Accelerator boards - no future? Bad business?

[ethan at 757.org]

> I'm most familiar with the Amiga accelerators.  I suspect those who
> produced them were helped out greatly by a couple of factors.  One is that
> the hardware specs were very well known and full schematics were available
> for most (all?) Amigas.  I doubt the same is true of SGI machines.

SGI stuff is still much under wraps as far as I know. Thats why NetBSD and 
similar for SGI are still pretty rudimentary. All propriatary, those 
groups don't want to use stolen info, and who knows if the documentation 
still even exists after the rackable purchase.


> The
> other is that many Amigas had processor "slots" (with edge connectors)
> rather than some tiny fiddly ball-grid array etc...  but I'm not a EE; so
> maybe that's bunk.
>

High clock rates for data busses of modern systems wouldn't work with old 
style card edge interconnects AFAIK. Also, I don't think the old PPC 
accelerators for the Amiga or the ones for the Macs (that sometimes had 
CPU upgrade slots) would really accelerate everything - you might get 
faster processor instructions and maybe L1/L2 cache -- but memory and 
I/O are still slow?

> When I look at these boards they seem like they'd be a LOT of work to
> develop and produce.  I wonder how they were even economical to make back in
> the day.  Plus, now the user base will probably only shrink.  It's not a
> great business model for a hard-to-produce item.  It doesn't keep my
> techno-lust from wanting it, though.

They were really expensive at the time :-) US dollars have lost a lot of 
value (especially given overpriced housing.) The old $3000 Tandy system 
with a 20 meg hard card and TGA/CGA graphics is like $7000 in todays cash.

Amiga stuff was always pretty expensive, and you had to pay VGA monitor 
prices for it's crappy TV display (1084S). And yea, at least Atari stuff 
was made in Taiwan so they were using cheaper labor then too.


> So, here's the question. Is my dream likely to ever be possible enough that
> a boutique shop could pull it off and not lose their shirt on the production
> costs and R&D to do it ? I'm encouraged by things floppy emulators that are
> produced for these old machines. However, that's probably significantly
> easier to make than a CPU accel board.
> What do you guys think?
> -Swift

You gotta build it!

There are better tools for doing it now (board layout apps and such) and 
you can get boards made in China cheap. So that's on your side.


--
Ethan O'Toole