cmos technology nodes: trends, challenges, opportunities?

9
IBM Research – T. J. Watson Research Center CMOS Technology Nodes: CMOS Technology Nodes: Trends, Trends, Challenges, Challenges, Opportunities? Opportunities? Ghavam Shahidi IBM Research Division 2 IBM Research CMOS Scaling: CMOS Scaling: Dennard Dennard’s Theory Theory Smaller Faster Lower Power Scaled technology generations

Upload: others

Post on 03-Feb-2022

3 views

Category:

Documents


0 download

TRANSCRIPT

IBM Research – T. J. Watson Research Center

CMOS Technology Nodes:CMOS Technology Nodes:Trends,Trends,Challenges,Challenges,Opportunities?Opportunities?

Ghavam Shahidi

IBM Research Division

2

IBM Research

CMOS Scaling: CMOS Scaling: DennardDennard’’ss Theory Theory

Smaller

Faster

Lower Power

Scaled technology generations

3

IBM Research

Device Performance: ~30% per GenerationDevice Performance: ~30% per Generation

4

IBM Research

Performance ChallengePerformance Challenge

Technology performance gain continuing

– Through 32-22 nm nodes

– Power-delay curve improving

Issue: Supply voltage not going down

– Power density (and power)

Solution: Industry moving to multi “low-power” cores

– Performance not through frequency

Do we care as much about performance?

– System performance: Number of cores vs. the frequency

Drive to higher VT’s and lower supply => Effectively devices not as fast

5

IBM Research

CMOS Density: MooreCMOS Density: Moore’’s Laws Law

30% Reduction in W and device pitch

– 50% area reduction

– 2X device/area (every 2 years)

Devic

e P

itch

W

6

IBM Research

Device Pitch ScalingDevice Pitch Scaling

2006230-35065

55-65

70-80

90-100

120-130

170-180

Device Pitch (nm)

2017-18

2014-15

2012

2010

2008

Year

15

22

11

32

45

Node

7

IBM Research

Lithography Roadmap (ASML)Lithography Roadmap (ASML)

70-80 nm pitch (15 nm node) is feasible with immersion

8

IBM Research

15 nm (Post 22) Node Cross Section:15 nm (Post 22) Node Cross Section:

9

IBM Research

15 nm (Post 22) Node Cross Section:15 nm (Post 22) Node Cross Section:

10

IBM Research

15 nm (Post 22) Node Cross Section:15 nm (Post 22) Node Cross Section:

11

IBM Research

Challenge: L ScalingChallenge: L Scaling

30

40

50

60

70

80

90

180n

m13

0nm

90nm

65nm

45nm

32nm

*

22nm

*

Technology

Gate

Length

(nm

)

12

IBM Research

Value of CMOS Scaling: Transistors/Unit AreaValue of CMOS Scaling: Transistors/Unit Area

0.9 B0.2 B65

~6.4 B

~3.2 B

~1.6 B

0.8 B

0.4 B

Transistors /cm2

25.8 B

14.4 B

7.2 B

3.6 B

1.8 B

Transistors /450 mm2

15

22

11

32

45

Node

Extrapolate from 65 nm Intel’s Xeon and IBM’s Power 6 dual cores

13

IBM Research

Value of CMOS Scaling: Transistors/Unit AreaValue of CMOS Scaling: Transistors/Unit Area

0.9 B0.2 B65

~6.4 B

~3.2 B

~1.6 B

0.8 B

0.4 B

Transistors /cm2

25.8 B

14.4 B

7.2 B

3.6 B

1.8 B

Transistors /450 mm2

15

22

11

32

45

Node

Extrapolate from 65 nm Intel’s Xeon and IBM’s Power 6 dual cores

3D will add another 2-3X more transistors/unit area

14

IBM Research

CMOS Paradigm ShiftCMOS Paradigm Shift

Device count will continue to go up– Moore’s Law goes on

Key challenges:– What to do with multi billion transistors / cm2

100’s of Millions of high

performance devices in 2D

10’s of Billion of low

performance devices in 3D

15

IBM Research

What About What About ““NanoNano””-Devices?-Devices?

K. H. Yeo et al. (IEDM , 2006)

N. Singh et al. (IEDM , 2006)

STM Image

IBM

16

IBM Research

What About What About ““NanoNano””-Devices?-Devices?

K. H. Yeo et al. (IEDM , 2006)

N. Singh et al. (IEDM , 2006)

STM Image

IBM

Key Attribute: Small width

17

IBM Research

When Will We Get to Real When Will We Get to Real NanoNano??

10-168

7-105

16-2511

25-35

35-45

50-60

70-80

Minimum W (nm)

15

22

32

45

Node

Nano-devices by 2020-2025 in production?

Key Attribute: Small width