Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

Lake Crest について調べてみた

385 views

Published on

Intel の Deep Learning 用チップ Lake Crest、というか Nervana Engine についてざっと調べてみました。
(2016年12月の学部学生向け昼休みイベントの資料)

Published in: Technology
  • Be the first to comment

Lake Crest について調べてみた

  1. 1. Intel Lake Crest Yutaka Yasuda, Kyoto Sangyo University, 2016/12/16
  2. 2. AI 2016.3 AlphaGO vs 2016.9 Google (AI) 2015 Google Photo “Google's AlphaGo AI Continues to Wallop Expert Human Go Player”, Popular Mechanics, 2016/3/10 http://www.popularmechanics.com/technology/a19863/googles-alphago-ai-wins-second-game-go/
  3. 3. Deep Learning 2014 ImageNet Google 20 2012 Google 

  4. 4. ”Deep Visual-Semantic Alignments for Generating Image Descriptions”, Andrej Karpathy, Li Fei-Fei, Stanford University, CVPR 2015
  5. 5. Neural Network = Neuron https://en.wikipedia.org/wiki/Artificial_neural_network
  6. 6. “Introduction to multi gpu deep learning with DIGITS 2”, Mike Wang
 http://www.slideshare.net/papisdotio/introduction-to-multi-gpu-deep-learning-with-digits-2-mike-wang/6
  7. 7. “Introduction to multi gpu deep learning with DIGITS 2”, Mike Wang
 http://www.slideshare.net/papisdotio/introduction-to-multi-gpu-deep-learning-with-digits-2-mike-wang/6
  8. 8. “Introduction to multi gpu deep learning with DIGITS 2”, Mike Wang
 http://www.slideshare.net/papisdotio/introduction-to-multi-gpu-deep-learning-with-digits-2-mike-wang/6
  9. 9. https://www.youtube.com/watch?v=BMEffRAvnk4
  10. 10. Why nVIDIA?
  11. 11. Lake Crest
  12. 12. Intel Artificial Intelligence Day 2016/11/17 -12:30 PM PT San Francisco
  13. 13. http://pc.watch.impress.co.jp/docs/column/ubiq/1030981.html
  14. 14. Intel Nervana Engine
  15. 15. https://www.nervanasys.com/technology/engine/
  16. 16. ASIC
  17. 17. CPU ASIC GPU ASIC Wikipedia
  18. 18. “ ASIC ” Nervana Engine Web
  19. 19. 2.5D Blazingly fast data access via high-bandwith memory (HBM)
  20. 20. Processing Cluster x12 (3x4) ICL (Inter Chip Link) x12 8GB HBM2 x4
  21. 21. HBM?
  22. 22. An Introduction to HBM - High Bandwidth Memory - Stacked Memory and The Interposer http://www.guru3d.com/articles-pages/an-introduction-to-hbm-high-bandwidth-memory,2.html • HBM DRAM • • GPU Interposer • 2.5D
  23. 23. GDDR5 HBM2 32-bit Bus With 1024-bit Up-to 1750 MHz (7 Gbps) 2 Gbps Up-to 28 GB/s per chip 125GB/s (2Tb/s) per unit 1.5V 1.3V
  24. 24. LGA 2011: CPU 2011 Xeon E5 1600/2600 v4 Broadwell-EP 2000 1024 x4 → Wikipedia: LGA 2011
  25. 25. http://pc.watch.impress.co.jp/docs/column/ubiq/1030981.html
  26. 26. Tensor
  27. 27. 
 
 https://www.tensorflow.org “TensorFlow: Large-Scale Machine Learning on Heterogeneous Distributed Systems”, Abdai, et. al, 2015, https://arxiv.org/abs/1603.04467v2
  28. 28. https://www.tensorflow.org/tutorials/mnist/beginners/ https://en.wikipedia.org/wiki/ Artificial_neural_network or CPU
  29. 29. Nervana Engine ASIC Tensor HBM2 4 unit HBM 1024bit! 2.5D
  30. 30. Nervana Engine 12 100Gbit/s 
 https://www.nervanasys.com/technology/engine/
  31. 31. 100Gbit/s *12
  32. 32. Deep Learning GPU GPU
  33. 33. GPU SIMD “ ” 
 
 http://logmi.jp/45705
  34. 34. GPU SIMD GPU 32bit 
 AI GPU nVIDIA CPU https://www.tensorflow.org/tutorials/mnist/beginners/
  35. 35. GPU Nervana Engine
  36. 36. Binary Neural Network GPU 32bit BNN - Binarized Neural Network ( -1 / +1 ) Nervana Accelerating Neural Networks with Binary Arithmetic https://www.nervanasys.com/accelerating-neural-networks-binary-arithmetic/
  37. 37. “Accelerating Neural Networks with Binary Arithmetic” (blog post) These 32 bit floating point multiplications, however, are very expensive. In BNNs, floating point multiplications are supplanted with 
 bitwise XNORs and left and right bit shifts. This is extremely attractive from a hardware perspective: binary operations can be implemented computationally efficiently at a low power cost. Nervana website (blog post) https://www.nervanasys.com/accelerating-neural-networks-binary-arithmetic/ 32bit BNN XNOR bit shift
  38. 38. Nervana Engine GPU SIMD BNN (ASIC) XNOR 
 -1 0, +1 1 Tensor
  39. 39. GPU nVIDIA
  40. 40. Intel Xeon Phi http://www.4gamer.net/games/049/G004963/20161007061/
  41. 41. Intel Nervana Engine
  42. 42. https://software.intel.com/en-us/blogs/2013/avx-512-instructions
  43. 43. Deep Learning nVIDIA GPU Deep Learning Nervana Binalized HBM2 nVIDIA FP16 Intel AVX-512 SIMD Google TPU (Tensor Processing Unit) 8bit CPU!
  44. 44. Google 

  45. 45. 
 XNOR / 
 CPU 
 100Gbps 
 SIMD
  46. 46. 'You've got to find what you love,' Jobs says Steve Jobs, 2005, Stanford University https://www.youtube.com/watch?v=UF8uR6Z6KLc “Follow your heart”

×