1. G-Number 1 Challenges and Opportunities for Technion and the Israeli Chip Industry Avinoam Kolodny EE Department Technion – Israel Institute of Technology July 2010

2. G-Number 2

3. 3 EE CS

4. 4 Thoughts about Technion and the Israeli chip industry Past Past Present: current research challenges Present: current research challenges Future: opportunities and directions Future: opportunities and directions

5. G-Number 5 My professional career: between Technion and Intel My professional career: between Technion and Intel The Past - from a Personal Viewpoint

6. G-Number 6 1975: I graduated EE

7. G-Number 7 Got a part-time job at Intel I was the first student employed by Intel Israel I was the first student employed by Intel Israel

8. G-Number 8 Why did Intel Hire Me? Because there was no computer at Intel! Because there was no computer at Intel! They needed to run simulations on the Technion’s computer They needed to run simulations on the Technion’s computer

9. G-Number 9 First thing I did for Intel I ported Intel’s simulator to the big IBM machine at the Technion computer center I ported Intel’s simulator to the big IBM machine at the Technion computer center

10. G-Number 10 First trip to Intel Santa Clara Intel Israel bought a minicomputer Intel Israel bought a minicomputer They wanted me to go get the software They wanted me to go get the software I was still a Technion graduate student… I was still a Technion graduate student… … I asked my advisor… … I asked my advisor…

11. G-Number 11

12. G-Number 12 Technion’s Microelectronics Infrared focus Research Goal

13. G-Number 13 Post Doc at Intel’s Non-Volatile Memory Group In the 1980s, Intel moved from memories to processors In the 1980s, Intel moved from memories to processors

14. G-Number 14 Design Automation for Processors Logic simulation RTL modeling Timing Analysis Cell libraries Logic Synthesis

15. G-Number 15 Synthesis-Based Chip Design Logic designers became programmers Logic designers became programmers Key for the ASIC business and “fabless” companies Key for the ASIC business and “fabless” companies Israel became “Land of chips and software” with synthesis Israel became “Land of chips and software” with synthesis The 486 Processor

16. G-Number 16 1988 2009 A lot of the CAD work was done in Haifa, by research collaboration with Berkeley…

17. G-Number 17 Intel’s CAD History

18. G-Number 18 Contribution to Education

19. G-Number 19 Back to Technion

20. G-Number 20 Observation about the Past It was all implied by Moore’s law: It was all implied by Moore’s law: – More and more transistors – From memories to processors to systems – Digital revolution: changed the world –New ways to live, learn and teach

21. G-Number 21 From Past to Present

22. G-Number 22 Technion and the Israeli Chip Industry Technion Israeli Chip IndustryPast Present

23. G-Number 23

24. G-Number 24 Present Challenges - Reflected in Current Research Interconnect Interconnect Energy efficiency Energy efficiency Programming of many-core systems Programming of many-core systems Correctness and security Correctness and security

25. G-Number 25 Connectivity and Complexity

26. G-Number 26 Adding more metal layers… Wires dominate: Wires dominate: –Speed –Power –Noise –Reliability –Cost

27. G-Number 27 Intel’s Pentium-M, low-power microprocessor designed in Haifa, 0.13 micron CMOS Intel’s Pentium-M, low-power microprocessor designed in Haifa, 0.13 micron CMOS Bit-Transportation energy is larger than computation energy!!! Bit-Transportation energy is larger than computation energy!!! Interconnect Power: 2004 case study * N. Magen, A. Kolodny, U. Weiser and N. Shamir, “Interconnect-Related Energy dissipation in a Low-Power Microprocessor”, Proc. SLIP, 2004. Interconnect 51 % Gate 34% Diffusion 15%

28. G-Number 28 Bits as Cars on the Information highway Energy efficiency in chip design Energy efficiency in chip design

29. G-Number 29 As System Complexity Goes Up Power Efficiency Goes Down Power-efficiency requires many parallel local computations Power-efficiency requires many parallel local computations –Chip Multi Processors (CMP) –Thread-Level Parallelism (TLP) Die Area (or Power) Uniprocessor Performance (F. Pollack. Micro 32, 1999)

30. G-Number 30 Possible future paradigm shift in VLSI: Network on Chip (NoC) From: Dedicated signal wires to: Shared network From: Dedicated signal wires to: Shared network Computing Resource Network switch Network link * E. Bolotin, I. Cidon, R. Ginosar and A. Kolodny., “QNoC: QoS architecture and design process for Network on Chip”, JSA special issue on NoC, 2004.

31. G-Number 31 Characteristics of a Paradigm Shift Solves a critical problem (or several problems) Solves a critical problem (or several problems) Step-up in abstraction Step-up in abstraction Design is affected: Design is affected: –Design becomes more restricted –New tools –The changes enable higher complexity and capacity –Jump in design productivity Initially: skepticism. Finally: change of mindset! Initially: skepticism. Finally: change of mindset! successful

32. G-Number 32 Driving Forces Behind NoC 3) Chip Multi Processors (key to power-efficient computing) 1) Global interconnect design problem: delay, power, noise, scalability, reliability 2) System integration productivity problem

33. G-Number 33 Emergent VLSI Systems R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R R Highly regular Classes of replicated cores E.g. DSP, HW accelerators, Cache banks NoC Different clocks & voltages

34. G-Number 34 Memory Access in a Multiprocessor A small number of data items, shared by many processors, is accessed numerous times What can be done better? Bring processors closer to the shared data! Bring processors closer to the shared data! * Z. Guz, I. Keidar, A.. Kolodny, U. Weiser, “Utilizing Shared Data in Chip Multiprocessors with the Nahalal Architecture”, BEST PAPER, SPAA 2008

35. G-Number 35 Aerial view of Nahalal cooperative village P0 P1 P2 P3 P4 P5 P6 P7 This Has Been Addressed Before Overview of Nahalal cache organization

36. G-Number 36 From Present to Future

37. G-Number 37 Beyond 10 Years Ahead: Future Challenges and Opportunities MOS Device Scaling will slow down MOS Device Scaling will slow down –New ways for system integration – Back to basic sciences – Interdisciplinary solutions – New architectures

38. G-Number 38 Long Term Issues for the Industry Flat world Flat world – Competition with Asia Imperatives for Technion and the industry Imperatives for Technion and the industry – Israel can win by quality and innovation – Education: more Ph.D. graduates – Strong research collaboration with Technion in computer engineering

39. G-Number 39 Mythical Academia-Industry Mismatch Industry Industry – Working results – Secret IP – Tight schedule – Low Risk – Control – Money Academia Academia – Paper results – Publish – Slow – High risk – Freedom – Fame

40. G-Number 40 The Innovation Paradox - Applied Creativity in Organizations The Innovation Paradox - Applied Creativity in Organizations * F. Chu, A. Kolodny, S. Maital and D. Perlmutter, “The innovation paradox: Reconciling creativity & discipline – How winning organizations combine inspiration with perspiration,” Proceedings of IEEE International Engineering Management Conference, October 2004. Dadi Perlmutter (Intel( Dadi Perlmutter (Intel( Shlomo Maital (TIM) Shlomo Maital (TIM) Avi Kolodny (Technion( Avi Kolodny (Technion(

41. G-Number 41 The Innovation Paradox Creativity and project implementation often require contradictory mindsets

42. G-Number 42 Need Both: Creativity and Organization The conflicting values must be reconciled! Not by tradeoff, by synergy! Not by tradeoff, by synergy!CreativityDiscipline DisciplineCreativity We found several methods for doing this We found several methods for doing this

43. G-Number 43 “Head in the Clouds, Feet on the Ground” Creative, wild brainstorming Creative, wild brainstorming –No idea is rejected (“head in the clouds”) Detailed evaluation of ideas Detailed evaluation of ideas –By practicality and benefit (“feet on the ground”) Let’s apply this for academia-industry joint research! Let’s apply this for academia-industry joint research!

44. G-Number 44 Past vs. Future for Technion and the Industry – Summary of Opportunities

45. G-Number 45 1) Research Partnership Past: Technion enabled Israeli hi-tech Past: Technion enabled Israeli hi-tech – Mostly as an engineering school – Partly as a research institute Future: Industry will need research to innovate Future: Industry will need research to innovate – More Ph.D.’s will be needed – Research collaboration essential

46. G-Number 46 2) Industry-Funded Research Past: Defense technology was the driver Past: Defense technology was the driver –Defense-funded research –Adapted later for commercial use Future: Commercial developments lead Future: Commercial developments lead –Consumer products are adapted for defense –Industry must fund academic research

47. G-Number 47 3) Organization / Culture Past: Past: –Technion focus on theory and individual research – Industry focus on practical short term results Future: Future: – Joint research in computer engineering – Long and short terms – Teamwork

48. G-Number 48 A Specific Plan: Computer Engineering Research Center

49. G-Number 49 Hardware Physics Electronic devices Computer hardware High Performance Computing Need for integrated view: Hardware and software Software Theory Algorithms Software

50. G-Number 50 The center as a focal point for collaboration Research in Israel, centered at the Technion, including Academia, R&D from Hi-Tech industry, as well as R&D from leading international companies (Microsoft, Intel, Google, etc.) Create a focal point

51. G-Number 51 Conclusion The Technion gave invaluable leverage for the past success of: The Technion gave invaluable leverage for the past success of: – Israel – Israeli hi-tech industry – Individual Alumni Let’s work together towards mutual success in the future! Let’s work together towards mutual success in the future!

52. G-Number 52