IP Library › Granted Patent US 12,581,663
Granted Patent B2
US 12,581,663 · App. 18/087,384 · Granted Mar 17, 2026

Heterogeneous integration structure with voltage regulation

Inventors: Mukta Ghate Farooq (Hopewell Jct, NY); Arvind Kumar (Chappaqua, NY)
Assignee: International Business Machines Corporation
H10B80/00H01L23/481H01L23/528H01L23/5386H01L24/02H01L24/08H01L25/18H01L24/16H01L2224/02372H01L2224/02373H01L2224/02381H01L2224/08145H01L2224/16225
View Patent ↗
Loading inventors, assignments & file history…
Monitor This Case
Get email alerts when status or documents change.
Order Certified Copies
Most orders are placed with the USPTO same day — all within 24 business hours.
Order via The Patent Place →
Pre-filled with this patent's details
Quick Facts
Patent No.
US 12,581,663
App. No.
18/087,384
Granted
Mar 17, 2026
Kind
B2
Abstract

Heterogeneous integration semiconductor packages with voltage regulation are described. A semiconductor device can include a chip including a memory device and a plurality of through-silicon-vias (TSVs). The semiconductor device can further include a processor arranged on top of the chip. The processor can be configured to communicate with the memory device via a plurality of interconnects. The semiconductor device can further include at least one voltage regulator arranged on top of the chip. The at least one voltage regulator can be configured to regulate power being provided from the plurality of TSVs to the processor.

Claims (36)

1 . A semiconductor device comprising:

a chip including a memory device and a plurality of through-silicon-vias (TSVs);

a processor arranged on top of the chip, wherein the processor is configured to communicate with the memory device via a plurality of interconnects;

at least one voltage regulator arranged on top of the chip, the at least one voltage regulator being configured to regulate power being provided from the plurality of TSVs to the processor; and

at least one optical chip configured to transfer regulated voltage from the at least one voltage regulator to the processor.

2 . The semiconductor device of claim 1 , wherein the at least one voltage regulator includes:

a first voltage regulator configured to regulate power being provide from the plurality of TSVs to the processor; and

a second voltage regulator configured to regulate power being provide from the plurality of TSVs to the memory device.

3 . The semiconductor device of claim 1 , wherein the at least one voltage regulator is configured to output regulated power to the processor via back-end-of-line (BEOL) wires.

4 . The semiconductor device of claim 1 , wherein:

the chip is arranged on top of a substrate;

the semiconductor device further includes another memory device arranged on top of the substrate; and

a memory controller is embedded in the chip, the memory controller being configured to perform memory operations between the memory device and said another memory device.

5 . The semiconductor device of claim 1 , wherein the chip and the processor are connected in a face-to-face (F2F) configuration.

6 . The semiconductor device of claim 1 , wherein the plurality of TSVs are arranged on a periphery of the chip.

7 . The semiconductor device of claim 1 , wherein the memory device in the chip is a three-dimensional (3D) stacked memory device.

8 . A semiconductor device comprising:

a chip arranged on top of a substrate, the chip including a memory device and a plurality of through-silicon-vias (TSVs), the memory device being configured to store data associated with an artificial intelligence application;

an accelerator arranged on top of the chip, wherein the accelerator is configured to communicate with the memory device via a plurality of interconnects, and the accelerator is configured to use data stored in the memory device to perform a specific task for the artificial intelligence application;

at least one voltage regulator arranged on top of the chip, the at least one voltage regulator being configured to regulate power being provided from the plurality of TSVs to the accelerator; and

at least one optical chip configured to transfer regulated voltage from the at least one voltage regulator to the accelerator.

9 . The semiconductor device of claim 8 , wherein the at least one voltage regulator includes:

a first voltage regulator configured to regulate power being provide from the plurality of TSVs to the accelerator; and

a second voltage regulator configured to regulate power being provide from the plurality of TSVs to the memory device.

10 . The semiconductor device of claim 8 , wherein the at least one voltage regulator is configured to output regulated power to the accelerator via back-end-of-line (BEOL) wires.

11 . The semiconductor device of claim 8 , further comprising another memory device arranged on top of the substrate, wherein a memory controller is embedded in the chip, and the memory controller is configured to perform memory operations between the memory device and said another memory device.

12 . The semiconductor device of claim 8 , wherein the chip and the accelerator are connected in a face-to-face (F2F) configuration.

13 . The semiconductor device of claim 8 , wherein the plurality of TSVs are arranged on a periphery of the chip.

14 . The semiconductor device of claim 8 , wherein the memory device in the chip is a three-dimensional (3D) stacked memory device.

15 . A method comprising:

forming a plurality of TSVs on a frame of a memory power chip, wherein the memory power chip includes the frame and a memory device;

attaching the memory power chip with the plurality of TSVs on top of a substrate;

attaching a processor and at least one voltage regulator to a top surface of the memory power chip; and

attaching at least one optical chip to the top surface of the memory power chip.

16 . The method of claim 15 , wherein the processor is an accelerator configured to perform a specific task of an artificial intelligence application using data stored in the memory device.

17 . The method of claim 15 , wherein attaching the processor to the top surface of the memory power chip comprises attaching a face of the processor to a face of the memory device of the memory power chip.

Assignments (1)
ASSIGNMENT OF ASSIGNOR'S INTEREST Recorded Dec 22, 2022
From: FAROOQ, MUKTA GHATE; KUMAR, ARVIND
To: INTERNATIONAL BUSINESS MACHINES CORPORATION
Reel/Frame 062188/0898 →
Continuity (1)
Related Publication 20240215270A1 · Jun 27, 2024
References Cited (29)
US 8053873B2 · Chauhan et al. · 2011 [cited by applicant]
US 8692368B2 · Pan et al. · 2014 [cited by applicant]
US 8982563B2 · Raj et al. · 2015 [cited by applicant]
US 9048112B2 · Pan et al. · 2015 [cited by applicant]
US 9101068B2 · Yun et al. · 2015 [cited by applicant]
US 9229466B2 · Saraswat et al. · 2016 [cited by applicant]
US 9831148B2 · Yu et al. · 2017 [cited by applicant]
US 10096582B2 · Mantiply et al. · 2018 [cited by applicant]
US 11211378B2 · Farooq et al. · 2021 [cited by applicant]
US 11380652B2 · Choi et al. · 2022 [cited by applicant]
US 20110109382A1 · Jin et al. · 2011 [cited by applicant]
US 20150242308A1 · Kim et al. · 2015 [cited by applicant]
US 20160379686A1 · Burger et al. · 2016 [cited by applicant]
US 20210335718A1 · Cheah · 2021 [cited by examiner]
US 20220278067A1 · Tseng · 2022 [cited by examiner]
US 20230094979A1 · Aleksov · 2023 [cited by examiner]
US 20230107103A1 · Youn · 2023 [cited by examiner]
US 20230197619A1 · Loh · 2023 [cited by examiner]
US 20230207544A1 · Loh · 2023 [cited by examiner]
US 20230253381A1 · Jeong · 2023 [cited by examiner]
US 20230260965A1 · Li · 2023 [cited by examiner]
US 20230268288A1 · Tong · 2023 [cited by examiner]
US 20240023239A1 · Kang · 2024 [cited by examiner]
US 20240038721A1 · Chen · 2024 [cited by examiner]
US 20240113071A1 · Liu · 2024 [cited by examiner]
CN 114121891A · 2022 [cited by examiner]
D. Kim, J. Kung, S. Chai, S. Yalamanchili and S. Mukhopadhyay, “Neurocube: A Programmable Digital Neuromorphic Architecture with High-Density 3D Memory,” 2016 ACM/IEEE 43rd Annual International Symposium on Computer Arc… [cited by applicant]
Gao et al. “TETRIS: Scalable and Efficient Neural Network Acceleration with 3D Memory,” ACM SIGARCH Computer Architecture News, vol. 45, Issue 1, Mar. 2017, pp. 751-764 https://doi.org/10.1145/3093337.3037702. [cited by applicant]
Hargrove, MJ et al., “Masterslice Logic Chip with Multilevel Power Circuits” Parent, RM; IBM TDB 06-80 p. 115-117; IP.com Electronic Publication Date: Feb. 13, 2005, 3 pages. [cited by applicant]