• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

Intel 10nm and GF 7nm at IEDM 2017

Page 10 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

plopke

Senior member
Jan 26, 2010
220
67
101
I hide nothing. In my forth post on this thread, I announced as an AMD fan and investor.
I really do not like doing this since my skill level of English spelling and grammar is very poor to begin with but after catching up on this thread.
Please use new lines,paragraphs and not 20 million ................... , structured text is always so much easier to read and correct.
 
  • Like
Reactions: Tarkin77

LTC8K6

Lifer
Mar 10, 2004
28,523
1,569
126
I hide nothing. In my forth post on this thread, I announced as an AMD fan and investor.
But it's naive to treat a company as a "good guy" (or a "bad guy").

Companies are out to make a buck off of you. That's what they do. If they don't make profits, they die.
 

Beemster

Member
May 7, 2018
34
30
51
baby, you just ain't seen nothing yet!

https://www.youtube.com/watch?v=7miRCLeFSJo

AMD at Computex. Lisa was just glowing about 7nm EPYC 2 results in the lab. She claims sampling to customers very soon. It looks like an April 2019 launch could be a lock. Having known Lisa well, that is about as animated as she will get while in "CEO mode".

https://www.amd.com/en/events/computex

http://markets.businessinsider.com/news/stocks/pushing-boundaries-for-cpus-and-gpus-amd-shows-next-generation-of-ryzen-radeon-and-epyc-product-leadership-at-computex-2018-1026647937

https://www.forbes.com/sites/tiriasresearch/2018/06/05/epyc-year-bodes-well-for-amds-server-future/#768392543651

.......enjoy
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,474
136
But it's naive to treat a company as a "good guy" (or a "bad guy").

Companies are out to make a buck off of you. That's what they do. If they don't make profits, they die.
Yes. Companies are out to make money, but not by breaking the law. Intel has been found guilty by multiple regulatory organizations and courts around the world of antitrust behaviour. Intel has earned the image of an abusive monopoly through their own actions and behaviour. Nobody can deny that.
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,474
136
Ok folks. Here you have it. Intel's 10nm process first look by techinsights.

http://techinsights.com/technology-intelligence/overview/latest-reports/intel-10-nm-logic-process/
  • Logic transistor density of 100.8 mega transistors per mm2, increasing 10nm density 2.7X over the 14nm node
  • Utilizes third generation FinFET technology
  • Minimum gate pitch of Intel’s 10 nm process shrinks from 70 nm to 54 nm
  • Minimum metal pitch shrinks from 52 nm to 36 nm
Process Highlights:
  • Deepest scaled pitches of current 10 nm and upcoming 7 nm technologies
  • First Co metallization and Ru usage in BEOL
  • New self-aligned patterning schemes at contact and BEOL
Design Highlights:
  • Hyperscaling via 6.2-Track high density library
  • Contact on active gate (COAG) cell-level usage
 

raghu78

Diamond Member
Aug 23, 2012
4,093
1,474
136
Hans de Vries predicted a 6.2 track cell layout when the IEDM 2017 paper was presented. Kudos to Hans for his fantastic observation and analysis. For those who do not have a semiwiki login I will post it here with full credit to Hans de Vries.

Just a tidbit: The "cell-size" MMP x CPP of Intel's process is actually 40 x 54 = 2160 nm2 (using M0) or even 44 x 54 = 2376 nm2 (using M2)

This is because the M1 lines run in parallel with the gates, with a pitch ratio of 36:54 = 2:3. You can see it in this photo over at Dick's article where M1 is just visible at the top of the image.

In Intel's 14nm they used M2P x CPP = 52 x 70 = 3640 nm2 and M2 scaled from 52nm to 44nm.

This makes the Track number 7.65 (=272:36) a bit odd. A better number, using M2 as in your example image the track count would be 272:44 = 6.2 tracks




The odd thing is the track size 272:36 = 7.55. The only 36nm (M1) lines run parallel with the gates...
Maybe they relaxed some layers from an original 36 nm to 44 nm and went from a 7.5 track design to a 6 track design to keep the Logic cell size equal...

That would look like this I presume:



 

formulav8

Diamond Member
Sep 18, 2000
6,998
521
126
Intel will officially lose the crown. Nothing they can do about it now either.

Edit: IIRC, they are changing their density target from 2.7x to 2.4 instead.
 
Last edited:

raghu78

Diamond Member
Aug 23, 2012
4,093
1,474
136
Intel will officially lose the crown. Nothing they can do about it now either.

Edit: IIRC, they are changing their density target from 2.7x to 2.4 instead.
Intel lost the density lead as soon as Samsung and TSMC started shipping 10nm wafers in HVM in H1 2017. Intel now has leadership transistor performance with 14++ which is the main reason for their clock frequency and ST perf lead in PCs. Intel will not be able to catch up with TSMC on density as N5 is expected to be in HVM in 2020 and N3 with nanosheets in 2022.
 
  • Like
Reactions: goldstone77

CHADBOGA

Platinum Member
Mar 31, 2009
2,023
636
136
Intel lost the density lead as soon as Samsung and TSMC started shipping 10nm wafers in HVM in H1 2017. Intel now has leadership transistor performance with 14++ which is the main reason for their clock frequency and ST perf lead in PCs. Intel will not be able to catch up with TSMC on density as N5 is expected to be in HVM in 2020 and N3 with nanosheets in 2022.
With BK at the helm, it makes Intel catching up far less likely, but who is to say that TSMC won't run into some problems with their future processes, like Intel has with 10nm(for different reasons no doubt)????
 

ASK THE COMMUNITY