This relates generally to integrated circuits, and more particularly, to integrated circuit components for supporting high bandwidth communication.
Computing systems such as cloud computing systems or data centers that are used for hosting, storing, or conveying large amounts of data typically include many high-performance computing devices interconnected to one another. A typical computing device includes a printed circuit board, an integrated circuit die mounted on the printed circuit board and a separate optical module mounted on the printed circuit board. An external optical cable is connected to the optical module to connect the integrated circuit die to the rest of the system.
Data is transferred between the integrated circuit die and the optical module using package traces formed on the printed circuit board. State of the art computing systems may have high bandwidth requirements, with communications exceeding 10 Gigabit per second (Gbps), 100 Gbps, or even 400 Gbps, either per lane, or aggregated. The package traces on the printed circuit board connecting the integrated circuit die to the optical module are, however, not optimized for high bandwidth density and low power. As a result, power dissipation poses a significant problem with such types of package-level interconnects (i.e., a large percentage of power in data centers, wireless applications, and other high-performance computing systems are due to interconnect power).
It is within this context that the embodiments described herein arise.
The present embodiments relate to integrated circuits, and more particularly, to a multichip package that includes a main integrated circuit die and an electro-optical tile. The electro-optical tile may include a transceiver and an optical engine, which eliminates the conventional chip-to-module interconnect. Configured in this way, the power consumption can be greatly reduced while supporting high bandwidth interconnect density.
It will be recognized by one skilled in the art, that the present exemplary embodiments may be practiced without some or all of these specific details. In other instances, well-known operations have not been described in detail in order not to unnecessarily obscure the present embodiments.
An illustrative system 100 of interconnected electronic devices is shown in
Interconnection resources 102 may include conductive lines and busses, optical interconnect infrastructure, and/or wired and wireless networks with optional intermediate switching circuitry may be used to send signals from one electronic device to another electronic device or to broadcast information from one electronic device to multiple other electronic devices. For example, a transmitter in device B may transmit serialized data signals at a given transmission rate as a data stream over a serial communication link 102 to a receiver in device C. Similarly, device C may use a transmitter to transmit serialized data signals as a data stream over a serial communication link 102 to a receiver in device B.
If desired, multiple serial communication links may be used to transmit data. For example, multiple transmitters in a transmitting device may each transmit a portion of the data as serial data streams over multiple serial communication links or “channels” to multiple receivers in a receiving device. Upon reception, receiver circuitry in the receiving device may restore the data by aggregating portions from the different channels received at the multiple receivers. The aggregated data may then be stored by memory circuit on the receiving device or processed and retransmitted to another device.
As integrated circuit fabrication technology scales towards smaller processing nodes, it becomes increasingly challenging to design an entire system on a single integrated circuit die (sometimes referred to as a system-on-chip). Designing analog and digital circuitry to support desired performance levels while minimizing leakage and power consumption can be extremely time consuming and costly.
One alternative to single-die packages is an arrangement in which multiple dies are placed within a single package. Such types of packages that contain multiple interconnected dies may sometimes be referred to as systems-in-package (SiPs), multichip modules (MCM), or multichip packages (MCP). Placing multiple chips (dies) into a single package may allow each die to be implemented using the most appropriate or optimal technology process (e.g., a core logic chip may be implemented using one technology node, whereas a memory chip may be implemented using another technology node) and may help increase the performance of die-to-die interface (e.g., driving signals from one die to another within a single package is substantially easier than driving signals from one package to another, thereby reducing power consumption of associated input-output buffers), may free up input-output pins (e.g., input-output pins associated with die-to-die connections are much smaller than pins associated with package-to-board connections), and may help simplify printed circuit board (PCB) design (i.e., the design of the PCB on which the multi-chip package is mounted during normal system operation).
As shown in
In accordance with an embodiment,
An external optical cable 306 may have a connector 308 configured to mate with optical engine component 304, as indicated by connection path 310. Mated in this way, external optical network signals can be fed directly to and from multichip package 200. By forming the optical engine component 304 directly on multichip package 200, interconnects between main die 300 and component 304 can be optimized for low power and high bandwidth density, which can help substantially reduce power consumption in a high-performance computing system while maintaining interconnect speeds of greater than 10 Gbps, 40 Gbps, 100 Gbps, 400 Gbps, etc.
Main die 300, transceiver die 302, and optical engine die 304 may be mounted on package substrate 200 using solder bumps 404 (e.g., controlled collapse chip connection (C4) bumps) and microbumps 406. It should be noted that the pitch width of solder bumps 404 may be greater than the pitch width of microbumps 406, such that microbumps 406 have greater connectivity density than solder bumps 404. The diameter of microbumps 406 are also generally smaller than the diameter of C4 bumps 404 (e.g., bumps 406 may be at least two times smaller, at least four times smaller, etc.). Solder bumps 404 are also smaller than BGA solder balls 402.
In order to facilitate communications between two chips on multichip package 200, package 200 may include one or more embedded multi-die interconnect bridge (EMIB) components 408. An EMIB is a small silicon die that is embedded in package substrate 400 and that offers dedicated ultra-high-density interconnection between dies within package 200. EMIBs generally include wires of minimal length, which help to significantly reduce loading and directly boost performance without consuming large amounts of power.
EMIB solutions may be advantageous over other multichip packaging schemes that use a silicon interposer, which is prone to issues such as warpage and requires a comparatively large number of microbumps and through-silicon vias (TSVs) to be formed on and within the interposer, thereby reducing overall yield and increasing manufacturing complexity and cost. The number of dies that can be integrated using an interposer is also limited to that supported by EMIB technology.
The EMIB technology described above may be used as an interface between two or more integrated circuit dies in package 200. In the example of
The exemplary multichip package stack-up of
The transceiver PHY may include a physical coding sublayer (PCS) and forward error correction (FEC) block 502, a serializer 504, a deserializer 506, a link management circuit 508, and/or other high-speed serial interface circuitry suitable for transmitting and receiving data. Block 502 may include, among others, decoders, encoders, data alignment circuitry, and registers such as first-in-first-out (FIFO) storage elements. Serializer 504 may be configured to transmit serialized data off of transceiver 302, whereas deserializer 506 may be configured to receive serialized data, deserialize the received data, and feed the deserialized data to block 502 for further processing. Management block 508 may be configured to control the operations of the transceiver PHY to ensure proper connection and data transfer.
Protocol processing circuit 500 may serve as a data link layer component that is used to provide address and channel access control mechanisms to support unicast, multicast, or broadcast communications services. Protocol processing circuit 500 that is used to support an Ethernet link is sometimes referred to as a media access controller (MAC). In general, protocol processing circuit 208 may be used as the interface between the transceiver PHY and the main die to support any type of network communications protocol.
Still referring to
Blocks 500 and 502 are configured to process signals in the digital domain. Blocks 504 and 506 in the transceiver PHY and blocks 510 and 516 in the optical engine are configured to process signals in the analog domain. Blocks 512 and 514 are configured to process signals in the optical domain. Operated in this way, transceiver 302 and optical engine 304 are configured as an electro-optical tile that converts signals between the digital/electrical domain and the optical domain.
At step 604, transimpedance amplifier (TIA) and limiting amplifier (LA) block 516 may receive signals from optical receiver 514. At step 606, deserializer 506 in the transceiver PHY may directly receive signals from block 516 in the optical engine. Deserializer 506 may then feed deserialized signals to the main die via blocks 502 and 500.
These steps are merely illustrative and are not intended to limit the present embodiments. The existing steps may be modified or omitted; some of the steps may be performed in parallel; additional steps may be added; and the order of certain steps may be reversed or altered. If desired, other ways of using the transceiver to directly drive optical engine within the same package via an EMIB or other embedded high-density interconnect component may be used.
The embodiments described above in which the main die, transceiver, and optical engine are formed as part of different dies within a multichip package are merely illustrative. In another suitable embodiment (see, e.g.,
In yet another suitable embodiment (see, e.g.,
The following examples pertain to further embodiments.
Example 1 is an integrated circuit package, comprising: processing circuitry; and an optical engine coupled to the processing circuitry.
Example 2 is the integrated circuit package of example 1, wherein the optical engine is optionally coupled to an external optical cable configured to mate directly with the integrated circuit package.
Example 3 is the integrated circuit package of any one of examples 1-2, optionally further comprising a transceiver coupled to the processing circuitry.
Example 4 is the integrated circuit package of example 3, wherein the transceiver is optionally coupled between the processing circuitry and the optical engine.
Example 5 is the integrated circuit package of example 4, wherein the processing circuitry, the optical engine, and the transceiver are optionally all formed on separate dies within the integrated circuit package.
Example 6 is the integrated circuit package of example 4, wherein the optical engine and the transceiver are optionally formed on the same die within the integrated circuit package.
Example 7 is the integrated circuit package of example 4, wherein the processing circuitry, the optical engine, and the transceiver are optionally all formed on the same die within the integrated circuit package.
Example 8 is the integrated circuit package of example 4, wherein the transceiver comprises a transceiver physical-layer circuit, wherein the optical engine comprises a channel driver, and wherein the transceiver physical-layer circuit optionally directly drives the channel driver.
Example 9 is the integrated circuit package of example 8, wherein the transceiver physical-layer circuit optionally comprises: a serializer that directly drives the channel driver of the optical engine; and a deserializer that directly receives signals from a transimpedance amplifier and limiting amplifier block in the optical engine.
Example 10 is the integrated circuit package of example 1, optionally further comprising an additional optical engine mounted on top of the transceiver.
Example 11 is a method of operating a multichip package, wherein the multichip package comprises a main chip and an optical engine chip, the method comprising: with the main chip, sending signals to the optical engine chip; and with the optical engine chip, receiving the signals from the main chip and outputting corresponding optical signals to an external optical cable that is directly mated to the multichip package.
Example 12 is the method of example 11, wherein the multichip package further comprises a transceiver chip interposed between the main chip and the optical engine chip, the method optionally further comprising: with the transceiver chip, conveying signals between the main chip and the optical engine chip.
Example 13 is the method of example 12, optionally further comprising: with a physical-layer component in the transceiver chip, directly driving the optical engine chip.
Example 14 is the method of example 13, wherein directly driving the optical engine chip with the physical-layer component in the transceiver chip optionally comprises: with a serializer in the physical-layer component of the transceiver chip, directly driving a channel driver in the optical engine chip.
Example 15 is the method of example 14, wherein directly driving the optical engine chip with the physical-layer component in the transceiver chip optionally further comprises: with a deserializer in the physical-layer component of the transceiver chip, directly receiving signals from a transimpedance and limiting amplifier block in the optical engine chip.
Example 16 is an integrated circuit system, comprising: a package substrate; an integrated circuit formed on the package substrate; a transceiver formed on the package substrate; and an optical engine formed on the package substrate, wherein optical engine is configured to interface directly with an optical cable.
Example 17 is the integrated circuit system of example, wherein the transceiver optionally comprises: a media access controller; a physical coding sublayer and forward error correction block configured to receive signals from the media access controller; a serializer configured to receive signals from the physical coding sublayer and forward error correction block; and a deserialzer configured to output signals to the physical coding sublayer and forward error correction block.
Example 18 is the integrated circuit system of example 17, wherein the optical engine optionally comprises: a channel driver configured to be directly driven by the serializer; an optical transmitter configured to receive signals from the channel driver; an optical receiver; and an amplifier block configured to receive signals from the optical receiver, wherein the amplifier block is configured to directly drive the deserializer.
Example 19 is the integrated circuit system of any one of examples of claims 16-18, optionally further comprising: a first embedded multi-die interconnect bridge coupled between the integrated circuit and the transceiver; and a second embedded multi-die interconnect bridge coupled between the transceiver and the optical engine.
Example 20 is the integrated circuit system of example 19, wherein the integrated circuit is optionally configured to conduct signals to the package substrate via solder bumps, wherein the integrated circuit is optionally configured to conduct signals to the first embedded multi-die interconnect bridge only via microbumps, and wherein the microbumps are at least two times smaller than the solder bumps.
Example 21 is an integrated circuit package, comprising: first means for outputting electrical signals; and second means for receiving the electrical signals and outputting corresponding optical signals to an external optical cable that is directly mated to the integrated circuit package.
Example 22 is the integrated circuit package of example 21, optionally further comprising: third means for conveying signals between the first means and the second means.
Example 23 is the integrated circuit package of example 22, wherein the first, second and third means are optionally formed on separate integrated circuit dies.
Example 24 is the integrated circuit package of example 22, wherein the second and third means are optionally formed on the same integrated circuit die.
Example 25 is the integrated circuit package of example 22, wherein the first, second and third means are optionally formed on the same integrated circuit die.
For instance, all optional features of the apparatus described above may also be implemented with respect to the method or process described herein. The foregoing is merely illustrative of the principles of this disclosure and various modifications can be made by those skilled in the art. The foregoing embodiments may be implemented individually or in any combination.
Number | Name | Date | Kind |
---|---|---|---|
6789959 | Conn | Sep 2004 | B1 |
6945712 | Conn | Sep 2005 | B1 |
8546955 | Wu | Oct 2013 | B1 |
8588561 | Zbinden | Nov 2013 | B2 |
8588562 | Zbinden | Nov 2013 | B2 |
8787711 | Zbinden | Jul 2014 | B2 |
9002155 | Li | Apr 2015 | B2 |
9184850 | Aronson | Nov 2015 | B2 |
9374165 | Zbinden | Jun 2016 | B2 |
9438355 | Blumenthal | Sep 2016 | B2 |
9577759 | Aronson | Feb 2017 | B2 |
9651752 | Zbinden | May 2017 | B2 |
9671572 | Decker | Jun 2017 | B2 |
9835811 | Zbinden | Dec 2017 | B2 |
9841572 | Zbinden | Dec 2017 | B2 |
9915796 | Zbinden | Mar 2018 | B2 |
9915797 | Zbinden | Mar 2018 | B2 |
10436992 | Zbinden | Oct 2019 | B2 |
10530614 | Wu | Jan 2020 | B2 |
10965439 | Sun | Mar 2021 | B2 |
10965501 | Wu | Mar 2021 | B2 |
20020154082 | Zavracky | Oct 2002 | A1 |
20020158823 | Zavracky | Oct 2002 | A1 |
20030053170 | Levinson et al. | Mar 2003 | A1 |
20070258551 | Sefidvash et al. | Nov 2007 | A1 |
20080292322 | Daghighian | Nov 2008 | A1 |
20090116845 | Li et al. | May 2009 | A1 |
20130148973 | Zbinden | Jun 2013 | A1 |
20130236186 | Zbinden | Sep 2013 | A1 |
20140286646 | Zbinden | Sep 2014 | A1 |
20150220472 | Sengoku | Aug 2015 | A1 |
20150365177 | Blumenthal | Dec 2015 | A1 |
20160065309 | Aronson | Mar 2016 | A1 |
20160085038 | Decker | Mar 2016 | A1 |
20160142155 | Kim | May 2016 | A1 |
20160291274 | Zbinden | Oct 2016 | A1 |
20160327760 | Zbinden | Nov 2016 | A1 |
20160327761 | Zbinden | Nov 2016 | A1 |
20170003465 | Zbinden | Jan 2017 | A1 |
20170163346 | Aronson | Jun 2017 | A1 |
20170194309 | Evan et al. | Jul 2017 | A1 |
20170219788 | Zbinden | Aug 2017 | A1 |
20170294906 | Wagh | Oct 2017 | A1 |
20170346617 | Bandi | Nov 2017 | A1 |
20190089520 | Dvir | Mar 2019 | A1 |
20190149154 | Patil | May 2019 | A1 |
20200003976 | Zbinden | Jan 2020 | A1 |
Number | Date | Country |
---|---|---|
02063800 | Aug 2002 | WO |
2014014846 | Jan 2014 | WO |
Entry |
---|
Extended European Search Report for EP Application No. 18204765.4 dated May 3, 2019. |
Communication Pursuant to Article 93(3) EPC for EP Application No. 18204765.4 dated Jan. 25, 2021. |
Ravi Mahajan et al., “Embedded Multi-Die Interconnect Bridge (EMIB)—A High Density, High Bandwidth Packaging Interconnect,” 2016 IEEE 66th Electronic Components and Technology Conference, pp. 557-565. |
Chenhui Li et al., “Chip Scale 12-Channel 10 Gb/s Optical Transmitter and Receiver Subassemblies Based on Wet Etched Silicon Interposer,” Journal of Lightwave Technology, vol. 35, No. 15, Aug. 1, 2017, pp. 3229-3236. |
Number | Date | Country | |
---|---|---|---|
20190041594 A1 | Feb 2019 | US |