Aquantia Corporation

Marvell Asia Pte., Ltd.

Eliyan Corporation

Other

Rambus Inc.

Ramin Farjadrad

Ramin Shirani

Paul Langner

Hossein Sedarat

Syrus Ziai

Kamal Dalmia

Seid Alireza Razavi Majomard

Kevin Donnelly

Curtis McAllister

Dance Wu

Shaoan Dai

Xing Wu

Fadi Saibi

Michael A Brown

Saied Benyamin

Dragan Labalo

Wensheng Sun

Ali Tabatabaei

Saman Behtash

Mohit Singh

Jerry Martinson

Thomas Wayne Gandy

Ali Khoshniat

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a multi-chip module (MCM) is disclosed. The MCM includes an active silicon substrate and a memory interface circuit configured to support N memory channels. The memory interface circuit has a primary interface for coupling to a host memory interface via the N memory channels. A first HBM stack of memory die is disposed on the active silicon substrate and coupled to a secondary interface of the memory interface circuit. The first HBM stack dedicated to a first subset of the N data channels and a first data transfer rate. A second HBM stack of memory die is disposed on the active silicon substrate. The second HBM stack is positioned inline with the first HBM stack and the memory interface circuit and coupled to the secondary interface of the memory interface circuit. The second HBM stack is dedicated to a second subset of the N data channels and exhibits a second data transfer rate. The first HBM stack and the second HBM stack are configured to collectively support the N channels and exhibit an aggregate data rate that is a sum of the first data rate and the second data rate.

Multi-chip module (MCM) with scalable high bandwidth memory

A physical layer interconnect between chips/chiplets provides high bandwidth communication with low power requirements using an organic substrate such as a printed circuit board. An interface between first and second chiplets uses a separate chiplet, or a combination logic die and interconnect interface, interfacing with the interconnect. A connection between a computing device and a memory can be longer, allowing the computing device to be coupled to more memories, expansion slots, or external connections. The interconnect can route memory commands between computing devices and memories, allowing multiple and different computing devices to be coupled to each other or to multiple and different memories. The memories can perform in-memory computing using chiplets coupled thereto. The interconnect couples to possibly different computing devices and possibly different memories, such as in a rack configuration, including CPUs or GPUs. The specialized processing devices can include one or more TPUs or VPUs.

Complex system-in-package architectures leveraging high-bandwidth long-reach die-to-die connectivity over package substrates

A multi-chip package includes first and second groups of integrated circuit (IC) chips and a transfer IC chip disposed in the multi-chip package. The transfer IC chip is communicatively interposed between the first and second groups of IC chips and is configured to transfer signals from at least a first IC chip of the first group of IC chips to at least a second IC chip of the second group of IC chips or an output interface. The output interface is configured to output first data from the multi-chip package. A first set of ultra-short reach (USR) signaling links connects the first group of IC chips to the transfer IC chip. A second set of USR signaling links connects the second group of IC chips to the transfer IC chip. Each of the USR signaling links comprises a trace length of less than one inch.

Efficient signaling scheme for high-speed ultra short reach interfaces

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a wireline signaling link is disclosed. The wireline signaling link includes a first transceiver circuit disposed on a first integrated circuit (IC) chip. The first transceiver circuit includes first transmit circuitry to transmit a first set of symbols representing first data, each symbol having a symbol time. The first transceiver circuit also includes first receiver circuitry to receive a second set of symbols representing second data. A second transceiver circuit is disposed on a second IC chip. The wireline signaling link includes a bidirectional channel having a first end coupled to the first transceiver circuit and a second end coupled to the second transceiver circuit. The wireline signaling link further includes circuitry to make an adjustment of an electrical delay of the wireline signaling link, the adjustment based on a phase relationship between interference associated with the first data that is received by the first receiver circuitry and the second data that is received by the first receiver circuitry.

Wireline link with crosstalk reduction based on controlled channel delay

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a chiplet-based multi-chip module (MCM) is disclosed. The chiplet-based MCM includes a package substrate and a first processing integrated circuit (IC) chiplet disposed on the package substrate. A second processing IC chiplet is disposed on the package substrate. At least one first row multi-port memory chiplet is disposed on the package substrate and coupled between the first processing IC chiplet and the second IC processing chiplet. Data transfers between the first processing IC chiplet and a destination IC chiplet are routed through the at least one first row multi-port memory chiplet.

Multi-chip module (MCM) with multi-port unified memory

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a chiplet-based multi-chip module (MCM) is disclosed. The chiplet-based MCM includes a first integrated circuit (IC) chiplet comprising a first interface to receive a first set of information-carrying signals associated with a memory access operation. Conversion circuitry generates a second set of support signals associated with the memory access operation. The conversion circuitry aggregates the second set of support signals with the first set of information-carrying signals to generate an aggregate set of signals associated with the memory access operation. A second interface transmits the aggregate set of signals. Memory, comprising a memory interface coupled to the second interface of the first IC chiplet receives the aggregate set of signals.

Chiplet gearbox for low-cost multi-chip module applications

An integrated circuit (IC) chip includes a Universal Chiplet Interconnect express (UCIe) interface circuit for transferring UCIe signals. The UCIe interface circuit includes a mainband sub-interface for transferring mainband signals and a sideband sub-interface for transferring sideband signals along a first number of sideband signal paths. A bump interface includes a second number of sideband bumps, each of the sideband bumps for coupling to a signal link. The second number of sideband bumps is less than the first number of sideband signal paths. A converter circuit is disposed between the UCIe interface circuit and the bump interface. The converter circuit includes a receiver circuit to receive first sideband data from the sideband sub-interface. The receiver circuit includes local clock circuitry, oversampling circuitry, and majority detection circuitry to receive oversampled data and to resolve states of sideband data bits based on a majority voting process.

Interface conversion circuitry for universal chiplet interconnect express (UCIe)

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a multi-chip module (MCM) is disclosed. The MCM includes a common substrate and a first integrated circuit (IC) chip disposed on the common substrate. The first IC chip includes a first memory interface. A second IC chip is disposed on the common substrate and includes a second memory interface. A first memory device is disposed on the common substrate and includes memory and a first port coupled to the memory. The first port is configured for communicating with the first memory interface of the first IC chip. A second port is coupled to the memory and communicates with the second memory interface of the second IC chip. In-memory processing circuitry is coupled to the memory and controls transactions between the first memory device and the first and second IC chips.

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a memory chiplet is disclosed. The memory chiplet includes memory control circuitry for controlling at least one memory die of a first memory type. A die-to-die (D2D) interface circuit couples to a host IC chiplet. The D2D interface circuit includes input/output (I/O) circuitry for coupling to multiple signaling lanes for transferring information packets with the host IC chiplet.

Universal memory interface utilizing die-to-die (D2D) interfaces between chiplets

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a memory chiplet is disclosed. The memory chiplet includes a D2D interface of a first type for coupling to a host IC chip via multiple lanes. The D2D interface includes multiple unit interface modules, each of the multiple unit interface modules corresponding to a first set of signal path resources of a lowest granularity provided by the multiple lanes. A memory port includes a memory physical interface of a first memory type for accessing memory storage of the first memory type. The memory physical interface of the first memory type includes a second set of signal path resources corresponding to multiple memory channels of the first memory type. Mapping circuitry maps the second set of signal path resources to the first set of signal path resources in a manner that utilizes all of the first signal path resources for an integer number of the multiple unit interface modules.

Memory chiplet with efficient mapping of memory-centric interface to die-to-die (D2D) unit interface modules

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a logic base die for inclusion in a chiplet-based multi-chip module (MCM) is disclosed. The logic base die includes a first port interface for coupling to an interface beachfront of a first integrated circuit (IC) chiplet. The first port interface is to receive data of a first bandwidth via the interface beachfront via a first set of traces formed in a micro-bump advanced-package routing layer. A memory port provides a first portion of the first bandwidth to at least a first memory stack configured for positioning on the logic base die. A second port interface couples to the first port interface and utilizes a second portion of the first bandwidth.

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a multi-chip module (MCM) is disclosed. The MCM includes a package substrate and an interposer disposed on a portion of the package substrate. A first integrated circuit (IC) chip is disposed on the interposer. A first memory device is disposed on the interposer and includes a first port interface including an interposer-compliant mechanical interface for coupling to the first IC chip via a first set of traces formed in the interposer. A second port interface includes a non-interposer-compliant mechanical interface for coupling to an off-interposer device. Transactions between the first IC chip and the off-interposer device pass through the first port interface and the second port interface of the first memory device.

Semiconductor devices, packaging architectures and associated methods are disclosed. In one embodiment, a multi-chip module (MCM) is disclosed. The MCM includes a package substrate and an integrated circuit (IC) processor chip disposed on the package substrate. The IC processor chip includes a data interface configured to support N channels. A scalable high bandwidth memory (HBM) is coupled to the IC processor chip. The scalable HBM includes a first HBM device disposed on the package substrate with a first primary data interface that supports a first set of N/2 data channels and a first data transfer rate. A second HBM device is disposed on the package substrate and supports a second set of N/2 data channels and a second data transfer rate. The first HBM device and the second HBM device are configured to collectively support the full N channels and an aggregate data rate that is a sum of the first data rate and the second data rate.

Growing community of inventors

Los Altos, CA, United States of America

Ramin Farjadrad