Apparatus and method for training device-to-device physical interface

Patent metadata

Field	Value
Publication number	US-2025094375-A1
Application number	US-202418965046-A
Country	US
Kind code	A1
Filing date	Dec 2, 2024
Priority date	Oct 19, 2020
Publication date	Mar 20, 2025
Grant date	—

Abstract

A method of training a physical interface between a first device and a second device includes performing a first training of the physical interface by communicating with the second device by using a first candidate group of lanes from among a plurality of lanes; performing a second training of the physical interface by communicating with the second device by using a second candidate group of lanes from among the plurality of lanes, the second candidate group being different from the first candidate group; determining a lane group based on a result of the first training and a result of the second training; and setting the second device so that the determined lane group is used for the physical interface.

First claim

1 .- 17 . (canceled)

18 . A method of training a physical interface between a first device and a second device, the method comprising:
performing a first training of the physical interface by communicating with the second device by using a first candidate group of lanes from among a plurality of lanes; performing a second training of the physical interface by communicating with the second device by using a second candidate group of lanes from among the plurality of lanes, the second candidate group being different from the first candidate group; determining a lane group based on a result of the first training and a result of the second training; and setting the second device so that the determined lane group is used for the physical interface, wherein the physical interface includes a high bandwidth memory (HBM) interface, and the second device includes a plurality of memory dies that are stacked.

19 . The method of claim 18 , further comprising:
selecting the second candidate group among the plurality of lanes based on the result of the first training.

20 . The method of claim 19 ,
wherein the performing of the first training comprises:
detecting a plurality of first valid window margins, each valid window margin in the plurality of first valid window margins corresponding to a lane from among the lanes in the first candidate group, and
wherein the selecting of the second candidate group comprises:
identifying a first lane corresponding to a minimum first valid window margin of the plurality of first valid window margins; and
selecting lanes from among the plurality of lanes excluding the first lane as the second candidate group.

21 . The method of claim 20 ,
wherein the performing of the second training comprises:
detecting a plurality of second valid window margins, each valid window margin in the plurality of second valid window margins corresponding to a lane from among the lanes in the second candidate group, and
wherein the determining of the lane group comprises:
identifying a second lane corresponding to a minimum second valid window margin of the plurality of second valid window margins;
identifying a higher valid window margin among the minimum first valid window margin and the minimum second valid window margin; and
selecting a candidate group corresponding to the identified valid window margin among the first candidate group and the second candidate group, as the determined lane group.

22 . The method of claim 19 , wherein the selecting of the second candidate group comprises:
identifying deactivation of at least one function of the physical interface; and selecting the second candidate group such that the second candidate group includes at least one lane that is unused as a result of the at least one function being in a deactivated state.

23 . The method of claim 22 , wherein the at least one function comprises:
at least one of a data masking function, a data bus inversion function, and an error correction function.

24 . The method of claim 18 , further comprising:
performing a third training by communicating with the second device by using a third candidate group of lanes from among the plurality of lanes, the third candidate group being different from the first candidate group and different from the second candidate group, wherein the determined lane group is determined based on a result of the first training, a result of the second training, and a result of the third training.

25 . The method of claim 18 , further comprising:
verifying lanes between the first device and the second device from among the plurality of lanes, wherein the first candidate group and the second candidate group include lanes, from among the plurality of lanes, that are determined to be valid lanes as a result of the verifying.

26 . The method of claim 18 ,
wherein the performing of the first training includes transmitting a first soft mapping command for temporarily setting the first candidate group to the second device, wherein the performing of the second training includes transmitting a second soft mapping command for temporarily setting the second candidate group to the second device, and wherein the setting of the second device includes transmitting to the second device a hard mapping command permanently setting the lane group.

27 . A first device configured to communicate with a second device based on a physical interface, the first device comprising:
a routing circuit configured to select some of a plurality of lanes based on a control signal, for the physical interface; and processing circuitry configured to,
generate the control signal for training the physical interface by using each of different candidate groups of lanes from among the plurality of lanes, and
determine a lane group to be used for the physical interface based on a result of the training
wherein the physical interface includes a high bandwidth memory (HBM) interface, and the second device includes a plurality of memory dies that are stacked.

28 . The first device of claim 27 , wherein the processing circuitry is further configured to detect a plurality of minimum valid window margins corresponding, respectively, to a plurality of trainings, and determine, as the determined the lane group, a candidate group of lanes, from among the plurality of lanes, corresponding to a highest value among the plurality of minimum valid window margins.

29 . The first device of claim 27 , wherein the processing circuitry is further configured to identify deactivation of at least one function of the physical interface, and generate the control signal for training the physical interface based on a candidate group of lanes, from among the plurality of lanes, that includes at least one lane that is unused as a result of the at least one function being in a deactivated state.

30 . The first device of claim 29 , wherein the at least one deactivated function includes at least one of a data masking function, a data bus inversion function, and an error correction function.

31 . The first device of claim 27 , wherein the processing circuitry is further configured to verify lanes between the first device and the second device from among the plurality of lanes, and perform a training for the physical interface among a candidate group that includes lanes, from among the plurality of lanes, that are determined to be valid lanes as a result of the verifying.

32 . A system comprising:
a plurality of interconnections; and a first device and a second device configured to communicate with each other via at least some of the plurality of interconnections based on a physical interface, wherein the first device and the second device are further configured to train the physical interface by using each of a plurality of different candidate groups, each candidate group including interconnections from among the plurality of interconnections, wherein first device is configured to determine an interconnection group to be used for the physical interface based on training results, wherein the physical interface includes a high bandwidth memory (HBM) interface, wherein the second device includes a plurality of memory dies that are stacked, and wherein the plurality of interconnections include at least one through silicon via penetrating at least one of the plurality of memory dies.

33 . The system of claim 32 , wherein the first device is further configured to,
detect a plurality of minimum valid window margins corresponding, respectively, to a plurality of trainings, and selec

Assignees

• Samsung Electronics Co Ltd

Inventors

• Yeo Taekyung
• Hwang Sangyun
• Kim Sujeong
• Oh Jihun
• Shin Joohee

Classifications

• G06F13/4295

  using an embedded synchronisation · CPC title

• G06N20/00

  Machine learning · CPC title

• G06F11/1479

  Generic software techniques for error detection or fault masking · CPC title

• G06F15/17

  using an input/output type connection, e.g. channel, I/O port · CPC title

• G06F13/382

  using universal interface adapter · CPC title