From: Eugene Leitl (eugene.leitl@lrz.uni-muenchen.de)
Date: Wed Jun 09 1999 - 14:28:41 MDT
mark@unicorn.com writes:
> Actually, I'd say we're seeing that already; I know of several chips
> which were months late because the design tools and processes just can't
> handle the complexity. I suspect we'll be able to get around it for a while
Then there is high time for innovation. I see no complexity issues
arising in reconfigurable architectures apart from the fab process
physics itself. Let's shift the complexity load to the software,
where it belongs.
> by changing design methodologies, but I think we're starting to hit the end
> of the curve. Intel, for example, has done a great job by simply taking the
Wintel has done a desastrous job. It has perpetuated and cemented an
insanely braindead way of doing things (not even inventing it: that
blame goes to Big Blue), and yet people praise it for all the
innovation they introduced. Bah, humbug. I can only hope the stalled
progress will suddenly give to a breathtaking wave of innovation. All
the ideas are there already. It is just nobody is willing to take the
risks, since new technologies are expensive.
> 8086 and pushing up performance while bolting on a few new bits here and
> there, but as far as I remember Merced was supposed to be out already, and
> even their new core logic chips seem to be well behind (e.g. see
> http://www.theregister.co.uk/990608-000022.html). Of course there's no
> particular technical reason for the new chips, so maybe they can drop them
> and push the 80x86 up to multi-gigahertz clock rates.
Oh, but there is. These gate delays add up after a while. One needs to
streamline/radically simplify the architecture to go to higher clock
rates. Structure shrink alone can't do it forever, you know.
I think Merced architecture is stale already well before it has become
available. The Sony Playstation 2 graphics synthesizer seems to be the
only truly innovative hardware design I am currently aware of. The
Xilinx people need some watching, as well.
> >but we'll still see an actual steepening of the performance/price curve
> >as most new systems come with two, then four, then eight processors.
>
> I agree, and I think we'll see a reversal of Intel's attempts to do
> everything on the CPU (e.g. soft modems) and more offloading of nonessential
> functions to dedicated chips.
There is no need to go dedicated. If there are hundreds or thousands
identical CPUs in each desktop there is sufficient horsepower to do
anything in software. God, we can do embedded RAM now. Nonvolatile
core is a the threshold of viability, including for reconfigurable logic
purposes. Let's dump code bloat and go WSI. There is a lot of silicon
real estate out there on these 300 mm wafers. And quantitive yield can
do wonders to prices.
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 15:04:05 MST