From: Jon M. Taylor <taylorj@ecs.csus.edu>
  To  : ggi-develop@eskimo.com
  Date: Thu, 9 Jul 1998 19:52:23 -0700 (PDT)

Re: [off-topic] RE: Fast lines

On Thu, 9 Jul 1998, David Waite wrote:

> > Jon M. Taylor wrote:
> <snip>
> >	It would be better to give your CPU instructions for fast >vector
> >and maxtrix logic, colorspace conversions, fourier transforms, and >any
> >other mathermatical or logical operations that OpenGL would use.  >It is a
> >bad idea to tie your hardware into one API.
> 
> Riight =-) Plus, if you tie it to the 1.2 version of an API, you are going
> to be in trouble when 1.3 comes out...

	Exactly.

> >	Six of one, half a dozen of the other.  It is a "media >processor",
> >according to current industry jargon.  A lot of people think that >this is
> >the way the industry is heading - eventually all the hardware logic >in a
> >computer except maybe some bus glue will all be in one ASIC.  Now >that
> >hardware is so cheap and powerful, the massive speed and >flexibility
> >increase that you get by unifying your logic and throwing away the >busses
> >more than compensates for lack of modular upgradeability.
> 
> Yes, but you then have problems such as:
> 
> -'my video card isn't working'... 'well, Replace your computer!'

	"My Nintendo 64 isn't working!"  "Replace your Nintendo 64".  
Whern things are cheap enough, it isn't worth it to de-integrate your 
design and lose speed - non-techie consumers will be using the product 
and they won't upgrade by parts anyway.

> -'well, we messed up the left channel of the integrated sound card, should
> we try making another round of silicon, or just ship it in mono?'

	Again, it is a cost and marketplace issue.  Nobody repairs VCRs, 
they throw them away and buy a new one.

> Also, since at my job we mess with large volumes of custom hardware, I can't
> see this in the forseeable future.

	How expensive is your hardware?  What is the nature of its target
market?  Those questions determine a lot of design issues.

> >	Here's something else to consider, too.  USB and its >successors
> >will bring us universal external peripheral expandability.  This >reduces
> >the need (and desireability) of having DACs associated with your >video and
> >audio hardware - just send the data in binary form to any external
> >peripheral (USB monitor, USB stereo, USB modem, etc).  Your >"computer"
> >should only have to deal with constructing a rectangular array of >colored
> >pixels, not the task of turning that data into analog electrical >signals!
> 
> Definately the successors, USB I believe is only 12 megabit/s, and I'm
> unsure whether it is packet-based or (I think) time-share based. Hopefully
> it isn't time-share based, because then two USB keyboards would make your
> USB camera not work.

	It is a daisy-chain bus, so there is the potential for devices
higher up the chain to hog the bus.  It isn't really meant for
high-bandwidth stuff.  Firewire (400 megabits IIRC) is more suited to
this, which is why it is used in digital video cameras and other releated
consumer electronics.  USB is meant for keyboards, mice, joysticks, etc. 

> USB soundcards sound very cool, the ones integrated into external speakers..

	Most decent stereo recievers and CD players already allow for a 
direct digital connection between them.  Cuts out a source of noise and 
signal lossage.  Even better would be to not go digital->analog until the 
speakers.

> they pull supposedly can pull off the system's power through the USB bus, so
> that the USB cable is the only cable needed. Also, USB is 'plug-and-play' so
> when you plug in the speakers, teh OS can look up the new device, plug in
> the drivers and play a soft 'bing!' to let you know your new sound device
> works.

	This would (and hopefully will) work well with KII/GGI Console.

Jon

---
'Cloning and the reprogramming of DNA is the first serious step in 
becoming one with God.'
	- Scientist G. Richard Seed