US 12,079,711 B2
Apparatus and mechanism for processing neural network tasks using a single chip package with multiple identical dies
Uday Kumar Dasari, Union City, CA (US); Olivier Temam, Antony (FR); Ravi Narayanaswami, San Jose, CA (US); and Dong Hyuk Woo, San Jose, CA (US)
Assigned to Google LLC, Mountain View, CA (US)
Filed by Google LLC, Mountain View, CA (US)
Filed on Feb. 26, 2021, as Appl. No. 17/186,598.
Application 17/186,598 is a continuation of application No. 15/819,753, filed on Nov. 21, 2017, granted, now 10,936,942.
Prior Publication US 2021/0256361 A1, Aug. 19, 2021
Int. Cl. G06N 3/063 (2023.01); G06F 7/50 (2006.01); G06F 13/16 (2006.01); G06F 13/40 (2006.01); G06F 15/78 (2006.01); G06F 17/16 (2006.01); G06N 3/04 (2023.01); G06N 3/0464 (2023.01); G06N 20/00 (2019.01); G11C 11/22 (2006.01); G11C 11/54 (2006.01); H01L 25/065 (2023.01); H01L 25/18 (2023.01)
CPC G06N 3/063 (2013.01) [G06F 15/7896 (2013.01); G06N 3/04 (2013.01); G06F 7/50 (2013.01); G06F 13/1668 (2013.01); G06F 13/4027 (2013.01); G06F 17/16 (2013.01); G06N 3/0464 (2023.01); G06N 20/00 (2019.01); G11C 11/22 (2013.01); G11C 11/54 (2013.01); H01L 25/0652 (2013.01); H01L 25/0657 (2013.01); H01L 25/18 (2013.01); H01L 2225/06541 (2013.01); H01L 2225/06589 (2013.01)] 20 Claims
OG exemplary drawing
 
1. An integrated circuit configured to process workloads of a neural network comprising a plurality of neural network layers, the integrated circuit comprising:
a plurality of processing dies, each of the plurality of processing dies being configured to process a respective portion of the workload that corresponds to one or more layers of the neural network; and
a plurality of communication paths configured to couple each of the plurality of processing dies, such that a first processing die is able to pass an output of a neural network layer to a second, different processing die.