486SX vs. 386DX
Douglas Stevens
386DX-40MHz, all other factors being equal?
--
----------------------------------------------------
Douglas L. Stevens | E-Mail:
University of Calgary | dste...@educ.ucalgary.ca
Faculty of Education | dlst...@acs.ucalgary.ca
Damien P. Neil
In a previous article, dlst...@acs.ucalgary.ca (Douglas Stevens) says:
>Would it be better to buy a 486SX-25MHz computer or a
>386DX-40MHz, all other factors being equal?
Definately the 486SX. The fast 386 might be a bit faster, and have a math
coprocessor, but you will be able to upgrade the SX to a 486DX2/50 at a
later time. The only reasons to go with the 386 are if you need a math
coprocessor now, or if you plan to never upgrade the computer.
--
+-------------+------------------+---------------------------------+
| Damien Neil | dp...@po.cwru.edu | Case Western Reserve University |
+-------------+------------------+---------------------------------+
Roger Petersen
Neither the 386DX nor the 486SX have a built-in math coprocessor,
so they're equal in that respect.
I'd expect the 486SX to run roughly 2 times faster
than the 386DX at the *same* clock speed.
Disclaimer: Of course this is highly dependent on
your application. Maybe 1.5X to 2.5X, depending.
Cache is another important factor.
Roger Petersen
ro...@sr.hp.com
Allen J Michielsen
>Would it be better to buy a 486SX-25MHz computer or a
>386DX-40MHz, all other factors being equal?
Without a single solitary doubt, any INTEL 486 system is a better choice
than any 386 system, systems being equal otherwise. All Intel 486's have
a built in cache, which has a very good hit rate, while all 386 systems
which have a cache, have a external cache, which may be larger, may have a
higher hit rate, and is slower over all. Almost all Intel 486 systems
will have clock doubler, tripler, or quadrupler chips which will allow
the end user to (eventually) drop in a faster cpu, and dramatically
increase performance. Intel is shipping clock doubler chips, and has
announced it is developing quadrupler chips. I see tripler chips as a
possible interium step if problems arise.
All in all, the performance gains (differences) between memory, disk, &
video speed (performance) are a bigger factor today than the difference
between the 2 cpus.
al
Vivek Khindria (CSU)
You may want to double check your information about the 486sx.
I understand that it does have a math coprocessor, but has a smaller
databus.
Vivek Khindria
McMaster University, Hamilton, Ontario, Canada
khin...@mcmaster.ca
Scott Wilken
>
> You may want to double check your information about the 486sx.
> I understand that it does have a math coprocessor, but has a smaller
> databus.
>
>Vivek Khindria
BUZZZZZZ! Thanks for playin!. The 386sx does indeed have a smaller databus
than its DX counterpart, however this is *NOT* true of the 486 line.
The 486DX has a math coprocessor built-in onboard, the 486sx is identical to
the 486dx (same bus, has a cache, etc), but it has *NO* math coprocessor.
Scott
--
Go FAST! | Internet: wil...@plains.nodak.edu | AMA #587126
Take Chances! | UUCP: ..!uunet!plains!wilken | DoD #0087
VF700F Interceptor | Bitnet: WILKEN@PLAINS |
Patrick Taylor
> You may want to double check your information about the 486sx.
> I understand that it does have a math coprocessor, but has a smaller
> databus.
>Vivek Khindria
>McMaster University, Hamilton, Ontario, Canada
>khin...@mcmaster.ca
No, you need to double-check your information. Please don't post about
things that you aren't pretty damn sure of unless you make it clear that you
aren't very damn sure.
The 486 SX does *not* perform any math-coprocessor functions, and it has the
same data bus size as a 486DX, 32 bits wide.
The 386 SX is the one with a smaller (16 bit wide) data bus, but it too has
no coprocessor.
----------------------------------------------------------------------------
"This must be Thursday. I never could get the hang of Thursdays"
- Douglas Adams
- Patrick Taylor (Arthur Dent)
Ericsson Network Systems
exu...@exu.ericsson.se "Don't let the .se fool you"
or exu...@ZGNews.Lonestar.Org, exu.e...@memo.ericsson.se
Steve Russell
That may be true if you ignore the economics involved.
Unless you really need the horsepower of a 50+mhz 486, a 40mhz 386 with
math is an acceptable substitute costing several hundred dollars less,
which lets you put the money saved into a bigger disk or better video
card or monitor.
Norton index (yeh, I'm ignoring the internal/external cache question)
for a 386-40 is 48, the 486DX-33 is 51. Three points of cpu poop is
not justification for an extra 400 bucks or so.
Sure, if you really need the speed for pixel graphics or simulation, then
pop for the 486-50 or 66 models (makes them JPEG files stroll instead
of creap). Norton says 486-50 = 77, or better than 30% improvement over
486-50. Maybe worth it for the money, and the application.
I can't see any economic advantage for any 386/486 "SX" design.
SX is just a marketing ploy where you pay to begin with and maybe
pay again afterwards (for the new cpu - and then what do you do
with the old one?).
You buy hardware to fit the application. No point over buying when
obsolescence is literally around the next quarter.
-steve
Sean Chan
>Norton index (yeh, I'm ignoring the internal/external cache question)
^^^^^^^^^^^^
>for a 386-40 is 48, the 486DX-33 is 51. Three points of cpu poop is
>not justification for an extra 400 bucks or so.
>Sure, if you really need the speed for pixel graphics or simulation, then
>pop for the 486-50 or 66 models (makes them JPEG files stroll instead
>of creap). Norton says 486-50 = 77, or better than 30% improvement over
>486-50. Maybe worth it for the money, and the application.
Actually, from what I've been told, Norton SI should not be used to test
the speed of systems (anyone knows why??). Recently, I posted a question
to check the results of SI run on 386DX-40 and had a wide range of
resutls. I was getting responses from 43.1 -> 57.0 (my 486SX-25 came up
to a 54.0). And those values applied to 386DX-40s with 256K cache. So, I
don't really trust SI that much anymore. Oh yes, I suppose I didn't take
chipsets (of M/B) into account, but I don't think that it some chipsets
are THAT much faster than others.
In Western Australia, the cheapest 386DX-40 128K cache (UMC chipset) is
$300, and the UMC 486SX-25 0K extern. cache is $390 (add 50 for 64K
cache).
In my opinion, the 486SX-25 is a better choice as it is upgradable to a
the Overdrive/DX2 chip (and I'm getting a 2nd hand DX2-50 for Aus$250 :)
)...
Sean
***********************************************************************
* * *
* Sean Chan * Inside every microchip *
* Student Number : 892561F * there is a little man *
* 3rd Yr. Computer Science * running around doing *
* Curtin University of Technology * everything you have *
* * asked him to do... *
* E-Mail : ch...@marsh.cs.curtin.edu.au * *
* pch...@cc.curtin.edu.au ******************************
* *
******************************************
Alok Dhir
>rus...@ampex.com (Steve Russell) writes:
>
>>Norton index (yeh, I'm ignoring the internal/external cache question)
> ^^^^^^^^^^^^
>>for a 386-40 is 48, the 486DX-33 is 51. Three points of cpu poop is
>>not justification for an extra 400 bucks or so.
>
>>Sure, if you really need the speed for pixel graphics or simulation, then
>>pop for the 486-50 or 66 models (makes them JPEG files stroll instead
>>of creap). Norton says 486-50 = 77, or better than 30% improvement over
>>486-50. Maybe worth it for the money, and the application.
>
>Actually, from what I've been told, Norton SI should not be used to test
>the speed of systems (anyone knows why??). Recently, I posted a question
I don't know whether Norton should or should not be used, but your figures
are quite a bit shy of my experience. My 486/33 at work scores 78 on Norton
SI v5.0 and my 486/50 at home scores 108.1.
--
============================================================================
Al Dhir Technical Consulting Staff
Internet: ad...@cygnus.umd.edu University of Maryland, College Park
Bitnet: adhir%cygnus.umd.edu@Interbit (301) 405-1500 * (301) 405-3014
John Ackermann
>The 486DX has a math coprocessor built-in onboard, the 486sx is identical to
>the 486dx (same bus, has a cache, etc), but it has *NO* math coprocessor.
FWIW, yesterday (in the process of going through this same decision
process), I ran Landmark V2.0 on three systems. For basic CPU (not FPU)
performance, these are the results:
386/40 64k cache -- 61MHz
486SX/25 no external cache -- 83 MHz
NCR 486/25 external cache ? -- 83 MHz
So, based at least on this test, and looking at integer CPU only, the
486SX has the same performance as the 486DX.
John
--
John R. Ackermann, Jr. Law Department, NCR Corporation, Dayton, Ohio
(513) 445-2966 John.Ac...@daytonoh.ncr.com
Packet Radio: ag9v@n8acv tcp/ip: ag...@ag9v.ampr [44.70.12.34]
fred j mccall 575-3539
> You may want to double check your information about the 486sx.
> I understand that it does have a math coprocessor, but has a smaller
> databus.
You understand incorrectly. The difference between a 486SX and a
486DX is that the SX part has no math coprocessor and they use a
smaller feature size on the die since the redesign. The SX part used
to just be a DX part with the math unit deliberately disconnect, but
Intel reengineered the chip to remove it altogether.
--
"Insisting on perfect safety is for people who don't have the balls to live
in the real world." -- Mary Shafer, NASA Ames Dryden
------------------------------------------------------------------------------
Fred....@dseg.ti.com - I don't speak for others and they don't speak for me.
fred j mccall 575-3539
>In article <> dlst...@acs.ucalgary.ca (Douglas Stevens) writes:
>>Would it be better to buy a 486SX-25MHz computer or a
>>386DX-40MHz, all other factors being equal?
>Without a single solitary doubt, any INTEL 486 system is a better choice
>than any 386 system, systems being equal otherwise. All Intel 486's have
>a built in cache, which has a very good hit rate, while all 386 systems
>which have a cache, have a external cache, which may be larger, may have a
>higher hit rate, and is slower over all. Almost all Intel 486 systems
>will have clock doubler, tripler, or quadrupler chips which will allow
>the end user to (eventually) drop in a faster cpu, and dramatically
>increase performance. Intel is shipping clock doubler chips, and has
>announced it is developing quadrupler chips. I see tripler chips as a
>possible interium step if problems arise.
I have one single rule. Never buy ANYTHING that ends in SuX.
Particularly not the piece of marketing cruft known as the 80486SX.
Given that the only difference in price between a 486SX system and a
486DX system OUGHT to be the difference in CPU cost, I would pony up
the extra bucks to go with the DX parts.
If you can't do that, buy clone parts. Cyrix and AMD are both doing
some interesting things with regard to i486 clones that run in i386
motherboards (for example).
The advice above is of the same sort as "always buy IBM" or "always
buy Microsoft"; nice safe choices in going for name brands, but not
always the best price-performance ratio in the world.
>All in all, the performance gains (differences) between memory, disk, &
>video speed (performance) are a bigger factor today than the difference
>between the 2 cpus.
So why not recommend buying a cheap clone CPU with faster peripherals,
if that is where all the speed difference is?
91066720 j r selleck
BZZZZZZZ !!! thanks for playin!. Seeing as though we are not discussing on intel machines here i feel it is relevant to point out that the Cyrix 486 has a 16 bit BUS but inclu
des maths coprocessor..
and even the Intel 486sx has a mathscoprocessor, it also has a small laser hole in it where it has been disabled...
So for all our budding MICRO surgeons ot
out there you 2 can have a 486 dx from an sx if you can fix it !!
James
David Shepherd
: wil...@plains.NoDak.edu (Scott Wilken) writes:
: and even the Intel 486sx has a mathscoprocessor, it also has a small laser hole in it where it has been disabled...
:
: So for all our budding MICRO surgeons ot
: out there you 2 can have a 486 dx from an sx if you can fix it !!
may have been the case with early 486sx's but i belive intel have
re-engineered the 486sx chip to leave out the co-processor area ....
mass market chips get to a point where margins are very slim
(especially if there's competition :-) so having x% less die per wafer
due to an unused co-processor is bad economics.
--
--------------------------------------------------------------------------
david shepherd: d...@inmos.co.uk or d...@inmos.com tel: 0454-616616 x 625
inmos ltd, 1000 aztec west, almondsbury, bristol, bs12 4sq
1992: celebrate the quincentenary of columbus getting lost
luf...@cuphub.cup.edu
> wil...@plains.NoDak.edu (Scott Wilken) writes:
>
>>In article <1992Oct21.1...@maccs.dcss.mcmaster.ca> khin...@maccs.dcss.mcmaster.ca (Vivek Khindria (CSU)) writes:
>>>
>>> You may want to double check your information about the 486sx.
>>> I understand that it does have a math coprocessor, but has a smaller
>>> databus.
>>>
>>>Vivek Khindria
>
>>BUZZZZZZ! Thanks for playin!. The 386sx does indeed have a smaller databus
>>than its DX counterpart, however this is *NOT* true of the 486 line.
>
>>The 486DX has a math coprocessor built-in onboard, the 486sx is identical to
>>the 486dx (same bus, has a cache, etc), but it has *NO* math coprocessor.
>
>>Scott
>>--
>>Go FAST! | Internet: wil...@plains.nodak.edu | AMA #587126
>>Take Chances! | UUCP: ..!uunet!plains!wilken | DoD #0087
> BZZZZZZZ !!! thanks for playin!. Seeing as though we are not discussing on
> intel machines here i feel it is relevant to point out that the Cyrix 486 has
>
> and even the Intel 486sx has a mathscoprocessor, it also has a small laser
> hole in it where it has been disabled...
>
> So for all our budding MICRO surgeons ot
> out there you 2 can have a 486 dx from an sx if you can fix it !!
> James
BZZZZZZZZZ !!! thank YOU for playing, but most people do not consider the Cyrix
486 a true 486. First, it is pin compatable with a 386; second, just because a
chip has a math coprocessor, it dosn't mean it's a 486.
Third, all new 486sx's do not have the coprocessor onchip, but it is excluded
entirely.
Bryan Luff
Cal. U. of PA
fred j mccall 575-3539
>wil...@plains.NoDak.edu (Scott Wilken) writes:
Well, I wouldn't normally respond to this, since it is MOSTLY correct,
but I just hate this rude "BZZZT" crap. So --
BZZZZZZZT!!!! Thanks for playing. While your statements about the
486SX was originally (sort of) correct, this hasn't been the case in a
long, LONG time. The i486SX part has actually been remasked (with a
smaller feature size, which dissipates less power) and that mask does
NOT contain the math coprocessor circuitry.
>So for all our budding MICRO surgeons ot
>out there you 2 can have a 486 dx from an sx if you can fix it !!
>James
Back to medical school for you, dude.