Ten emerging technologies to watch in 2010
Brett Davis
http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=221900117&pgno=4
8. Resistive RAM or the memristor
So products are launching soon.
4. Maskless lithography
No more million dollar mask sets slowing down tech upgrades.
Brett
Brett Davis
Brett Davis <gg...@yahoo.com> wrote:
> Ten emerging technologies to watch in 2010
> http://www.eetimes.com/news/semi/showArticle.jhtml?articleID=221900117&pgno=4
>
> 8. Resistive RAM or the memristor
>
> So products are launching soon.
TSMC is rumored to have killed its 32nm node
http://www.semiaccurate.com/2009/11/20/tsmc-rumored-have-killed-their-32nm-node/
Could this also be related to RRAM needing 28nm and below to work?
That also would force all designs to move to 28nm, and 32nm to get ignored.
I have never heard of fabs ever skipping a generation.
The economics are changing, fewer and fewer chips are made at
the bleeding edge. In which case all half node plans are dead.
Bernd Paysan
> Could this also be related to RRAM needing 28nm and below to work?
> That also would force all designs to move to 28nm, and 32nm to get
> ignored.
Why should that be the case? RRAM first and most needs a usable
memresistor material to work, the scaling down has no theoretical
problem, but neither should the scaling up. We have to wait and see who
comes first and delivers working chips.
> I have never heard of fabs ever skipping a generation.
Well, 28nm is a "half-node", just like 40nm, it's between two
generations (were a generation to the next doubles transistor count,
i.e. reduces feature size by sqrt(2)). Stopping with half node plans,
and going straight ahead one full node (i.e. from 40nm to 28nm) is
probably a good way to keep costs down. The half-node step for
foundries like TSMC exists to reduce the lead of Intel, i.e. even though
still trailing Intel TSMC can at least reduce the gap to one-half node
instead of a full node.
> The economics are changing, fewer and fewer chips are made at
> the bleeding edge. In which case all half node plans are dead.
The majority of PC chips, like CPUs, GPUs, and memory, are still made at
the bleeding edge, the rest never has. Globalfoundries has to follow
TSMC for bulk silicon, since that's what ATI needs for their GPUs, and
the only way to make the fab profitable of course is to fill the line.
--
Bernd Paysan
"If you want it done right, you have to do it yourself"
http://www.jwdt.com/~paysan/
Del Cecchi
Ebeam direct write is so 20th century. :-)
Del Cecchi
>
> The majority of PC chips, like CPUs, GPUs, and memory, are still made at
> the bleeding edge, the rest never has. Globalfoundries has to follow
> TSMC for bulk silicon, since that's what ATI needs for their GPUs, and
> the only way to make the fab profitable of course is to fill the line.
>
IBM does bulk as well as SOI.