Xref: utzoo comp.sys.amiga.tech:13171 comp.sys.amiga.hardware:2424
Path: utzoo!utgpu!news-server.csri.toronto.edu!cs.utexas.edu!usc!sdd.hp.com!zaphod.mps.ohio-state.edu!uwm.edu!psuvax1!wuarchive!kuhub.cc.ukans.edu!markv
From: markv@kuhub.cc.ukans.edu
Newsgroups: comp.sys.amiga.tech,comp.sys.amiga.hardware
Subject: Re: GVP 040 board
Message-ID: <24864.26931f97@kuhub.cc.ukans.edu>
Date: 5 Jul 90 16:08:06 GMT
References: <2836@zipeecs.umich.edu> <4698@munnari.oz.au>
Distribution: na
Organization: University of Kansas Academic Computing Services
Lines: 66

>> Does anybody know if GVP is working on an 040 board for the 2000?
>   Even here in Australia we are told GVP is working on a 68040 board.
>   Truth is, the only people that know for sure are Great Valley Products.

I'd like to see a 040 board for the 3000s CPU slot (it has some of the
fancy signals the available that the 040 uses for cache control).
 
>> What things will they have?
>   If you are familiar with 68040 architecture, it is a turbo charged 68030.
>   The 68040 runs somewhat faster than the 68030, it has much larger

That is like saying a Formula I racer runs somewhat faster than a VW
bug. At 25MHz the 68040 does 20 MIPs (which puts it in the ballpark
with a Sparc processor, but obviously the 040s CISC MIPs do a whole
lot more than the Sparc's RISC MIPs).  The 030 I belive does 7-8 MIPs
at 25 MHz, which puts the 040 being about 3 times faster.

>   data and instruction caches, both running in at around 4K, and it also

The data and instruction sections of the chip have completely seperate
pipelines, MMUs, etc (the FPU has some of its own stuff too), giving
rather incredible performance.  Effective throughput of the 040 is
supposed to be around 1.2 cycles per instruction.

>   has a fair few of the 68881 maths co-processor instructions to reduce
>   the familiary 'cpXXX' instructions.

Actually it has all the arithmatic operations and does them 5 times
faster than a 882 (3.5 MFLOPs at 25MHz -vs- .5 MFLOPs for an '882).
Given that the 882 is about twice as fast as an '881, the 040 is a
screamer.  Even though the transcendentals have to be done in
software, given the 040s much faster arithmatic Motorola claims that
the 040 should still be faster than an 882 on even the
transcendentals.  Of course the OS (are you listening Commodore ?)
needs to support the traps.
 
>   As for external caches, that remains to be seen. They could be emulated
>   in software, but not even the fastest RAM chips could approach the speed
>   of the internal cache.

No, but a large cache properly implemented could be tons faster than
normal memory at about 95% of the performance at much less memory
(which is why they are so popular, a 64K cache of 0 ws memory with a
multimeg system is a lot cheaper than making all multi megs of RAM 0
ws).

>> -UNIX
> 
>   As far as I know, C= is currently arranging licensing agreements with
>   two large UNIX distributors and developers. I am lead to believe they
>   are Sun Corp and BSD (Berkeley). However, UNIX on the Amiga will be
>   * EXPENSIVE * and will require a minimum of 2 megs of RAM, and 
>   preferably 4 to even start resembling something on a Sun or the like.

Commodores Unix will be System V, Rel 4 with OpenLook which means it
is liscensed from AT&T and Sun.
 
>     krooglik@moroka4.ecr.mu.oz.OZ.AU
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
Mark Gooderum			Only...		\    Good Cheer !!!
Academic Computing Services	       ///	  \___________________________
University of Kansas		     ///  /|         __    _
Bix:	  markgood	      \\\  ///  /__| |\/| | | _   /_\  makes it
Bitnet:   MARKV@UKANVAX		\/\/  /    | |  | | |__| /   \ possible...
Internet: markv@kuhub.cc.ukans.edu
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~