Unix close for 486 - commens requested

Unix close for 486 - commens requested

Post by Curtis Hrisch » Sat, 07 Aug 1993 00:20:12



Hi.  I am looking for comments about using 80486 PCs as Unix boxes.
Before the flame throwers come out, here is the reason:  I am in an
acedemic environment that runs a large amount of simulations.  Rather
than have a $40K workstation run 10 simulations, it would be faster to
run the simulations on 10 80486 PCs (and more reliable).

To achieve this, there are some functions that are necessary:
- Transparent NFS.  Should be able to access files on the PC just like
any other NFS connected workstation.
- Use gcc or g++ compiler, and other GNUish tools.  If it came with
its own compiler even better.
- Remote login capabilities, so that users could remotely login and
start batch simulation jobs.

To be perfectly honest, I don't know what this means in terms of
kernel capabilities:  Is TCP/IP networking necessary? ...... If you
could provide some comments about this, please do.

This brings us to the question of hardware - what are typical
requirements for Unix (clones+) for 80486 platforms?  Your experience
is invaluable.

I will summarize the comments and repost.

Thank you for your time.

Curtis
--
----------------------------------------------------------------------------
| Curtis Hrischuk          | Boundaries of Science:                        |

| Carleton University      | 2) Why am I here?                             |
| Ottawa, Canada, K1S-5B6  | 3) What happens when I leave?                 |
| Ph  (613) 788-2600 x1762 | "The proof is almost identical to the previous|
| FAX (613) 788-5727       | proof and hence omitted here."(actual quote)  |

 
 
 

Unix close for 486 - commens requested

Post by Nate Willia » Sat, 07 Aug 1993 00:23:01




>Hi.  I am looking for comments about using 80486 PCs as Unix boxes.
>Before the flame throwers come out, here is the reason:  I am in an
>acedemic environment that runs a large amount of simulations.  Rather
>than have a $40K workstation run 10 simulations, it would be faster to
>run the simulations on 10 80486 PCs (and more reliable).

                                          ^^^^^^^^^^^^^

In my opinion, you should consider getting BSDI's commercial
BSD *nix.  For $2K you get one-year unlimited licenses + support,
and you can't beat that with ANY other PC unix product that I'm aware
of.    There are free *nices around as well, but if you desire stability
and you require it to work right, then you really need to look at
something that is supported commercially.

Some of the other free *nix folks might disagree, but you can't touch
BSDI's support staff for knowlege and for having a rock-solid product.

Quote:>To achieve this, there are some functions that are necessary:
>- Transparent NFS.  Should be able to access files on the PC just like
>any other NFS connected workstation.
>- Use gcc or g++ compiler, and other GNUish tools.  If it came with
>its own compiler even better.
>- Remote login capabilities, so that users could remotely login and
>start batch simulation jobs.

All of this and more is available from BSDI, and the networking code is
very stable and VERY usable.  (At times more so than some of the
workstations on campus)  Note, I have not run BSDI, but I run one of the
freely available *nices based on the same code, and although my box is
not as stable as far as uptimes as BSDI boxes, my box is not crashing due
to any network problems (knock-on-wood.)

Quote:>This brings us to the question of hardware - what are typical
>requirements for Unix (clones+) for 80486 platforms?  Your experience
>is invaluable.

We have seen very acceptable (!) performance out of 486/66 EISA boxes
with 32MB of memory + SCSI systems, and also out of 486/66 EISA/VLB +
SCSI systems.  

The reason for EISA is that it is a better bus for doing higher I/O and
video, and it allows you to have free access to all your memory, whereis
the ISA machines are limited to DMA up to 16MB, or you have to implement
a slower bounce-buffer system to get to the memory above 16MB.

Get a good EISA SCSI controller (adaptec 1742 works well), and you can
put multiple disks, tape drives, cd-players, etc... on the machie plus
get much better performance using a multi-tasking, multi-user system.

I have not seen an IDE system come close to the performance of a EISA
SCSI box.

Note, you can run *nix on 386SX/16 + 4MB, but it isn't recommended.
(Right Jaye!)

If you want to use the system with X11R5, I suggest getting an
accelerated graphics card with an S3 chipset, which is supported by
BSDI.  A good  monitor is a must if you're doing X, so get one of the PC
rags and find out what they think are good monitors.  I have a Nanoa
550i, and I absolutely love it.

--


work #: (406) 994-4836       |  Operating Systems for [34]86 machines.
home #: (406) 586-0579       |  (based on Net/2, name changes all the time :-)

 
 
 

Unix close for 486 - commens requested

Post by Jaye Mathis » Sat, 07 Aug 1993 01:50:41




>Note, you can run *nix on 386SX/16 + 4MB, but it isn't recommended.
>(Right Jaye!)

Damn straight, and it works...
--
 Jaye Mathisen, COE Systems Manager                (406) 994-4780
 410 Roberts Hall,Dept. of Computer Science

 
 
 

Unix close for 486 - commens requested

Post by Jaye Mathis » Sat, 07 Aug 1993 05:44:24





>My disagreements with this article are legion.


>to BSDI's product is ~$1K NOT $2K

Bzzt.  License for *1* machine is 1k.  University license w/source,
and unlimited binary run-time license, with a single source of support
is 2k.  the $1k only lets you run it on 1 machine, additional RTL's are
something like $200 a pop, which would put him over the 2k figure
anyway.  (these numbers from the last price sheet ftp'd from bsdi.com).

Quote:

>2.) You don't need accelerated video, 32M RAM, EISA, etc etc ad nauseum
>    for what you've asked for. It's nice to have more and faster, but a
>    lot less will produce very nicely for you. Especialy I have to wonder
>    about the display recomendation given the stated application is for
>    BATCH simulations.

Well, there's a bit of a disagreement here.  If the machines are only
batch, and there's no possiblity that they won't/can't be used for
X terminals or such, then you're right.  But *if* you're going to buy
all those machines, and you have the opportunity to use them for computse
servers, and X, then accelerated video is the way to go.

As to EISA vs ISA, it depends.  If your simulation is so big that
you need more memory, then EISA is a better choice if you opt to run one
of the free unices, because of the 24bit addressing problem with the
Adaptec in an ISA box.  The problem doesn't exist in the EISA box with
a EISA controller.   EISA boxes aren't that much more than ISA now days
anyway.  I don't recall offhand if BSDI implements bouncebuffers for the
154x series to support more than 16MB's of RAM on the ISA.

If the stuff is I/O intensive and reads and writes a lot of data, there's
no comparison between ISA and EISA, the EISA box blows it away.

If this guy gets stuck with 1 "server" machine to store the disk, and
minimal hardware on the other 9, then making the server EISA is a wise
choice.

I've done some fairly extensive testing with the free unices (not Linux
however), and BSDI, and there is no comparison between IDE and good SCSI,
the SCSI blows it away.

Anyway, my 2 bits.
--
 Jaye Mathisen, COE Systems Manager                (406) 994-4780
 410 Roberts Hall,Dept. of Computer Science

 
 
 

Unix close for 486 - commens requested

Post by Jon Gefae » Sat, 07 Aug 1993 02:51:13


My disagreements with this article are legion.


to BSDI's product is ~$1K NOT $2K

2.) You don't need accelerated video, 32M RAM, EISA, etc etc ad nauseum
    for what you've asked for. It's nice to have more and faster, but a
    lot less will produce very nicely for you. Especialy I have to wonder
    about the display recomendation given the stated application is for
    BATCH simulations.

3.) Other than mispelling Nanao :) (I agree, that 550i is swell) I think
    this fellow has well stated what is higher end equipment and why.
--
 ______
 \ \  / Jon Gefaell, Computer Systems Engineer      | Amateur Radio - KD4CQY
  \/\/  Information Technology and Communications   | -Will chmod for food-

Any opinions expressed herein are not intended to be construed as those of UVA

 
 
 

Unix close for 486 - commens requested

Post by Nate Willia » Sat, 07 Aug 1993 06:14:12





>My disagreements with this article are legion.


>to BSDI's product is ~$1K NOT $2K

$2K is the University price, and since he was a .edu site and wanted 10
copies, $1K + ( ( * $200/binary) > $2K University price.

Quote:>2.) You don't need accelerated video, 32M RAM, EISA, etc etc ad nauseum
>    for what you've asked for. It's nice to have more and faster, but a
>    lot less will produce very nicely for you. Especialy I have to wonder
>    about the display recomendation given the stated application is for
>    BATCH simulations.

Regarding accelerated video, as I stated, IF you decide to do X work, you'll
want accelerated video.  If they sit in a machine room, Hercules or MDA does
quite nicely. :-)

As far as EISA and 32M memory, if these machine are indeed going to be used
for serious computational work, then the more memory you have the better.
And if you want more than 16MB of memory then EISA is really the only choice
you have on a PC.  (And you get the added advantage of faster I/O, video,
etc. if you need/want it).

32M is a nice # for any CISC based workstation.  I run 16MB on my ISA
box running a free variant of BSD *nix, and I've run my machine up to
48MB of processes w/out trying very hard.  Can you say swap yourself
silly?

If you've got the $$, spend them wisely.  In my opinion, you'd be much
better off buying 10 high end 486's than 15-20 lower end 386/486's.

(That is assuming you want to buy PC's)

Nate

--


work #: (406) 994-4836       |  Operating Systems for [34]86 machines.
home #: (406) 586-0579       |  (based on Net/2, name changes all the time :-)

 
 
 

Unix close for 486 - commens requested

Post by Jon Gefae » Sat, 07 Aug 1993 07:25:24


All very good points. :) I love synthesis...
--
 ______
 \ \  / Jon Gefaell, Computer Systems Engineer      | Amateur Radio - KD4CQY
  \/\/  Information Technology and Communications   | -Will chmod for food-

Any opinions expressed herein are not intended to be construed as those of UVA
 
 
 

Unix close for 486 - commens requested

Post by Rui Pedro Mendes Salguei » Sat, 07 Aug 1993 07:31:21



: > For $2K you get one-year unlimited licenses + support,

: My disagreements with this article are legion.
:

: to BSDI's product is ~$1K NOT $2K

You are talking about one source license and Nate Williams is talking
about a SITE license. One site license is $2K/year.

BTW, since we are talking money, for 10 machines you could buy one
source license + 9 aditional binary licenses ($200 each ?) = $2800.
This includes support for 60 days.
So you see that in the first year it's cheaper to pay a site license.

: 2.) You don't need accelerated video, 32M RAM, EISA, etc etc ad nauseum
:     for what you've asked for.

I agree. I'm using BSDI with 16Mb in a ISA machine with a ET4000 video.
But the display is a little slow (mostly text scrolling).
Accelerated video may be worth it if you want to use X.

--
 Rui Salgueiro |   Dpt. de Matematica    |"In my life / Why do I smile


 
 
 

Unix close for 486 - commens requested

Post by Jeff Kell » Sat, 07 Aug 1993 11:07:17




 > I don't recall offhand if BSDI implements bouncebuffers for the 154x
 > series to support more than 16MB's of RAM on the ISA.

BSDI's BSD/386 _does_ support more than 16MB on ISA bus machines with
Adaptec 154x series SCSI controllers.  FYI...

                -jeff

Jeff Kellem

 
 
 

Unix close for 486 - commens requested

Post by IE.. » Sun, 08 Aug 1993 01:15:20




Quote:

>Hi.  I am looking for comments about using 80486 PCs as Unix boxes.

The June 15, 1993 issue of PC Magazine evaluates the following Unixes for
Intel:

   Consensys V 4.2
   Dell Unix System V Release 4
   Interactive Unix
   SCO Open Desktop
   UnixWare
   NeXTStep for Intell Processors
   Solaris for x86

Check it out...


 
 
 

Unix close for 486 - commens requested

Post by Superus » Sat, 07 Aug 1993 21:23:25


: Hi.  I am looking for comments about using 80486 PCs as Unix boxes.
: Before the flame throwers come out, here is the reason:  I am in an
: acedemic environment that runs a large amount of simulations.  Rather
: than have a $40K workstation run 10 simulations, it would be faster to
: run the simulations on 10 80486 PCs (and more reliable).
:
: To achieve this, there are some functions that are necessary:
: - Transparent NFS.  Should be able to access files on the PC just like
: any other NFS connected workstation.
: - Use gcc or g++ compiler, and other GNUish tools.  If it came with
: its own compiler even better.
: - Remote login capabilities, so that users could remotely login and
: start batch simulation jobs.
:

seems like linux would fit the bill extremely well.
it is also free.

it seems to satisfay all of the requirements you listed in this post....


 
 
 

Unix close for 486 - commens requested

Post by Stephen Youn » Sun, 08 Aug 1993 21:07:24





>>Hi.  I am looking for comments about using 80486 PCs as Unix boxes.

>The June 15, 1993 issue of PC Magazine evaluates the following Unixes for
>Intel:

>   Consensys V 4.2
>   Dell Unix System V Release 4
>   Interactive Unix
>   SCO Open Desktop
>   UnixWare
>   NeXTStep for Intell Processors
>   Solaris for x86

>Check it out...



Better yet, pick up a copy of the FAQ on the same subject. It discusses
all these and more in great detail.

  ftp.uu.net:/usenet/comp.unix.sysv386/PC-clone_UNIX_Software_Buyer_s_Guide.Z

Is one place to get it. I'm sure there are many others.

Steve
--
Stephen M. Youndt           | geek - n. [Perh. < dial. geek, fool < MLG geck.]

UUCP:  uunet!hacker!steve   |        consists of biting the head off a live
(703)978-6352               |        chicken or snake.

 
 
 

Unix close for 486 - commens requested

Post by Jim Baches » Sun, 08 Aug 1993 09:11:00


I recommend getting 386BSD from agate.berkeley.edu or NetBSD from
sun-lamp.cs.berkeley.edu. These are free and complete BSD systems.
Its real UNIX. If you also get XFree86 you'll have some reasonable
UNIX workstations at a university buget price. They also have
complete networking and support for a number of ethernet cards.

Hope this helps.

Thanks

Jim Bachesta

 
 
 

Unix close for 486 - commens requested

Post by Greg Leh » Mon, 09 Aug 1993 18:06:57





>>2.) You don't need accelerated video, 32M RAM, EISA, etc etc ad nauseum
>>    for what you've asked for. It's nice to have more and faster, but a
>>    lot less will produce very nicely for you. Especialy I have to wonder
>>    about the display recomendation given the stated application is for
>>    BATCH simulations.

>Well, there's a bit of a disagreement here.  If the machines are only
>batch, and there's no possiblity that they won't/can't be used for
>X terminals or such, then you're right.  But *if* you're going to buy
>all those machines, and you have the opportunity to use them for computse
>servers, and X, then accelerated video is the way to go.

>As to EISA vs ISA, it depends.  If your simulation is so big that
>you need more memory, then EISA is a better choice if you opt to run one
>of the free unices, because of the 24bit addressing problem with the
>Adaptec in an ISA box.  The problem doesn't exist in the EISA box with
>a EISA controller.   EISA boxes aren't that much more than ISA now days
>anyway.  I don't recall offhand if BSDI implements bouncebuffers for the
>154x series to support more than 16MB's of RAM on the ISA.

The 16 MB limit has been discussed elsewhere - to the best of my
knowledge,

I'd just *love* to see some real numbers here. There's been a lot of
hype about the performance improvements that EISA and Local Bus
(particularly VESA) bring when compared to ISA, but I have never seen
any numbers, and nobody has correlated these claims with the chip set
in use.

I have recently completed a reasonably comprehensive test of
accelerated video boards under UNIX, and have found:

1. The price increment for a reasonably fast accelerated board (say
   the STB X-24, which runs an S3 801 and is about 15 times as fast as
   an ET4000-based board like the Diamond SpeedStar) is in the order
   of $100. Add the cost of a server (about $100 - $200 ) if your UNIX
   doesn't support accelerated boards (most System V don't, BSDI
   does).

2. With accelerated boards, the performance improvement through using
   EISA or Local Bus instead of ISA is hardly measurable.

3. The difference in motherboard chip set performance can more than
   offset the performance improvement of an EISA or Local Bus board.
   In my particular test, I compared S3 928 and CL5426 chipsets (like
   Elsa Winner 1000/#9 GXE and Genoa 8500 respectively) running under
   ISA, EISA and VESA local bus. The VL bus results were (slightly)
   *worse* than the ISA results. Running the test with the ISA board
   in the VL bus motherboard, I got results which were worse than in
   the vanilla ISA motherboard: obviously there is something wrong
   with the VL bus board. But nobody talks about relative motherboard
   performance, just these buzzwords EISA, ISA and VL Bus. I'll get
   round to more details later (maybe), but here are some orders of
   magnitude, measured on a 486DX/2-66 with 16 MB of memory and
   running SVR4.2:

 board    bits/pixel   line      fill       blt       text       arc       cmplx    xstones
 Elsa Winner 1000 (S3 928, 2 MB):
 EISA          4      312938    142525    115081     307656    2251175    195098    193895
 ISA           4      309830    134609    115879     292875    1935545    146209    184360
 VL Bus        4      311791    136694    122223     289437    1745757    120980    183283
 ISA/VL board  4      295064    135495    120990     284281    1785385    106359    178251

 STB X-24 (S3 801, 1 MB)
 ISA           4      195786     89979     78486     206937    1476028    123790    126570

 ATI Ultra Pro (Mach 32, 2 MB)
 ISA           8      339402     58934     49199     183562    4786756    116078    100635

 Genoa 8500 (CL5426, 1 MB)
 VL Bus        8      149458     28331     24077     177375    1983207     53398     53053
 ISA           8      120519     28505     23929     207625    1801266     53856     52676

 Diamond SpeedStar (ET4000, 1 MB)
 ISA           8       41113      5113      2663      68062     547235      5882      7823

   In each case, I have chosen the pixel depth (4 or 8 bits/16 or 256
   colours) which gave the best performance for the board).

4. Compared to motherboard performance, server performance is much
   more significant. There's been a reasonable amount of flaming
   recently about the relative performance of Metro Link and PPC. I
   haven't tested these servers yet, but I have tested the SGCS
   server. I didn't quite get their claimed performance (missed it by
   about 5% :-), but the results I did get were higher than Metro Link
   or PPC claim.

Quote:>If the stuff is I/O intensive and reads and writes a lot of data, there's
>no comparison between ISA and EISA, the EISA box blows it away.

>I've done some fairly extensive testing with the free unices (not Linux
>however), and BSDI, and there is no comparison between IDE and good SCSI,
>the SCSI blows it away.

How about publishing your results? Please keep personal mail down to
reasonable proportions.

Greg
--
Greg Lehey                       | Tel: +49-6637-1488              
LEMIS                            | Fax: +49-6637-1489
Schellnhausen 2, 36325 Feldatal, Germany

 
 
 

Unix close for 486 - commens requested

Post by Wm E. Davidsen » Tue, 10 Aug 1993 11:27:00


| The June 15, 1993 issue of PC Magazine evaluates the following Unixes for
| Intel:

  Eric Raymond's list of hardware posted to news.answers is also valuable.
--

    TMR Associates, +1 518-370-5654
    C programming, training, data gathering, porting to open systems,
    heterogeneous environments, computer controlled housing, custom software