🚀 go-pugleaf

In article <1giql5INNo5@hamblin.math.byu.edu> bill@mathnx.math.byu.edu (Bill  
Smith) writes:
> Well, I'll pass another rumor on from a friend.  While the
> 88110-flavor box has been dropped, another design based
> on it appears to be the version they will release.  It has
> a multiprocessing feature with plug-in sites for several
> cpu's, you just have another plugged in if you want it.  The
> alpha is not going to be the chip they use, it will also
> not be P5. No waiting for 68060 either.  Not R4000-etc.
> Not Sparc.  Not IBM.  Want to take a guess what's driving
> it? 

What does that leave?

HP PA-RISC is the obvious answer.  It's fast, cheap, has a good rep, and might  
pave the way for a NeXTSTEP port to Snakes.

I suppose something exotic like an AMD is conceivable, but much less likely.
Would you dare to speculate on *when*?

I still like the theory that the NRW is merely a method to find leaks.  They  
tell a story about the NRW to each suspected leaker, but change the CPU.  Thus,  
by seeing which CPU makes the rumour circuit, then know who the leak is! :-)

- Ernie P.

--
Ernest N. Prabhakar                  Caltech High Energy Physics
Member, League for Programming Freedom (league@prep.ai.mit.edu)
CaJUN President        NeXTMail:ernest@pundit.cithep.caltech.edu
"...and ourselves, your servants for Jesus sake." - II Cor 4:5b
#import <std/disclaimer.h>

In article <1gitslINNnat@gap.caltech.edu> ernest@pundit.cithep.caltech.edu (Ernest Prabhakar) writes:
>In article <1giql5INNo5@hamblin.math.byu.edu> bill@mathnx.math.byu.edu (Bill  
>Smith) writes:
>> Well, I'll pass another rumor on from a friend.  While the
>> 88110-flavor box has been dropped, another design based
>> on it appears to be the version they will release.  It has
>> a multiprocessing feature with plug-in sites for several
>> cpu's, you just have another plugged in if you want it.  The
>> alpha is not going to be the chip they use, it will also
>> not be P5. No waiting for 68060 either.  Not R4000-etc.
>> Not Sparc.  Not IBM.  Want to take a guess what's driving
>> it? 
>
>What does that leave?
>
>HP PA-RISC is the obvious answer.  It's fast, cheap, has a good rep, and might  
>pave the way for a NeXTSTEP port to Snakes.
>
>I suppose something exotic like an AMD is conceivable, but much less likely.
>Would you dare to speculate on *when*?
>

Not a rumor, not even speculation, not even a wish ... just a remote
possibillity ... since multi-processing is supposed to be one of the
great things about Mach, why not use the one processor that was
designed specifically _for_ multiprocessing ? I am talkingabout the
Inmos transputers -- the T9000, or whatever their latest design is
called ... I think even a machine with a couple or 4 of the (now a
couple of years old) T-800:s would be pretty sexy, and then NeXT could
boast about having not only the best UI, but also the only
multi-processor-box for home use ...

Besides, the transputers are not on the above list of processors they
didn't go with so ... maybe my wild shot actually hit a pheasant ? :)

>
>- Ernie P.
>
>--
>Ernest N. Prabhakar                  Caltech High Energy Physics
>Member, League for Programming Freedom (league@prep.ai.mit.edu)
>CaJUN President        NeXTMail:ernest@pundit.cithep.caltech.edu
>"...and ourselves, your servants for Jesus sake." - II Cor 4:5b
>#import <std/disclaimer.h>


-- 
/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/\/
// Christian Brunschen // Husmansv. 26 // S - 227 38 Lund // Sweden //
// voice# +46 (0)46 139345  // internet d89cb@efd.lth.se //        //
// irc snooker Senster gcb // ------------------------------------++

In article <1992Dec16.132918.8472@cyantic.com> mark@cyantic.com (Mark T. Dornfeld) writes:

   If PA is in the plans for NeXT, I hope either HP or NeXT builds a box
   specifically for NS or one without any HP proprietary gotchyas.  We have
   been working with the HP 9000 800 series for the last few weeks and HP has
   quite a bit to learn about putting a decent OS on what is really some 
   wonderful hardware.  Also, the 700's and 800's are not binary compatible
   due most probably to the multiprocessor architecture of the 800.
   -- 

   Mark T. Dornfeld                              Voice: (416) 234-9048
   CYANTIC Systems                               Facsimile: (416) 234-0477
   101 Subway Crescent Suite 2103                Email: mark@cyantic.com

Having administered a 9000/720 and a 9000/835 that sat on the same table,
I can tell you that the the CPUs _are_ binary compatible.  Indeed, one
can compile something on the new box (720) with an option to not use
one or two new instructions, and the executable will run on the old box
(835).  Also, everything we've compiled on the 835 runs on the 720.
The only exceptions to these observations are due to having version
8.07 of the OS on the 720 and version 7.0 on the 835 (where things like
shared libraries are not implemented).

Also, we have not had any out-of-the-ordinary problems with porting
stuff to either HP.  Many software packages include the appropriate
flags in the makefile for the HPs, and those that don't were typically
easy to fix.
--
------------------------------------------------------------------
Dr. Stephan V. Schell                   schell@gauss.ece.psu.edu
121 Electrical Engineering East
Dept. of Electrical and Computer Engineering
The Pennsylvania State University
University Park, PA 16802               (814) 863-8102

In article <1gitslINNnat@gap.caltech.edu> ernest@pundit.cithep.caltech.edu (Ernest Prabhakar) writes:
>In article <1giql5INNo5@hamblin.math.byu.edu> bill@mathnx.math.byu.edu (Bill  
>Smith) writes:
>> Well, I'll pass another rumor on from a friend.  While the
>> 88110-flavor box has been dropped, another design based
>> on it appears to be the version they will release.  It has
>> a multiprocessing feature with plug-in sites for several
>> cpu's, you just have another plugged in if you want it.  The
>> alpha is not going to be the chip they use, it will also
>> not be P5. No waiting for 68060 either.  Not R4000-etc.
>> Not Sparc.  Not IBM.  Want to take a guess what's driving
>> it? 
>
>What does that leave?
>
>HP PA-RISC is the obvious answer.  It's fast, cheap, has a good rep, and might  
>pave the way for a NeXTSTEP port to Snakes.

While the performance of the PA architecture is obviously desirable, I'd
like to check out the cost of complete systems from HP.  Many of the
advertised prices from HP are for diskless boxes with minimal RAM and the
low-end has very little internal expansion.

Does anyone have any list prices of HP systems that are fully configured?

If PA is in the plans for NeXT, I hope either HP or NeXT builds a box
specifically for NS or one without any HP proprietary gotchyas.  We have
been working with the HP 9000 800 series for the last few weeks and HP has
quite a bit to learn about putting a decent OS on what is really some 
wonderful hardware.  Also, the 700's and 800's are not binary compatible
due most probably to the multiprocessor architecture of the 800.
-- 

Mark T. Dornfeld                              Voice: (416) 234-9048
CYANTIC Systems                               Facsimile: (416) 234-0477
101 Subway Crescent Suite 2103                Email: mark@cyantic.com

In article <1giql5INNo5@hamblin.math.byu.edu> bill@mathnx.math.byu.edu (Bill  
Smith) writes:
> Well, I'll pass another rumor on from a friend.  While the
> 88110-flavor box has been dropped, another design based
> on it appears to be the version they will release.  It has
> a multiprocessing feature with plug-in sites for several
> cpu's, you just have another plugged in if you want it.  The
> alpha is not going to be the chip they use, it will also
> not be P5. No waiting for 68060 either.  Not R4000-etc.
> Not Sparc.  Not IBM.  Want to take a guess what's driving
> it?  

That leaves HP. Yippee!

In article <1992Dec16.132918.8472@cyantic.com> mark@cyantic.com (Mark T. Dornfeld) writes:
>                                                         Many of the
>advertised prices from HP are for diskless boxes with minimal RAM and the
>low-end has very little internal expansion.
>
The prices quoted for the new PA-7100 systems are fully-usable
color systems w/ decent RAM and disk storage.


Ken
#include <disclaimers>  My opinions do not reflect those of my employer
-- 
Kenneth K.F. Lui, klui@corp.hp.com                  3000 Hanover Street MS20BJ
Corporate Administrative Information Systems        Palo Alto, CA  94304   USA
Core Application Technologies                1(415)857-3230 Fax 1(415)852-8026

I don't think the Mach architecture can support multiple threading across
seperate CPUs.  I may be wrong but I don't think this "drop in CPU boards"
stuff is going to work without massive OS and architecture restructing.

IMHO

- Hacker
-- 
Dark Hacker @ Black Silicon, Fortress Of Computation
hacker@black-silicon.mclean.va.us

"Life itself is... COMPUTATION!"

The news from Dr. Smith is that NeXT won't be using...
... Sparc
... 88k
... etc.

Everyone seems to be assuming that this means NeXT will go with
the PowerPC. I don't think so.

It would jibe with recent history if NeXT dialed R4000 to dish up
some great performance - with a little help from Silicon Graphics
coming along with the chips.

I recall that NeXT and SGI have had high level meetings in the last
few months, or so the net rumours say. Also, SGI does very well at
the high end of the workstation market, but can't really compete with
SUN.... With a multiprocessor NeXT using MIPS chips at the low end
of the market, and NeXT recommending SGI machines for the high end,
both companies win big time.

I know there have been flame wars in the past over which machine
(Indigo etc. vrs NeXT) has better development tools. Perhaps we'll
soon see machines with the best of both worlds.

Just my misguided, but optimistic opinion.

Rob Harrap
Dept. Geological Sciences
Queen's University
Kingston, Ont.

Harrapr@qucdn.queensu.ca