IBM and a group of partners have managed to produce the first processor using a 7 nanometer process. This is just a test chip, and it’ll still be a while before processors based on 7nm technology used in real-world products. But this is the first time anyone’s demonstrated a working chip smaller than 10nm.
Why is smaller better? Broadly speaking, when you can pack more transistors into a smaller space, you get chips that are both faster and more efficient. That’s a large part of the reason we’ve seen Intel, AMD, and other chip makers move to smaller and smaller processes in recent years. The latest Intel chips, for example, are 14nm processors.
Intel has said that it expects to produce 10nm chips in 2016 and 7nm processors 2 years after that.
But IBM, GLOBALFOUNDRIES, and Samsung are already showing off a 7nm test chip which could put more than 20 billion transistors onto a chip the size of a fingernail.
IBM says it used Silicon Germanium channel transistors and Extreme Ultraviolet lithography to create the chip, which could offer up to 50 percent more power and/or 50 percent better power efficiency than the best chips available today.
via Ars Technica
What? IBM just sold its semiconductor division to the Abu Dhabi owned Globalfoundries (sic) company. (By the way, Globalfoundries is what absorbed the fabrication division of the U.S. semiconductor company AMD.) Man, talk about America in decline… American engineers and scientists (what’s left of them) invent stuff, then the bean-counters, lawyers, government regulators and tax collectors make sure it gets shipped out of the country ASAP.
IBM is a global company these days. They might be registered as a US corporation, but many of their finest research and development achievements have long been produced elsewhere (or by imported talent).
The UK has been there for years. ARM have no fab plant.
Getting down to the 7nm size is beyond amazing! I remember reading just a year or two ago how going that small was impossible. Of course, the real trick it to be able to do it in volume and at a price that the market can sustain. If IBM & Co. can do it, then Intel, Qualcom, Samsung, etc. could have a big problem.
Uhh..Samsung is part of that group of partners with IBM on this effort, and Qualcomm is fabless, so it has nothing to do with them.
Oops, I missed the Samsung bit. Qualcomm is still in trouble fabless or not. They still have to compete unless they can get IBM and company to produce their chips. Good luck with that.
which they do plan to do
Comments are closed.