Let's not cut off our nose to spite our interface - Arron Rouse
TAIWANESE fabber of chip wafers, TSMC, announced it has produced the first "functional" 28nm SRAM 64Mb cells.
This might sound a bit too geeky, but it looks like TSMC is holding all the 28nm cards and can move on to several levels of 28nm silicon, each one targeted at a specific customer segment.
In case you hadn't noticed, chips like other stuff come in a variety of forms, flavours and packages. In this case, TSMC has planned out production of High-performance, High-performance Low Power and Low Power silicon, using High-K/Metal gate (HKMG) technology on the first two and Silicon Oxynitride (SiON) on the latter.
This particular R&D achievement allows the company to start attracting designs from just about all market segments, from embedded designs to server and desktop PCs, and starting mid-Q1 2010 the company will begin risk production of said technology. SiON has been around for a while now, and TSMC has pushed its use in 28nm despite triggering some industry disbelief with at least one high-ranking IBM heavyweight having challenged the company directly.
The SiON gate as a component is an old acquaintance of TSMC's and it expects that to allow the company to jump straight into the action in no time by quickly ramping up production and flooding the market with the low-power - and as of now, cheaper - alternative to HKMG.
To give you an idea of what's in store with 28nm circuitry, you can do the maths right now. SiON gate technology is expected to outperform TSMC's current 40nm low-power silicon by being 50% faster or up to 50% less power consuming, and it also packs twice the circuit density.
The reality will be somewhere in between faster and lower power, as usual, but what TSMC is really doing is offering clients design options.
Qualcomm is first in line for this, eagerly waiting to start work on 28nm devices in early Q3 2010.
Rival reaper of wafers, Global Foundries, announced sampling of 28nm back in March, although in not as great a variety as this - but you can rest assured it'll get there too. µ
Pretty arrogant coming from a bunch of clowns that can't even master 40nm, right, AMD and NVidia?
TMSC = Those Masturbating Sycophantic Clowns.
28 NM ARTICLE HAS GONE TO CLOUDS, YET BASILY FIRST WILL BE LARGE CHIPS THAT TAKE ADVANTAGE OF SIZE directly as performance, Desktop/Server. then smaller stuuf, like mobiLap & then really small, mostly locked in by design, yet better yeilds & battery life.
890 on its way in Slow Lane, expected now in early spring 2010. No difference from 790Fx is expected. same with 58XX game cards, only sooner to market.
28 nm is one step from bare bottom, 14 nm. one more refresh & you be judge. Was IT Worth It?
Theres never going to be DNA CPU(except bS Pundits use of coliform), yet there will be SSD CPU & its scale is massive number of transisotrs, scale becomes something new. Poor Gordie, you should have worked it into Life Expectancy thanger. moore? Never Moore? moore?
food Pellets for Thought, These Days.
DRASHEK
Why is 14nm bare bottom? Even current public roadmaps hit 11nm with current immersion litho, EUV & co. Sure it looks like bottom from where we are but once we get down to it, we will think otherwise.
I'm wondering if they can make chips without many defects given the size and leakage on those size of gates. This could be one of those situations where the economy is pushing too hard and too fast.