TSMC announces 6nm process: the intermediate step between 5 and 7 nm
TSMC is working on a 6nm production process, which is a bit of a surprise as it never appeared on earlier roadmaps, these went from 7 and 7+ directly to 5 and 5+.
Both N5, N6, and N7 are produced by a euv process (extreme ultraviolet ). The production of the 5 nm process started earlier this year, 6 nm would not be around until the first half of 2020.
-- EETimes --
N6, TSMC’s terminology for the 6nm process, will have three advantages, according to CEO CC Wei, speaking at the company’s latest quarterly results announcement last week. N6 will have design rules that are 100% compatible with N7, allowing customers to directly migrate from N7, he said. In addition, N6 will increase logic density by 18% from N7 and provide a highly competitive performance-to-cost advantage. Finally, N6 will offer shortened cycle time and reduced defect density.
“The numbers N6 and N5 look pretty close, but actually the performance -- they still have a big gap,” Wei said. “N5 compared with N7, actually, the logic density increases by 80%. N6 compared with N7 is only 18%.”
N5 will deliver a significant jump in density and performance, but it will also be a full node upgrade. Unlike N6, it will require customers to design products from the ground up.
Risk production of N6 is scheduled for the first quarter next year with volume production starting before the end of 2020, according to TSMC.
Senior Member
Posts: 483
Joined: 2005-02-23
It's just rebranded 7FF++. The same way they did with 12FF that was rebranded 16FF++. And the same way Samsung did with 11LPP (14LP++++) and 8LPP (10LP+++). Or GloFo with 12LP, which is rebranded 14LP+, which is 14LP++++, including Samsung 14LPP (14LP+)
And they are not really ahead of Intel. Intel 10+ is denser. Samsung 7LPP (7LP+) that goes mass production later this year is even more denser.
Senior Member
Posts: 11532
Joined: 2012-07-20
It's just rebranded 7FF++. The same way they did with 12FF that was rebranded 16FF++. And the same way Samsung did with 11LPP (14LP++++) and 8LPP (10LP+++). Or GloFo with 12LP, which is rebranded 14LP+, which is 14LP++++, including Samsung 14LPP (14LP+)
And they are not really ahead of Intel. Intel 10+ is denser. Samsung 7LPP (7LP+) that goes mass production later this year is even more denser.
False. Look at nVidia's 16nm vs. 12nm chips. They both have between 23~25M transistors per mm^2. But then look at power draw per transistor, per clock.
GTX 1080Ti and RTX 2080Ti have practically same boost clock, and same TDP. But RTX2080Ti does that with almost 60% more transistors.
Or take that GTX 1080Ti against RTX 2070... almost same transistor count, same performance per transistor, And RTX 2070 eats good 30% less power.
nVidia could have likely go for higher transistor density instead of going for power efficiency... and enabling even bigger GPUs... Instead, they went with power efficiency and clock as they usually do.
And as written in article 7 to 6 is supposed to enable 20% area saving, that's not small thing as that can be traded for other properties.
- - - -
As for that intel being ahead. Where? I do not own intel's "Superior" 10nm chip. Very few people do and they seem to be unhappy. While there are actually dozens of different 7nm products people can buy, and buy in volumes.
- - - -
notebookcheck's review of that intel's 10nm thingy was hilarious. Especially since they historically used to ignore AMD's APUs. If someone like that tells that they are worse than older generation...
Senior Member
Posts: 483
Joined: 2005-02-23
Nvidia is fabless company. It doesn't have own 16 and 12 nm. They hire TSMC for that. But, they have a bit optimised for Nvidia process 12FFN (FinFet Nvidia). Density and effectiveness varies between generation variations. It's not always same. It could be all up to +/-20% each different directions. 12FF, 11LPP, 8LPP, 12LP, 14++ are all just variations of 16FF, 14LP, 10LP etc.While 5FF is a new process that is over 70% denser. Each generation is usually 100% density improvement.
Senior Member
Posts: 8141
Joined: 2008-07-31
You can say that Intels 10nm is denser all day long, which is all fine and dandy. No arguments there, it's great that it's denser.
......
But what's the point if it doesn't work? Or, what's the point if it doesn't improve....anything? Speed, power consumption, heat, etc.?
Intels 10nm may be "denser" but it is not "superior" because of that, if Intel decided right now to use TSMC's 7nm and put out a brand new line, full line from top to bottom CPUs, we'd see intel likely improve their current architecture a decent amount, in heat, power consumption, and potentially performance.
So.......what exactly is the point of continually stating that Intels 10nm is denser then 7nm, implying that it's a good thing, if it doesn't....bring anything good? And before you say "well it will!" no no no, i'm not interested in theoretical, we are here, now, with TSMCs working, and seemingly efficient, 7nm process, and we KNOW from previous release of intels 10nm that it is not beneficial over their 14nm. What we DON'T know is if that will ever change, or if they will skip over 10nm to their next process (my bet) for mainstream anything and keep their 10nm parts in the background.
So again, what's the point? Until Intel releases something on their 10nm that actually proves all the work they have done is worth it in the end, what is the point, right NOW, in talking about intels 10nm being..."superior" to TSMCs 7nm?
Senior Member
Posts: 1127
Joined: 2017-02-14
This is very smart on TSMC's part. This is exactly why they are ahead of Intel now. As these shrinks get harder and harder to achieve doing iterative processes instead of a revolutionary process like Intel was doing with 10nm which was more dense than even TSMC's 7nm looks like the way to go.