With much of the performance computing market sitting on top of 14nm (or similar) nodes, it's time to look to the future. GlobalFoundries is looking past 10nm, as it sees 7nm as the next major step in process technology, and it's not alone.
In a new announcement, the semiconductor foundry said that it will be able to deliver 7nm production soon, based on FinFET designs. Complementing that, it'll be able to "significantly" re-use much of the same equipment that was used for 14nm, which is currently being used at its Fab 8 campus in New York.
Jim McGregor, founder and principal analyst at TIRIAS Research, said that this was a good move for the foundry. According to McGregor, the move to 10nm simply doesn't offer enough gains to warrant moving from 14nm, adding that we'd "see only marginal performance and power benefits". Whereas 28nm > 14nm was a massive leap, 14nm > 7nm will be as well. It'll give chip-makers the ability to cram twice as many transistors into the same surface area, and improve power efficiency at the same time.
A rough estimate of performance gains is about +30%, which means that if current products were magically ported over to 7nm, those are the kinds of improvements we could expect. However, moving to a smaller process also means the ability to add more performance hardware, such as more cores in graphics cards, so 30% is just a safe estimate. It could be that the first GPUs, for example, to come out on the 7nm process will gain much more than 30% in performance.
What could really stand to benefit greatly from this shift will be mobile devices. It could technically be possible to make our mobile SoCs twice as fast as they are today, or at least far more power efficient.
GlobalFoundries says that 7nm testing is already going on at its Fab 8 location, but at this point, it's unclear when us consumers could expect products to hit the market based on the super-small process.