By using this site, you agree to our Privacy Policy and our Terms of Use. Close
fatslob-:O on 06 May 2014
Edit Thread
Soleron said:
fatslob-:O said:

Moore's law is dying as we speak right now ... The whole semiconductor industry can't continue to depend on just miniturization for performance gains and all innovations to performance gains such as SIMD model processing have being exhausted for the most part. The last miniturization technology will likely be at a 5nm process node according to ex-Intel microprocessor designer, Bob Colwell. Moore's law is not about surpassing physical limits anymore because at that point an entity would also have to surpass economical limits too as transitioning to a new process node gets prohibitively more and more expensive each time thus becoming harder to justify. Once Intel releases broadwell processors they only have 3 more shrinks left ... 

It's worse than that. The whole industry is now a gen and a half behind Intel. Their "14nm" is worse than Intel's 22nm and doesn't represent an area shrink, so there's no cost reduction.

I don't think it will last to 5nm commercially, at least not with all the current players (Intel, Samsung, TSMC, Globalfoundries).

Really ? I wasn't expecting samsung's 14nm process node to match Intel's 14nm process node but I was assured that it's chip area scaling was 15% better than TSMC's 20nm or 16nm process node so relative to Intel's 14nm process node the size of samsung's transistors would be roughly 18.5nm so to speak. The reason why global foundries went to the trouble of licensing samsung's 14nm process node over their own 14nm process technology was because they could meet their customers demand on schedule while being able to produce chips at a marginally smaller size. This would literally put globalfoundries ahead of TSMC in semiconductor manufactoring technology by roughly 12 months due to the fact that they will have implemented FinFet technology first while having better chip area scaling too! Mass production is slated to be around early 2015 for global foundries.

http://www.eetimes.com/document.asp?doc_id=1321974

I honestly do think that 5nm might be the last shrink ever. At this point it gets insanely expensive to transition to a new prcoess node and the scariest part of all this is that we don't even have next generation lithography technologies ready yet such as extreme ultraviolet lithography! The only time that I can see us pushing under 5nm is that EVERYONE would have to collaborate instead of competing.