Ruminations on the electronics industry from David Manners, Senior Components Editor on Electronics Weekly.
Intel Makes 14nm ICs
Intel has a 14nm process producing ICs in its laboratories, according to its managing director for northern Europe Pat Bliemer, in an interview with Nordic Hardware.
Although 14nm does not mean anything specific like a gate length, it does designate the successor process generation to Intel’s 22nm process which is due to move into production next year. “We actually have the next generation after 22nm running,” says Bliemer, “our engineers have the path to actually go and produce 14nm products.”
Intel’s date for having 14nm in volume production is 2014, and the company is said to be building two fabs capable of running the process.
The implications for Intel’s competition, and for the foundry industry, are profound.
“The announcement that Intel is running 14nm test chips in its labs must be sending icy cold shivers down AMD and GloFo’s spines, seeing they are still wrestling with the 32/28nm node transition,” says Malcolm Penn, CEO of Future Horizons, “even if the (as yet undenied) seriously bad press GloFo has been receiving of late were only half true, it still puts them a good three to four years behind Intel.
“That’s a tough situation for any chip firm to recover from, even the best run ones – historically recovery takes 1.5x to 2x the length of the starting gap, due to the fact you are trying to close in on a moving target,” adds Penn, “so with their key foundry on the ropes, it can’t be much longer before it’s time to call in the priest for AMD as well.”
“AMD could go to Intel as their foundry partner; their only other options are TSMC, which is still lagging Intel by a couple of years but at least they’re supremely competent, or Samsung … who now find themselves lagging TSMC in logic, thanks to their also being hamstrung
by the Common Alliance technology mistakes.”
The IBM club, which includes Samsung and GloFo, has elected to go for a gate first approach at 28nm which is proving to be a difficult technology to master.
Bliemer himself puts Intel’s lead at one and a half nodes ahead of the opposition.
Intel’s achievement also makes the fab-lite approach taken by so many companies look wrong-headed, says Penn.
“Selling off your fabs and getting out of manufacturing was the smart way to guarantee an easy and profitable future (and fat bonuses all round); until of course you can’t get your
wafers. And still some people say fab-ite is the future,” says Penn, “when hell freezes over maybe. When will chip CEOs give Wall Street the well-deserved single-figure salute and get on with being proper chip CEOs again?”