On Fri, 22 Jan 1999 20:23:05 +1000, Lance Birch 
	<zon### [at] satcomnetau> wrote:
>I think that you are the one speaking rediculously...
>
>An obscene amount of memory?  You've got to be joking?  Besides, NT is for
>those that need and use large amounts of memory.  It's also far more stable
>with large amounts of memory.

I agree.  NT is for those who need and use large amounts of memory.  But 
how many desktop users with 32M of RAM are going to "upgrade" to win2k
because they think it's the next logical step?  Do you think MS is going 
to tell them that's a bad idea?

>As do I.  I'd much prefer NT over 98.  You say NT takes longer to boot?  The
>only response I've got for that is to ask you a question...  Would you
>rather boot once a day and have it take a little longer than have to reboot
>10 times a day?  In my personal experiance, the former is better.  I'm sick
>to death of getting illegal operations!!!

If I only had to boot once a day, that'd be nice.  Since I'm doing real work,
however, developing device drivers and applications, I have to boot NT far
more often.  Real operating systems don't require you to reboot to change 
your IP address, or to install or remove a video driver, or to change the 
size of your swap file.

>Hmm, that's interesting.  From my benchmarks, NT renders 5 times faster than
>98 in 3D Studio MAX R2.5.  Now if that isn't a substantial increase I don't
>know what is.  Obviously POV-Ray is an exception to this as it wasn't
>specifically built for NT.

Obviously raytracing is a completely different task than 3DSMAX and has
different priorities.  I thought we covered this.

>For the first:  Why not?  The rendering station at my school seems to cope
>well with it.  It is also only a P233 with 64Mb of RAM.  

Then it's four times the machine most consumers currently running 98 have.
The average machine in the field is probably no better than a P166, and
I'm guessing most users have 32M or less RAM.

>It runs a
>Perception DSP drive and a video output subsystem.  I've NEVER had a crash
>or the slightest problem.  A DSP drive won't even run under 98 or 95 because
>of it's useless subsystem, even if the processor is much more powerful and
>you have more RAM.

Then that is an application for which NT is suited.  But it's hardly an
average desktop machine.

>And for the second statement:  What else are you planning on using?  Some
>out-of-date operating system that is totally incompatible with all major
>network standards... 

There is only one major network standard, and there's only one OS
that was designed from the ground up to support it.  The standard is
TCP/IP and the OS is Unix.

>I don't think so.  At work we run NT perfectly well.
>It acts as a server for a network and also as an internet pipeline and proxy
>server.  It has never crashed in the time I've been there and I don't
>believe it ever will.  It usually has to cope with being online for several
>days, even weeks at a time without restarting.  

Most Unix servers stay online for months or years without restarting.  Before
MS came along, rebooting or even shutting down a machine was virtually unheard
of.

>If Windows 98/95 DID support multi-processors I'd suppose that they could be
>viable for rendering, however the lack of memory support and fixed system
>resource setting is a pain and causes high-end apps like 3D Studio MAX to
>crash frequently.  

Ah.  Perhaps you're thinking of that dialog box that popped up on our NT
server a week or two ago, telling us that our swap file was too small
and telling us it would be happy to resize it for us if we didn't mind
rebooting?  This is dynamic memory allocation?  I'm not saying Unix does
it any better, mind, but I think it at least allows you to add a swap 
partition if you have the extra space without having to reboot.

>One of the biggest problems of running MAX on 95/98 is
>fixed resources.  Not many people understand the way 95/98 handles system
>resources.  I have to make this clear, memory has nothing to do with
>resources in 95/98.  The resources are ultimately fixed.  NT doesn't have
>this problem, it has infinite resource allocation, something MAX needs to
>run correctly.  

Pardon me, did you miss the part where I said I write software for NT and
Windows 9x?  I know how resource allocation works, in excruciating detail.
Because of what I do, I have had to reverse-engineer the resource allocation
scheme by disassembling the kernel to find out things MS didn't want me to 
know (and they told me so, in person).  It is not infinite, and it is still 
shared between processes.  The result is the same: one resource-hogging app 
can kill every other app on the system, no matter what they tell you.  
Granted, the pool of available resources is a bit larger, but it is still a 
fixed-size pool.

Post a reply to this message