiling or with a terminal window on doesn't count.
> 
> Oh, if you really NEED the multitasking of the 386 (I'm going to offend
> a lot of people, now) you should forgo the sx and get a 20-25-33MHz 'dx.
> With cache.  Then you have a REAL computer.
>          
Depends on what you want.  I'd probably opt for a dx too, but sometimes the
price difference between a 386sx and a 386dx machine is bigger than the 
286/386sx gap.  

Douglas Baer, 
University of Western Ontario, London, Canada N6A 5C2
Internet: BAER@UWO.CA    Bitnet: BAER@UWOVAX

freewill@pawl.rpi.edu (Jenni L. Lexau) (11/02/90)

  
Reasons why a 386sx is more usefull than a 286:
  
1.  It's a lot easier to debug on a 386, if the debugger can run in 386
mode, because the debugger is (almost) transparent to the program.  
Crashed programs don't crash the debugger, and breaking on memory
changes is much, much faster.  

  
2.  Windows 3.0 runs much better w/ DOS applications on a 386, even if
you don't multitask.  It allows better control of ports, cutting and
pasting of text and graphics to and from DOS applications, termination
of hung DOS applications, operation of DOS graphics applications, etc.
  
3.  Multitasking might not be *needed*, but it can be damn handy.  Handy
for impatient people who always want their machine to be faster.
  
4.  Number crunching software seems to run about twice as fast on a
387sx as a 287 of the same speed.  (trig is faster)
  
Quimby
  
(quimby@rpitsmts.bitnet, quimby@mts.rpi.edu)
  

harry@aeshq.UUCP (Harry Pulley) (11/03/90)

I have one benchmark for a 286-10 w/co-processor vs. 386-25, but it is very
one-sided in favor of the 286.  I wrote a fractal program in Turbo Pascal.  
This program calculates the newton iterations of root approximation for cubics
in the complex plane.

The only integer variables in the program are the loop counters; the float     
equations for the calculations take up most of the program.  

Anyway, here are the numbers: 286-10 w/cp finished a 640x480 screen in 33 min.
386-25 at the same res. took 47 min. to complete the task.  Note: these numbers
are approximate.

This sort of benchmark does not take advantage of any of the features of the 
386 (32-bit addressing, v-mem., etc).

The conclusion that I have drawn from these results, is that any small program
which uses little memory, disk-time, etc. and requires a large amount of 
number crunching time will probably run faster for less money if you go for a 
286.  You can get a nice 286 system with co-processor for <$2000.  The 80287-10
is ~$300 (up here in Canada...), while the cp's for the 386 line are over $700.

If you wish to do a large amount of math (spreadsheets, fractals, neural-nets,
graphics in perspective, etc...) and you have a limited budget, I would 
recommend a 286-12.  If you have the cash, a high speed 386/486 with a cp to
match will give you an incredibly fast system.
  
-- 
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
~     HCPIV (2b||!2b)     ~        Harry Pulley        ~   harry@aeshq.UUCP   ~ 
~    Computo, ergo sum    ~     Environment Canada     ~     TOR, ON, CA      ~
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

phil@brahms.amd.com (Phil Ngai) (11/03/90)

In article <13126@encore.Encore.COM> elliot@xenna.encore.com writes:
|STOP!!  I dislike reading responses from people who speculate rather than who
|know the facts.  All *new* 286 boards have LIM 4.0. A 286-16 board WITH

Too bad you don't know all the facts yourself. Nearly all
286 boards which have LIM 4.0 have software 4.0, not the
full hardware 4.0 and this a major failing if you want to
run DESQview.

(this is because most of the important features in the LIM 4.0
spec are "optional" and not required to be 4.0 compliant)

--
I voted. Did you?

silver@xrtll.uucp (Hi Ho Silver) (11/05/90)

$>I was given some advice claiming that a 286 w/co-processor
$>is a better buy today than getting a 386SX or a 386.  If anyone
$>has any benchmark data or references, please post it.
$If you just look at "benchmarks" (which can prove anything, depending on
$the benchmark), the fast 286 may be a better deal. Perhaps that assumes
$that you're only going to use your computer for running benchmarks.;-}

   A 10 MHz 8088 with an 8087 may well be faster at pure number-crunching
than a 16 MHz 386SX, but very few people perform strictly number-crunching.
A 16 MHz 80286 often proves marginally faster than a 16 MHz 386SX, but
the 386SX is probably the better choice for most people.

   The most important question to keep in mind when considering a hardware
purchase is "What will I be using it for during the time I expect to keep
this particular machine?"  If you're going to be doing a lot of numerical
stuff with software that uses a math coprocessor, then you'd probably be
best off with the 286+287 combination because it will do math significantly
faster than a 386SX or 386DX without a coprocessor.  But if that's only
going to be a small part of what you'll be doing, go for the 386SX or 386DX.

$You should also be considering the longer term factors. With the far
$superior instruction set and memory management of the 386, more and more
$(don't ask me for numbers) programs are being written that won't work on
$a 286. The prices of the 2 are so close, that I think you should buy for
$the future!			-ted-

   One big consideration for many people is whether or not you'll be using
Windows 3.  If so, don't even think about a 286.  Also, as Ted points out,
the memory management of the 386 is quite useful; it can be used for such
jobs as loading device drivers outside your 640K address space, leaving
you more memory, or for turning your extended memory into expanded memory,
with the ability to switch between the two at any time.
-- 
HI ROGER |Nikebo says "Nikebo knows how to post.  Just do it."| silver@xrtll
_________|-----------------------|_______________|------------|_____________
yunexus!xrtll!silver (L, not 1)  | Hi Ho Silver  | costing the net thousands
Silver:  Ever Searching for SNTF |i need a grilf | upon thousands of dollars