riscy
[Top] [All Lists]

Re: Desperate idea?

To: riscy@sunsite.unc.edu
Subject: Re: Desperate idea?
From: aki@akix.cts.com (Aki Atoji)
Date: Fri, 25 Feb 94 01:26 PST
In-reply-to: <9402250233.AA20216@rei.com>
References: <9402250233.AA20216@rei.com>
6692 writes:

 > Look, I'm not singling out anyone for criticism, and I'm certainly
 > not a big fan of 486 PCs.  But if _cheap_ is what you want, I really
 > think you'll be disapointed again, just as you all were when Andy
 > Busse announced Waldorf's costs for his board.
 > 
 > I originally signed onto this mailing list because I thought some
 > people wanted to _design_our_own_ main board and put it under
 > copyleft so everyone would have access to complete documentation,
 > and so anyone could manufacture it, and so anyone could improve
 > it, as long as they kept it under copyleft.
 > 
 > I never imagined it would be cheap, but I figured it might not be
 > completely unreasonable if we kept realistic goals.  I've continued
 > to lurk here, keeping my peace, but I pretty much lost interest when
 > the discussion shifted from 32-bit R3000's to 64-bit pie-in-the-sky,
 > because I figured the probability of anything actually being built
 > went from 1 chance in 10 to 1 in 100, or less.

I must say I share this sentiment.  It seems that somewhere along the
line, the focus shifted from a design with reasonable hardware with
reasonable cost that was to be copylefted, to 'this would be neat if
we could do (take your pick: PCI/VLB/FDDI/ISDN/megsof 0.7nS 2nd level
cache/whichisthegreatest64bitCPU/pentiumsucks/etc.)'

Although I'm a software type now, I've done hardware designs in the
past including an i960CA-33 (that's Intel's superscalar RISC CPU)
board with SCSI, and have gone through the customary part price
scrutiny that's required for any commercial hardware product.

Parts that run at high speed just aren't cheap.  If you want
gloriously fast Xstones and SPECmarks, you might as well buy the whole
package (including benchmark figures) from DEC or HP.  I would much
prefer to see a reasonable design at a reasonable cost that is for
everyone to share, tinker, improve up on in terms of both hardware and
software.  That goal in itself probably isn't as easy to achieve as it
sounds, but it's not as bad as trying to make a 2000 SPECmark machine
yourself.

I agree that if cheap is what you want, you can get it down at your
local computer store.  To be honest, an ISA bus PC is as close to
public domain computer design as you get these days.

As for myself, I have i960 and 29K setup right here at home that I
tinker with.  They were from my embedded programming/design days, and
none are suited for running general purpose Unix-like kernel.  

Now, I would really love to get my hands on something like a 3081 or
AM29030 board with built in SCSI, Ethernet, serial/parallel ports and
a couple of ISA slots for video and other things.  No, this isn't your
super-fast workstation killer.  It probably won't score very well on
SPECmark neither.  Worse yet, it'll probably cost a few hundred bucks,
excluding memory.  What's on it isn't much more than the eval boards
you can buy from IDT or LSI Logic or AMD.  But that little extra such
as SCSI, ethernet and video are a real pain to have to wirewrap and
write drivers for myself.

But then, if such a design existed, then I can try a lot of things
that I always would have liked to do, and some of it people wouldn't
mind playing with themselves.

Now, if any of you have a 3081 or 29030 eval boards laying about,
please let me know.  


---------------------------------------------------------------------------
Aki Atoji             Unix, X, Networking, Japanese and Embedded Consulting
                                                           aki@akix.cts.com
---------------------------------------------------------------------------
<Prev in Thread] Current Thread [Next in Thread>