second thoughts on buying a 3000UX

Tom Limoncelli limonce at pilot.njin.net
Thu Feb 28 19:13:51 AEST 1991


[Excuse the length of this.  It's late and my writing gets more
verbose as I get tired.  Though, it does reveal some nice things about
the 3000UX.]

In article <1135 at applix.com> scotte at applix.com (Scott Evernden) writes:

> In my opinion, if you have no need for the AmigaDOS stuff, then you've
> eliminated the only good reason to get the 3000UX.

A very common misconception!

The 3000UX makes a great Unix box on its own.  First of all, it's a
very complete Unix.  Nothing is left out (though you can disable
anything you want :-) ).  That alone makes it a step ahead of many
other systems.

[description of the X performance you want, among other things.]

> In the end, I decided to brew myself a 486/33 box.  It has up to now,
> satisified these requirements (and others) for X and unix development
> purposes.

You spend that much to get the performance you wanted?  Ugh!  You
really should look into a Amiga 3000UX.  You'll get the same
performance (no, not on CPU-bound tasks... but you were talking about
development and X windows; both are disk & video i/o intensive).
Here's why:


I was recently at a "normal" party (not too many computing
professionals there) and found myself participating in a discussion
about how SVR4 was slow and what could be done about it.

A person there explained what was done to speed SVR4 up, and what
things were added that resulted in slowing it down.  The conclusion
was that to get the best performance you HAD to have a really good I/O
system.  You *can* have a good I/O system on a 486 or 386 box, but not
if you use a machine designed for MS-DOS use.  In other words, 486/386
SVR4 Unix must be on a totally "designed-for-Unix" box.  [I believe
NCR's new line is like this.]

This tuning/performance information was according to the person at the
party.

Then I described the Amiga 3000UX's I/O system.  How memory was quite
fast and I/O was on this blindly fast system, DMA, 32-bits, etc. etc.

She said that such a system would beat a 486 on I/O so much so that
all the "speed problems" of SVR4 would diminish, and you could get the
performance you want.  In other words, the Amiga 3000UX would make for
one of the best possible SVR4 boxes.  SVR4 wants a machine with a
superior DMA system and fast RAM. (Most 386/486 boxes disable their
DMA because it doesn't work or isn't standard.)

Lack of a good DMA system and horridly slow video-ram on the 486/386
is a major problem for SVR4.  The system works the worst on the
machine with the biggest market!


This all seemed strange to me.  How could a person at some random
party know so much about this?  Well, I asked her for some
credentials.

It turned out she is on the performance and tuneing enhancement team
for SVR4.  The party happen to be at a house only 25 minutes from the
(new) home of Unix, Short Hills, New Jersey.

Strange thing about living in Morris County.  You never know who
you'll find yourself talking to at a party.

> -scott
> 
> Flames, anyone?

Hmmm... I'm not sure if this was a flame.  I guess I feel sad for you
because you spent so much money on a 486 machine; you paid for a lot of
CPU power to offset a bad I/O system.

-Tom
P.S.  The conclusion was that the best SVR4 system would have to have
superior DMA for all I/O, very fast video ram, (get this!) the Unix
file system (she said that her tests showed that a well-tuned UFS was
actually faster than a well-tuned Berkeley FFS!), a disk block size
and page size had to be equal, and a couple things that don't come to
mind right now.  Basically, in my mind, she unintentionally described
the Amiga 3000UX.



More information about the Comp.unix.amiga mailing list