GP,
Good point on the design expertise. Its not that easy to clone
anymore. Not only are designs getting bigger, but the current
generation of elec design automation tools are having a much harder
time with the smaller geometries (.25 and below). Designing at
.18 um is much more difficult and takes a lot more design iterations
to get a working layout. Current tools are based on the assumption
the gate delays out weigh the interconnects, but at these small sizes,
it is the other way around (ok...copper is supposed to help here).
Sorry got a bit offtrack... just trying to support your positon...
Anyways i740 is a great example of trying to buy into a tech and
basically failing (for the time being).
Here 3 design teams in different companies/locations, with different
tools/flows tried to bring together a single chip. The thing was
what a year late, poor performance, yada yada.... now they are
just using their muscle (like get a free pII if you buy our
i740) to get rid of these pieces of silicon. But, it eventually
will pay off for intel... something called marketing muscle,
integrate it into a north bridge... bingo!
On the engineering thing though, i'de have to agree more with K's
comments. Intel has the best fabs and process, but i also am not that
impressed with their design teams, even outside the auburn/c&t/real3d
fiasco. If you saw say a startup like nVidia, you would see quite a
difference. Much less engineers using more advanced tools (and
fabless i might add). Intel's solution... throw armies at a chip...
and through brute force... we get it done. Hey whatever works i
guess.... well it almosts works... merced is another one that...
well looks like late again.
Sid |
