From: Eugene Leitl (eugene@liposome.genebee.msu.su)
Date: Thu Jul 16 1998 - 06:57:22 MDT
Harvey Newstrom writes:
> > Intel plans even faster chips --
> > but who needs them?
>
> I think the problem is that chip speed is not the limiting factor for
> most computers. Bus speeds, disk speeds, network speeds can't keep up
> with the chip. Faster chips won't make Microsoft Word run faster, won't
> make a web page load faster, won't make an e-mail arrive faster.
Oh, I think I could use a 10 GFlop CPU with current generation's
800 MByte/s memory bandwidth just fine (for QM, MD). Bus speed is
a bottleneck? Agreed. Let's kill the motherboard, move the bus
onto die, so you could make it few kBit wide, and run at several
GHz. Disk speed is a bottleneck? Let's go totally solid state, or
holooptical. Network too slow? Go GBit/s, and beyond (just now you
can't feed even GBit Ethernet). If this is still insufficent, go
analog FPGA WSI. Or molecular circuitry.
Yes, and we will saturate each new resource generation easily. Current
high end PCs are way insufficient for robust realtime speech input,
not even mentioning vision. Their neural DSP performance is pathetic.
And complex VR simulation games can absorb essentially infinite
resources as can household robotics. And likewise for x, y, z.
The argument 'nobody needs that much power' is recurring
periodically. This hadn't occured in the past, this won't occur in the
future. And thanks god, for it keeps life being interesting.
'gene
This archive was generated by hypermail 2.1.5 : Fri Nov 01 2002 - 14:49:21 MST