Lessons from an Emerging Technology:Superconducting Computing

Dr. D. Scott Holmes

Booz Allen Hamilton, IARPA contractor

2017-11-30

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Future Supercomputing Vision

2

Hybrid technologies: digital (CMOS, SFQ), probabilistic, analog, neuromorphic, reversible, and quantum computing (QC) — whatever works best!

SFQ digital platform supports multiple cryogenic technologies

Requires optical interconnects between room temperature and cryogenic nodes

Courtesy of the Oak Ridge National Laboratory, U.S. Department of Energy

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Superconducting Computing Approach

3

Low temperature operation (~4 K)Allows different physics

Commercially available refrigeration

LogicSFQ (Single Flux Quantum)

Switching energy ~ 2x10-20 J

Memorycompatible with SFQ logic

InterconnectsSuperconducting in the cold space

Input/Output: electrical or optical

Major energy reductions in all 3 areas!

~2 mV

~1 ps

~c/3, nearly lossless

S

F (soft)IF (hard)

S0

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Notional Prototype, IARPA C3 Program

4

Metric Goal

Clock rate for superconducting logic

10 GHz

Throughput (bit-op/s) 1013

Efficiency @ 4 K (bit-op/J)

1015

CPU count 1

Word size (bit) 64

Parallel Accelerator count

2

Main Memory (B) 228

Input/Output (bit/s) 109

Cryogenic Refrigerator

Input/Output

Input/Output

~ 4 K (-270 oC)

Host

CPU

Main Memory

Cache

Parallel

Accelator

PA Controller

Room

Temperature

www.iarpa.gov/index.php/research-programs/c3

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Status of Superconductor Electronics

5

Presented in this workshop:

IARPA Programs for Superconducting ComputingMarc Manheimer, IARPA

Energy efficient, high bandwidth digital data links between 4 and 300 KDr. Deborah Van Vechten, ONR

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Digital-RF Receiver (Hypres)

6

Commercial product with applications in:

Software-defined radio, satellite communications

Directly digitizes RF (no analog down-conversion)

Ultra-wide bandwidth, multi-band, multi-carrier

Hybrid temperature heterogeneous technology

Different technologies between ambient and 4 K

Closed-cycle cryogenic refrigerator48 in.

rack

Gupta, et al., IEEE Trans. Appl. Supercond., June 2011

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Quantum Annealing (D-Wave Systems)

7

D-Wave® TwoXTM (2015 August 20), a commercial superconducting quantum annealing processor

128,000 Josephson junctions

1000 qubit array

15-20 mK operating temperature

D-Wave® TwoXTM quantum annealing processor“Washington” chip

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Electronics Technology Roadmaps

- 1993-1997 NTRS: National Technology Roadmap for Semiconductors

- 1998-2013 ITRS: International Technology Roadmap for Semiconductors

• Applied Moore’s Law to integrated circuits

• Physical scaling worked until about 2004, then cores, 3D, …

• 2010: First selection of post-CMOS devices

- 2014-2015 ITRS 2.0• Driver changed from scaling to applications

• 2015: Post-CMOS map of devices

- 2016+ IRDS: International Roadmap for Devices and Systems

• Opened the door to non-semiconductortechnologies

• 2017: First roadmapsunder development

8Paolo Gargini

IRDS Chair

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

IRDS BC Chapter Organization

9

Beyond CMOS

Emerging memory and

storage devices

Emerging logicand information

processing devices

Emerging application

areas

Emerging device and architecture

interfaceAssessment

• Cryogenic electronics

• Emerging devices for hardware security

• Memory devices• Selector devices• Storage class

memory devices

• CMOS extension • Beyond-CMOS

charge-based• Beyond-CMOS

non-charge-based

• Map Emerging architecture to suitable devices

• Define FOMs and key challenges

• Define criteria• Based on the

quantitative benchmarking reported by NRI

Matt Marinella

Name of person-in-charge

Shamik Das

Scott HolmesErik DeBenedictis

Mike Niemier

Mike FrankPaul FranzonMatt MarinellaGeoff Burr

An Chen

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Energy–Delay Metrics: Wiring▸ RQL : Reciprocal Quantum Logic,

a superconducting single flux quantum (SFQ) logic, Jc = 100 µA/µm2

- JTL: Josephson transmission line(0.13 fJ/bit, 5.5 ps)

- PTL: passive transmission line(0.26 fJ/bit 0.0120 mm, 6.5 ps)

▸ 4.2 K operation; energy per bit at room temperature with 1000 W/W refrigeration(range I : 400–10,000 W/W)

▸ Source for RQL data:Dorojevets, Chen, Ayala, Kasperek, “Towards 32-bit Energy-Efficient Superconductor RQL Processors: The Cell-Level Design and Analysis of Key Processing and On-Chip Storage Units,” IEEE Trans. Appl. Supercond., 2015. doi: 10.1109/TASC.2014.2368354 (Fig. 1)+ private communication for delays

▸ Added to:Pan, Chang, Naeemi, “Performance analyses and benchmarking for spintronic devices and interconnects,” 2016 IEEE International Interconnect Technology Conference / Advanced Metallization Conference (IITC/AMC), San Jose, CA, 2016. doi: 10.1109/IITC-AMC.2016.7507679

10

RQL (PTL)

RQL (JTL)

Superconducting

Fig. 5. Comparison between CMOS and spintronic devices in terms of (a) wire energy versus delay

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Can superconducting computing compete?

11

Shadow of the Colossus/

Wander to Kyozō

SCEI

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

A better way to view the relationship

12

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

Lessons from an Emerging Technology

13

Fair metrics are needed to evaluate alternative computing technologies level the playing field to allow different technologies to compete

relevant lessons from hiring for diversity?

Ramping up requires time and resources the real Moore’s Law

Government funding alone is not sufficientcost to develop energy-efficient, large-scale computers is large

ramp up using smaller products and markets

Don’t go it alone use your mother elephant

Go big or go home! small improvements are not worth the effort

large disruptions require even larger advantages

?

?

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

References

14

D.S. Holmes, A.L. Ripple, and M.A. Manheimer, “Energy-efficient superconducting computing – power budgets and requirements”, IEEE Trans. Appl. Supercond., vol. 23, no. 3, pp. 1701610, June 2013. DOI: 10.1109/TASC.2013.2244634

D.S. Holmes, "Superconducting computing: Lessons from an emerging technology," 2015 Fourth Berkeley Symposium on Energy Efficient Electronic Systems (E3S), Berkeley, CA, 2015. DOI: 10.1109/E3S.2015.7336778Video online: https://www.youtube.com/watch?v=3Whh9VXHqOQ

M.A. Manheimer, “Cryogenic Computing Complexity Program: Phase 1 Introduction”, IEEE Trans. Appl. Supercond., vol.25, 1301704, June 2015, DOI: 10.1109/TASC.2015.2399866

D.S. Holmes, A.M. Kadin, M.W. Johnson, “Superconducting Computing in Large-Scale Hybrid Systems”, Computer, vol. 48, pp. 34-42, December 2015. DOI: 10.1109/MC.2015.375

International Roadmap for Devices and Systems (IRDS), http://irds.ieee.org/

I N T E L L I G E N C E A D V A N C E D R E S E A R C H P R O J E C T S A C T I V I T Y ( I A R P A )

2017-11, Holmes, Lessons from an Emerging Technology: Superconducting Computing

15

Computer, vol. 48, Dec. 2015