4 ms·
Is there a chance that we never see a traditional 10 nm Intel processor? I wonder whether they could embrace whatever problems they’re having through instead o
by ianai 8y ago
Is there a chance that we never see a traditional 10 nm Intel processor?
I wonder whether they could embrace whatever problems they’re having through instead of trying to model strictly 1s and 0s including intermediate states.
- bhouston 8y agoDefine traditional. I think Intel has to follow AMD's many CCX per CPU approach, especially after Intel hired Jim Keller who ran the Zen program at AMD. If they have started the many CCX program before hiring Jim Keller I would have hope it could come out in 2019, otherwise I think it would be 2020 at the earliest. Thus the "traditional Intel processors" will likely have time to launch on 10mn before then. 10mn must be ready at some time right?
- mbell 8y ago> I think Intel has to follow AMD's many CCX per CPU approach, especially after Intel hired Jim Keller who ran the Zen program at AMD. AMD's approach as a lot of downsides, I think AMD only went that way due to yield issues and trying to cut down on the number of designs they are fabing for cost cutting.
- wmf 8y agoIntel is working on chiplets connected by EMIB, but it's not clear whether they would be cheaper than a monolithic die. Intel's ring and mesh interconnects with shared L3 cache have far better performance than AMD's CCX+crossbar arrangement so there's no reason for them to go backwards.
- mikhailt 8y agoNo, we will definitely see a 10nm Intel CPU and 7nm and 5nm. That's a sure thing, the problem isn't that they can't do it, they can and they already have developed some 10nm wafers a few years ago; they just can't scale the yield production to meet the yearly supply. So, while they keep working on 10nm yield issues, they're refining the current fabs in order to keep the sales going. 7nm fab is already in development: https://www.businesswire.com/news/home/20170208005977/en/Intel-Supports-American-Innovation-7-Billion-Investment https://www.businesswire.com/news/home/20170208005977/en/Int... We just can't expect them to shrink at the same pace they had a decade ago, it was easier to drop from 130nm to 14nm but from now on, it becomes incredibly difficult to shrink without new technologies. Quantum computing is a few decades away from being feasible for consumer use.