Location via proxy:   [ UP ]  
[Report a bug]   [Manage cookies]                
SlideShare a Scribd company logo

From R&D to Market: the industrialization challenge

•
•
Q
QCB-Conference

From R&D to Market: the industrialization challenge Maud VINET - Quantum hardware program manager, CEA Leti, France Explore the multiple challenges of the industrialization path (reliability, scaling, design, value chain) and how to overcome them. See how collective intelligence a key element for success is. Discover QuCube framework, which main objective is to demonstrate a quantum processor for simulation applications.

1 of 27
Download to read offline
ORGANIZED BY JUNE 20TH 2019 Maud VINET Quantum Hardware Program Manager CEA Leti, France From R&D to Market: the industrialization challenge
FROM R&DTO MARKET THE INDUSTRIALIZATION CHALLENGE MaudVinet Director of quantum hardware program CEA-Leti
How to benefit from the quantum speed up? Operation frequency Algo optimization Computer architecture 1 million 1 second 1 day RVanMeter, Communications of the ACM, 2013 DISAPPOINTED LONELY
QUANTUM COMPUTING
Risks What if I fail? What will my community think of me? My research is very veryinteresting Figures of merit vs applications I need to understand everything
From R&D to Market: the industrialization challenge
12 years ago, 2007 71nm Samsung 7nm Fin pitch is 30nm VLSI 2019, June 13th Kyoto What he missed is sidewall image transfer
Risks ✓ Where is the end point? ✓ Who are the competitors? ✓ Who is on my team (why do I need a team ☺)?
Semiconductor context Year of 1st introduction2014 2016 2018 2020 2022 14nm 10nm 7nm 5nm 3nm Artificial intelligence Matrix formalism HPC Low Power General purpose Computing CMOS Circuit Power management Standard cells height scaling Architecture Multicore 3D heterogeneous integration Photonic interconnects In memory computing 25nm TBOX 20nm LG ISPD SiC RSD Si channel Von Neumann Heterogeneous computing In memory computing Quantum computing FDSOI
FROM R&DTO MARKET SOMETHOUGHTS ONTHE INDUSTRIALIZATION PATH
Challenges for hardware industrialization Reliability Technology for scaling Design enablement Ecosystem & value chain
Reliability Technology for scaling Design enablement Ecosystem & value chain Phase 1 Understand variability Control and decrease variability Yield Phase 2 Phase 3 Device optimization and fabrication influenced benchmarking Components developments Components interfaces Architecture definition System integration Packaging Libraries Physical Design Kit Synthesis tools Full EDA ecosystem Friendly used interfaces Ecosystem creation Value chain identification Framework for industrialization Start-ups and licensing
Reliability Technology for scaling Design enablement Ecosystem & value chain Phase 1 Understand variability Control and decrease variability Yield Phase 2 Phase 3 Device optimization and fabrication influenced benchmarking Components developments Components interfaces Architecture definition System integration Packaging Libraries Physical Design Kit Synthesis tools Full EDA ecosystem Friendly used interfaces Ecosystem creation Value chain identification Framework for industrialization Start-ups and licensing
1st quantum revolution at work • Industry • Entrepreneurship Engineering ➢ Large scale facilities ➢ Nanotechnologies ➢ Computer sciences ➢ Mathematics ➢ Philosophy ➢ Social sciences students, researchers in fundamental quantum sciences • Create the feeling of belonging to an adventure • Leverage the physicits pessimism for risk analysis and system design Quantum physics ➢ Condensed matter ➢ Nanosciences ➢ Photonics Collective intelligence, scout for the skills
Research organizations Grenoble QuTech Netherlands (Delft) NQIT UK (Oxford) IQC Canada (Waterloo) NCCR-QSIT Switzerland (Zurich) CQCCT Australia Creation date 2019 2014 2014 2002 2011 2000 Structure Leti + fundamental research TUDelft + TNO Hub national University Waterloo IBM + universities and EPFL 6 Universities Funding Public PPP Public PPP + private patronage PPP PPP Size ~ 100 researchers ~145 researchers ~ 40 researchers ~30 researchers ~50 researchers ~100 researchers QUANTECA Grenoble • Critical mass • Skills diversity • Leverage local specialties
QuantECA Superconducting circuits Quantum optic electronics Spin qubits Cryogenic electronics Theory, modelization and simulation Fabrication technology Large scale quantum integration Q-Q interfaces
Speed up in the number of quantum algorithms
Speed up in the quality of the qubits Adapated from Schoelkopf et al
Speed up in the proposition of architectures for large scale M. Veldhorst et al. ,Nature Comm. (2017) M. Vinet et al., IEDM (2018) R. Li et al., arXiv 1711.03807 (2017) L.M.K. Vandersypen et al., npj Quant. Inf. (2017) J Gorman et al., npj Quant. Inf. (2016)
QuCubeState of the art ? Risk management
Indentifying synergies Commonalities identification Physics, Electronics, Instrumentation Algorithm Alternative paths
Si based QuCube technological developments Motivation and positioning • The main objective is to demonstrate a quantum processor for simulation applications • The novelty of is combine VLSI Si technology with quantum engineering and algorithms to find a full stack path for quantum computing • Quantum chips rely on Si spin qubits, quantum interface is based on cryoCMOS, compiler interacts at multilevel, thus saving overheads State of the art Technological modules QuCube quantum accelerator and up scaling Materials Integration for good qubits 2D tile Control electronics Optimized multi-scale compiler A programmable 100 qubit quantum accelerator
Si based QuCube technological developments Motivation and positioning • The main objective is to demonstrate a quantum processor for simulation applications • The novelty of is combine VLSI Si technology with quantum engineering and algorithms to find a full stack path for quantum computing • Quantum chips rely on Si spin qubits, quantum interface is based on cryoCMOS, compiler interacts at multilevel, thus saving overheads State of the art Technological modules QuCube quantum accelerator and up scaling Materials Integration for good qubits 2D tile Control electronics Optimized multi-scale compiler A programmable 100 qubit quantum accelerator CMOS technology Memory technology 5G Communication protocols Patterning
Cryoelectronics example COMET Programme l 28.03.2019 Short term Mid term Long term Quantum computing High performance computing Spatial applications
| 25 1 qubit 2021 2024 Gen0 Application processor 2030 Quantum simulation algorithm Logical qubit demonstration 6 entangled qubits Tackle scientific questions • 2 qubits gate • Fidelity increase • 2D tile for large scale 2018 Technological challenges • Development of technological modules for a million of qubits • First 2D array of 100 to 256 interconnected qubits Technology and biz • Demonstration of all the modules together • Architecture yield • Fabrication of a million of qubits • Value chain consolidation • Cloud accesss 100 qubits prototype Error correction Start of industrialization? QuCube silicon roadmap
GPU CPU HPC Parallel computing CLOUD/HPC Architecture CPU Processor System Tensor Flow Processor (TPU) Quantum computing QPU Applications Programming Languages And APIs Compilers Quantum error correction Critical components Qubits Arrays & integration system architecture SoftwarefocusHardwarefocus Source BostonConsulting Group High performance computing users Researchecosystem Another brain Mythic Habana Syntiant
COMET Programme l 28.03.2019

More Related Content

From R&D to Market: the industrialization challenge

  • 1. ORGANIZED BY JUNE 20TH 2019 Maud VINET Quantum Hardware Program Manager CEA Leti, France From R&D to Market: the industrialization challenge
  • 2. FROM R&DTO MARKET THE INDUSTRIALIZATION CHALLENGE MaudVinet Director of quantum hardware program CEA-Leti
  • 3. How to benefit from the quantum speed up? Operation frequency Algo optimization Computer architecture 1 million 1 second 1 day RVanMeter, Communications of the ACM, 2013 DISAPPOINTED LONELY
  • 5. Risks What if I fail? What will my community think of me? My research is very veryinteresting Figures of merit vs applications I need to understand everything
  • 7. 12 years ago, 2007 71nm Samsung 7nm Fin pitch is 30nm VLSI 2019, June 13th Kyoto What he missed is sidewall image transfer
  • 8. Risks ✓ Where is the end point? ✓ Who are the competitors? ✓ Who is on my team (why do I need a team ☺)?
  • 9. Semiconductor context Year of 1st introduction2014 2016 2018 2020 2022 14nm 10nm 7nm 5nm 3nm Artificial intelligence Matrix formalism HPC Low Power General purpose Computing CMOS Circuit Power management Standard cells height scaling Architecture Multicore 3D heterogeneous integration Photonic interconnects In memory computing 25nm TBOX 20nm LG ISPD SiC RSD Si channel Von Neumann Heterogeneous computing In memory computing Quantum computing FDSOI
  • 10. FROM R&DTO MARKET SOMETHOUGHTS ONTHE INDUSTRIALIZATION PATH
  • 11. Challenges for hardware industrialization Reliability Technology for scaling Design enablement Ecosystem & value chain
  • 12. Reliability Technology for scaling Design enablement Ecosystem & value chain Phase 1 Understand variability Control and decrease variability Yield Phase 2 Phase 3 Device optimization and fabrication influenced benchmarking Components developments Components interfaces Architecture definition System integration Packaging Libraries Physical Design Kit Synthesis tools Full EDA ecosystem Friendly used interfaces Ecosystem creation Value chain identification Framework for industrialization Start-ups and licensing
  • 13. Reliability Technology for scaling Design enablement Ecosystem & value chain Phase 1 Understand variability Control and decrease variability Yield Phase 2 Phase 3 Device optimization and fabrication influenced benchmarking Components developments Components interfaces Architecture definition System integration Packaging Libraries Physical Design Kit Synthesis tools Full EDA ecosystem Friendly used interfaces Ecosystem creation Value chain identification Framework for industrialization Start-ups and licensing
  • 14. 1st quantum revolution at work • Industry • Entrepreneurship Engineering ➢ Large scale facilities ➢ Nanotechnologies ➢ Computer sciences ➢ Mathematics ➢ Philosophy ➢ Social sciences students, researchers in fundamental quantum sciences • Create the feeling of belonging to an adventure • Leverage the physicits pessimism for risk analysis and system design Quantum physics ➢ Condensed matter ➢ Nanosciences ➢ Photonics Collective intelligence, scout for the skills
  • 15. Research organizations Grenoble QuTech Netherlands (Delft) NQIT UK (Oxford) IQC Canada (Waterloo) NCCR-QSIT Switzerland (Zurich) CQCCT Australia Creation date 2019 2014 2014 2002 2011 2000 Structure Leti + fundamental research TUDelft + TNO Hub national University Waterloo IBM + universities and EPFL 6 Universities Funding Public PPP Public PPP + private patronage PPP PPP Size ~ 100 researchers ~145 researchers ~ 40 researchers ~30 researchers ~50 researchers ~100 researchers QUANTECA Grenoble • Critical mass • Skills diversity • Leverage local specialties
  • 16. QuantECA Superconducting circuits Quantum optic electronics Spin qubits Cryogenic electronics Theory, modelization and simulation Fabrication technology Large scale quantum integration Q-Q interfaces
  • 17. Speed up in the number of quantum algorithms
  • 18. Speed up in the quality of the qubits Adapated from Schoelkopf et al
  • 19. Speed up in the proposition of architectures for large scale M. Veldhorst et al. ,Nature Comm. (2017) M. Vinet et al., IEDM (2018) R. Li et al., arXiv 1711.03807 (2017) L.M.K. Vandersypen et al., npj Quant. Inf. (2017) J Gorman et al., npj Quant. Inf. (2016)
  • 20. QuCubeState of the art ? Risk management
  • 21. Indentifying synergies Commonalities identification Physics, Electronics, Instrumentation Algorithm Alternative paths
  • 22. Si based QuCube technological developments Motivation and positioning • The main objective is to demonstrate a quantum processor for simulation applications • The novelty of is combine VLSI Si technology with quantum engineering and algorithms to find a full stack path for quantum computing • Quantum chips rely on Si spin qubits, quantum interface is based on cryoCMOS, compiler interacts at multilevel, thus saving overheads State of the art Technological modules QuCube quantum accelerator and up scaling Materials Integration for good qubits 2D tile Control electronics Optimized multi-scale compiler A programmable 100 qubit quantum accelerator
  • 23. Si based QuCube technological developments Motivation and positioning • The main objective is to demonstrate a quantum processor for simulation applications • The novelty of is combine VLSI Si technology with quantum engineering and algorithms to find a full stack path for quantum computing • Quantum chips rely on Si spin qubits, quantum interface is based on cryoCMOS, compiler interacts at multilevel, thus saving overheads State of the art Technological modules QuCube quantum accelerator and up scaling Materials Integration for good qubits 2D tile Control electronics Optimized multi-scale compiler A programmable 100 qubit quantum accelerator CMOS technology Memory technology 5G Communication protocols Patterning
  • 24. Cryoelectronics example COMET Programme l 28.03.2019 Short term Mid term Long term Quantum computing High performance computing Spatial applications
  • 25. | 25 1 qubit 2021 2024 Gen0 Application processor 2030 Quantum simulation algorithm Logical qubit demonstration 6 entangled qubits Tackle scientific questions • 2 qubits gate • Fidelity increase • 2D tile for large scale 2018 Technological challenges • Development of technological modules for a million of qubits • First 2D array of 100 to 256 interconnected qubits Technology and biz • Demonstration of all the modules together • Architecture yield • Fabrication of a million of qubits • Value chain consolidation • Cloud accesss 100 qubits prototype Error correction Start of industrialization? QuCube silicon roadmap
  • 26. GPU CPU HPC Parallel computing CLOUD/HPC Architecture CPU Processor System Tensor Flow Processor (TPU) Quantum computing QPU Applications Programming Languages And APIs Compilers Quantum error correction Critical components Qubits Arrays & integration system architecture SoftwarefocusHardwarefocus Source BostonConsulting Group High performance computing users Researchecosystem Another brain Mythic Habana Syntiant
  • 27. COMET Programme l 28.03.2019