I'm new to this channel, but I have scanned the last month and a half
of mail (it took hours), so I'm not as uninformed as I might appear to
be at first glance.
I originally heard about the riscy project earlier in the summer, and
thought, "Oh, that's nice. How the heck are they going to get this
thing *built* for a reasonable price? Who has the necessary tools?"
That question having been answered in probably the most satisfactory
way possible, I'd like to put forword my suggestions. Take them with
a grain of salt, as I have done very little digital design...
I would prefer a mips4k cpu. I think that it will be a better
long-term solution. About 3 years ago, I bought a 386sx16. It was
all I could afford, but I now know that I would have been better off
to wait a little longer and get a real 386. I'm still using the sx16,
as linux makes it useable, but I'm looking for a step up that will not
feel completely obsolete in a year, and I'm willing to pay a little
extra for it, if necessary. It's not clear to me that it would
actually cost more for a mips4k solution than a mips3k solution by the
time this board gets made. I don't think we know enough about prices
yet to know.
I would also like to make a general comment about integrated
functionality. This board has to be fairly flexible, since there are
not a really large number of us buying this. It has to cover the
needs of several different kinds of users. Therefore, where any
improvement that is significantly popular (i.e. IDE controller) and
costs little ( oh, say, < $5 ) to add to the current design, I think
that it ought to be added. This strategy will probably make the whole
design cost $25-$50 (or more) more than it might if we tried to hold
down the costs of everything, but I think the returns for that
For instance, as it appears that a controller for 2 '550 uarts, 2.88
FDC, IDE, EPC Parallel, and a few other things that skip my mind at
the moment is available, and that that controller will cost less than
$10 more than a very minimal version of that part, and will provide
several very useful features that might not otherwise be included (EPC
Parallel, IDE, and '550 uarts in addition to some uart capacity and
some FDC capacity), it makes the most sense to use this part. (Sorry,
I read the part number at about 3 am this morning, and don't recall it
Adding most of the parts people want when those parts are cheap is
probably the most economically viable option, not only for us hackers,
but also for Waldorf. The more MB's they sell, the less they cost for
the features added, and the more hackers they have working on linux.
Consider this: By making the motherboard *too* bare-bones, the number
of people who want to purchase one could drop enough that the
price/piece could rise to just as much as if a few extra chips were
added. Leveraging the $5000 setup cost across significantly more
purchasers could literally pay for itself, by making the cost/piece
drop by as much as the extra pieces on the board, and then we could
all be happy.
I know that the preceeding paragraph is speculation, but I can't say
anything more without taking a cumbersome and inaccurate vote. I
tried to get a feeling for what people wanted to buy as I read the
archives, but since many people kept changing their minds (reasonably
so, as this is a design discussion), it is hard to tell just how many
more people would buy a well-populated board than a sparsely-populated
board, and what a general consensus on what well vs. sparse population
In short, I want to buy one, but only if it is going to have all the
functionality that I want on the motherboard, and only if the
performance will be reasonable, not only now, but for at least a few
years to come.
I'm sorry to have rambled so long; I didn't mean to. I hope that it
is worth it, and I'd like to help. I am a linux kernel hacker, so I
think I can be of at least some help to this project.