Path: utzoo!utgpu!jarvis.csri.toronto.edu!cs.utexas.edu!yale!mintaka!bloom-beacon!bu.edu!mirror!rob
From: rob@mirror.tmc.com (Rob Limbert)
Newsgroups: comp.sys.ibm.pc
Subject: Re: Brain-dead 286 - summary
Message-ID: <37554@mirror.tmc.com>
Date: 12 Mar 90 16:32:33 GMT
References: <8681@rosevax.Rosemount.COM> <206900171@prism> <29449@amdcad.AMD.COM>
Reply-To: rob@prism.TMC.COM (Rob Limbert)
Lines: 104

In article <29449@amdcad.AMD.COM> phil@pepsi.AMD.COM (Phil Ngai) writes:
>In article <206900171@prism> rob@prism.TMC.COM writes:
>|  Also the various DOS multitaskers like Windows 386, VM/386, and so on.

>DESQview 286 on a machine with EEMS seems to work about as well as
>DV 386. (I've tried both) The virtual 8086 mode may in principle make
>multitasking real mode processes easy but in practice it doesn't seem
>to make a whole lot of difference in products that have been delivered
>at this time. Neither Windows 386 or VM/386 are without their
>own problems, I hear.

   I've used both versions of DESQview, and DV 386, while not perfect,
is distinctly better than DV 286 (at least for my purposes). I've used
Windows 386 and VM/386 with good results. Most real-mode multitaskers are,
in my opinion, so unstable as to be of little use. V86 multitaskers offer
enough extra robustness and capability to make them worthwhile. 

   V86 is no more a guarentee of stable multitasking than regular protected 
mode. Each theoretically permits completely safe multitasking, but it's 
possible, usually due to an OS bug, to crash a machine under protected or 
V86 mode. It's just much more difficult for an application level program to 
do it, by accident or on purpose.

>|>Windows, on the other hand, has a lot of software. And when 3.0
>|>comes out, it will run in protected mode, allowing 16 megabytes of
>|>memory, ON A 286.
>|
>|   It will still run better (offering multitasking of DOS applications
>|and better virtual memory support) on a 386.

>Have you seen this or are you assuming it? None of the "previews" of
>Windows 3.0 that I've seen have mentioned this.

   Windows 3.0 isn't an officially announced product yet, so discussing it 
isn't entirely appropriate. But I stand by the statement, which I wouldn't 
if it were just speculation.

>|The problem is that 286 protected mode is far 
>|less versatile and powerful than 386 protected mode. That's why there's 

>If you mean it doesn't have a large flat address space, that's true.
>But for multi-tasking, the 286 has plenty of potential for running
>protected mode programs.

   There are numerous other advantages to 386 protected mode (paging, I/O
permission mapping, the ability to mix real and protected mode tasks, 32-bit 
operations, etc). 

   I'll grant, though, that the 286 has more multitasking potential than many
people realize. This is a key point with regard to the original 'brain-dead 
286' question. By many measures (and considering its age), it's a capable 
chip, with decent memory management and good performance. But in a world
where DOS compatibility is everything, a world which the 286 designers
understandably didn't foresee, it has real problems unless you're content 
to use it in real mode. And the 64K segment limit has gone from being a
nuisance to a severe problem, as memory prices drop and programs grow in
size and complexity.

>|I agree it's unfortunate 
>|that the 286 hasn't been used more fully, but at this point, that's not 
>|very relevant.

>It's quite relevant if you are interested in what software could
>be written for the 286. The fact that it may be painful is made
>up for by the incredible market that exists.

   I'm not saying that software developers should ignore the 286; given the 
size of the 286 market, that'd be unwise. I do maintain that, at this point 
in time, buying a 286 machine is an increasingly questionable decision unless
you're not planning on keeping it for long, or unless you're certain that you
won't want to run 386 software - which there's a growing volume of - in the 
foreseeable future.

>|And the extra cost of a 386, especially
>|if you're willing to buy a non-name brand, is pretty marginal.

>That's becoming less and less true as the shortage of 386s gets
>worse. And the 386 market is a true monopoly.

   You may be right; only time will tell. This could lead down a completely 
different rathole about second-sourcing. But, as machines are shipped with 
more and more memory, costlier video adapters, bigger hard disks, etc., 
the CPU itself becomes a progressively smaller factor in the overall system 
cost. Based on a recent look through PC Magazine, the cost difference
between comparable 286 and 386 systems is as little as $150. 

>|Given the narrowing cost difference between 286 and 386 machines, the group 
>|of tasks for which the 286 is the CPU of choice is shrinking, and will
>|continue to do so. 

>On the contrary, I think that with software like Windows 3.0 coming
>out, the type of tasks which the 286 does well is going to increase
>dramatically.

   True, but it will also dramatically increase the type of tasks which
the 386 does better. OS/2 version 2.0 will do the same thing. 

   Maybe underneath it all we're both bemoaning the same problem - the large 
time lag between hardware and software. In the Intel world, that lag is 
several years, and it's condemned the 286, 386, and 486 to be run mostly as 
fast 8088s. All these chips, including the 286, have lots of untapped 
potential. My argument is that the 386 has much more than the 286 (and that 
potential has already been more fully realized), and, given the relatively 
small price difference, is a wiser purchase for most purposes.