[comp.sys.ibm.pc.misc] '386's SX's and DX's

ttak@uhura.cc.rochester.edu (Timothy Takahashi) (07/21/90)

In article <46500144@uxa.cso.uiuc.edu> mms00786@uxa.cso.uiuc.edu writes:
>
>This post (unwarranted?) may not be true for all people. First, the 386SX
>is a *full* 32 bit microprocessor; only thing is, its interface to the 
>external world is through 16 bits. All software intended to run on the 
>386DX should run absolutely correctly on the SX.

Agreed, my 386sx clone (a swan 386sx) is subjectively identical in speed
to the lab's 386dx (a 16mhz ibm ps/2 model 70). Our 16mhz 386dx inboard/386
xt feels slower due to the slow hard drive and video (my everex ega seems
much faster than our ati vga-wonder 256k)

with 20mhz 386sx's available, there is no need to buy anything less than 
a 25mhz 386dx (no real speed advantage with today's software)


tim

YTHRROUS@MTUS5.BITNET (Roger Rouse) (07/23/90)

The 386SX is a very nice middle ground between a 386 and a 286.  I just
purchased a 386SX-16 machine, and am throroughly impressed.  It combines
decent speed with the special modes that the 386 offers over the 286, and
all for about $300-$400 less than a pure 386.

jdu4855@cs.rit.edu (Unrue Jack D) (07/24/90)

In article <1990Jul22.234030.25727@utstat.uucp> philip@utstat.uucp (Philip McDunnough) writes:
>
>But isn't the 386sx restricted in the amount of memory it can address? If
>that's the case, then would a 386sx be able to run OS/2 ver 2.0 ? I guess
>I'm wondering if the SX includes ALL the 32 bit machine instructions of a
>DX? Furthermore, what is the size of the data path between a 386SX and a
>387SX?
>

Yeah, I think you and I are saying the same thing.  SX's use a 16 bit bus, and
OS/2 ver 2.0 inhabits a 32 bit address space, so the two don't go together.  I
think that's as far as I'll go, because I'm not sure about the data path 
between the 386SX and 387SX.


--
--
Jack Unrue
ultb::"jdu4855@cs.rit.edu"

jca@pnet01.cts.com (John C. Archambeau) (07/24/90)

v081nhdb@ubvmsd.cc.buffalo.edu (Nina Banerjee) writes:
>If you are going to get a '386, get a REAL '386! Namely a DX. The SX's are
>actually 16 bit microprocessors, or so a friend of mine in the Computer 
>business tells me. If you are going to spend the money on a '386 (either a
>new motherboard, or a whole new system), you might as well get a good one
>that has a 32 bit Microprocessor.

That depends on what you're going to do.  Keep in mind that the ISA (or AT)
bus only is a 16-bit bus.  The register size of the 386SX is 32-bit.  The
address space is 24-bit (or 16 Mb of physical memory).  The 386SX does have a
place in the industry.  If all you need is 386 compatability on an ISA bus
machine, then the 386SX is perfect.  My only regret in getting a 386SX is not
waiting for the 20 MHz version with cache.  ALR has already started shipping
the 20 MHz 386SX PowerFlex modules w/32K cache.

I run SCO Xenix 386 2.3.3 on my 16 MHz 386SX and it performs well.  Now if
you're going to get a MicroChannel or EISA machine, then you might as well go
full blown 386DX since you have a 32-bit system bus you're dealing with, but
with respect to ISA, if you can live at 16 or 20 MHz, then the 386SX is for
you.
 
     // JCA

 /*
 **--------------------------------------------------------------------------*
 ** Flames  : /dev/null                     | Small memory model only for
 ** ARPANET : crash!pnet01!jca@nosc.mil     | Unix?  Get the (*bleep*) out
 ** INTERNET: jca@pnet01.cts.com            | of here!
 ** UUCP    : {nosc ucsd hplabs!hd-sdd}!crash!pnet01!jca
 **--------------------------------------------------------------------------*
 */

werner@aecom.yu.edu (Craig Werner) (07/24/90)

	An analogy that probably should be added to the FAQ:

	8086 : 8088 :: 386DX : 386SX

that is (16/16 & 16/8,  32/32 and 32/16), with identical instruction sets
and addressing capabilities.
-- 
	        Craig Werner   (future MD/PhD, 5.5 years down, 2.5 to go)
	     werner@aecom.YU.EDU -- Albert Einstein College of Medicine
              (1935-14E Eastchester Rd., Bronx NY 10461, 212-931-2517)
          "Disinformation is one thing, but misinformation is unforgiveable."

rob@conexch.UUCP (Robert Collins) (07/24/90)

In article <31122@eerie.acsu.Buffalo.EDU> v081nhdb@ubvmsd.cc.buffalo.edu writes:
>If you are going to get a '386, get a REAL '386! Namely a DX. The SX's are
>actually 16 bit microprocessors, or so a friend of mine in the Computer 
>business tells me. 
This is like saying:  "Do as I say, as I am an authority on this matter...or
so a friend in the computer business tells me."  Actually, the SX is a 32-bit
uP, with a 16-bit address and data bus.  Internally, it is a full 32-bit uP.
In fact, the SX, and DX are indistinguishable from the software's point of
view.  Even with Intel's help, I haven't figured up a software method to tell
the two CPU's apart.

>If you are going to spend the money on a '386 (either a
>new motherboard, or a whole new system), you might as well get a good one
>that has a 32 bit Microprocessor.

Did you consider lap-top computers?  A DX is a power-hog compared to an SX.
If you were to buy a 386 laptop,it certainly would have an SX in it.


-- 
"Worship the Lord your God, and serve him only."  Mat. 4:10
Robert Collins                 UUCP:  ucbvax!ucivax!icnvax!conexch!rob
HOMENET:  (408) 225-8002       UUCP:  uunet!ccicpg!conexch!rob
WORKNET:  (408) 432-6200 x4356

shack@cs.arizona.edu (David Michael Shackelford) (07/25/90)

In article <1776@cs.rit.edu> jdu4855@cs.rit.edu (Unrue Jack D) writes:
>
>But isn't the 386sx restricted in the amount of memory it can address? If
>that's the case, then would a 386sx be able to run OS/2 ver 2.0 ? I guess
>I'm wondering if the SX includes ALL the 32 bit machine instructions of a
>DX? Furthermore, what is the size of the data path between a 386SX and a
>387SX?
>

I might be mistaken, but the *only* difference between the SX and DX chips
is the *data* bus size, 16 for SX and 32 for DX.  This results in the
requirement for SX's to do two physical memory accesses for each 32 bit
instruction.  The address bus sizes are the same (I think--haven't seen
the actual chips)  and the SX contains all instructions and registers
from the DX.

The effect of all of this is:  OS/2 2.0 should run on an SX the same as
any other 386 specific software would: slower for memory access, but otherwise
identically.

Additionally, even if the system bus has a 24 bit address size, OS/2 v2.0
would run in 16Mb easily.  It would work better on more memory, but I think
the minimum will be about 8mb if they follow the historical path.

This is all opinion, I don't have the spec sheets sitting before me!

Dave.

steveha@microsoft.UUCP (Steve Hastings) (07/26/90)

In article <1990Jul22.234030.25727@utstat.uucp> philip@utstat.uucp (Philip McDunnough) writes:
>>
>>But isn't the 386sx restricted in the amount of memory it can address? If
>>that's the case, then would a 386sx be able to run OS/2 ver 2.0 ?

In article <1776@cs.rit.edu> jdu4855@cs.rit.edu (Unrue Jack D) writes:
>Yeah, I think you and I are saying the same thing.  SX's use a 16 bit bus, and
>OS/2 ver 2.0 inhabits a 32 bit address space, so the two don't go together.


The 386SX has a 16-bit bus, which means that to do things in 32 bits
requires two bus cycles.  You could therefore run OS/2 2.0, or any other
386-specific software, on a 386SX.  It will just be slower.


For the ultimate in slowness, consider this:  you can buy a pin-compatible
replacement for a 286 -- a small daughterboard with a 386SX and some glue
logic.  (It costs about $300.)  It will run at speeds up to 16 MHz, but of
course it will run slower if you want it to.  I have this mental image of
an original IBM AT, running at 6 MHz, with one of these 386SX conversions
installed, running OS/2 2.0 -- it boggles my mind to think about it.  :-)
-- 
Steve "I don't speak for Microsoft" Hastings    ===^=== :::::
uunet!microsoft!steveha  steveha@microsoft.uucp    ` \\==|

phil@brahms.amd.com (Phil Ngai) (07/27/90)

In article <42788@conexch.UUCP> rob@conexch.UUCP (Robert Collins) writes:
|This is like saying:  "Do as I say, as I am an authority on this matter...or
|so a friend in the computer business tells me."  Actually, the SX is a 32-bit
|uP, with a 16-bit address and data bus.  Internally, it is a full 32-bit uP.

16-bit address bus, huh? Are we supposed to consider you an authority?

--
--
Phil Ngai, phil@amd.com		{uunet,decwrl,ucbvax}!amdcad!phil
We're saving water: we use disposable diapers.

jdu4855@cs.rit.edu (Unrue Jack D) (07/27/90)

In article <56069@microsoft.UUCP> steveha@microsoft.UUCP (Steve Hastings) writes:
>
>The 386SX has a 16-bit bus, which means that to do things in 32 bits
>requires two bus cycles.  You could therefore run OS/2 2.0, or any other
>386-specific software, on a 386SX.  It will just be slower.
>

I have received a good amount of e-mail and have read several postings which
point out the error in my previous posting, wherein I stated my mistaken
opinion that OS/2 and the SX don't go together. 
 
In particular, Doug MCDonald illustrated this by writing,

      "You are arguing the same as 'the 8088 has an 8 bit bus, so it can't
       run MS-DOS, a 16 bit OS,' which is equally wrong."


--
--
Jack Unrue
ultb::"jdu4855@cs.rit.edu"