From: Eugene.Leitl@lrz.uni-muenchen.de
Date: Wed Nov 29 2000 - 11:39:22 PST
Jim Whitehead writes:
> It seems to me that custom game machines can be optimized for games in ways
> that PCs cannot. For example, PC performance is severly limited by slow bus
It's largely homogeneity and the economies of scale. I sure hope asian
markets will kindle, and increase the potential PC clientele. The
amount of semiconductor waste and the power burn will become truly
stupendous.
> speeds, and one of the reasons the bus speed is so slow is because it takes
> the development of a new industry standard to increase the bus speed, so all
> the cards, chipsets, motherboards, etc. work with it. A specialized game
I don't know, the innovation cycle in the PC is shorter, so you can
profit from progress increments sooner. The dichotomy between dumb
frame buffer and smart CPU isn't true even since 1985/86, or so, when
Amiga 1k came out. The bus between the GFX chipset and store is not
standartized, and since you need CPU mips for game behaviour
complexity and the endstage rendering engine for speed AGP (it's 8x
now, is it?) should not be the bottleneck. Eventually, hardware
dictates that you move your intelligence into RAM -- but this equally
applies to both game consoles and video accelerators. I'm actually
surprised that nVidia didn't include embedded memory into their next
generation GFX chipset -- that might even give game consoles a slight
edge, for a short wave. After embedded memory it has to be
parallelism, and then, what, neuronal architectures? No one knows.
> machine can up its bus speed with impunity, since it doesn't have to worry
> about compatibility.
Yes, but you can't expect the game developers to grok each
architecture change instantly. A PC has much more continuity than each
console generation. Even so, 3d accelerators' capabilities take a while
to be exploited fully by game programmers.
This archive was generated by hypermail 2b29 : Wed Nov 29 2000 - 11:54:22 PST