Hybrid bridge fanout chiplet connection

Abstract

A chip (100) for hybrid bridge fan-out chiplet connectivity, the chip comprising a central chiplet (106), one or more first chiplets (102a-102n) each coupled to the central chiplet using a plurality of fan-out traces (110), and one or more second chiplets (104a-104m) each coupled to the central chiplet using one or more interconnect dies (ICDs) (108a-108m).

Description

Background Art

[0001] A chip composed of a plurality of chiplets may require an interconnection between each of a central chiplet and the remaining chiplets. For example, an interconnecting die (ICD) or a bridge can be used to connect a central chiplet to a chiplet adjacent to the central chiplet. However, an active bridge die covering a plurality of chiplets can affect the power supply connection and the ground connection to the die. Fan-out traces can be used to connect the chiplets to the central chiplet. However, even with a high-density fan-out routing layer, it is impossible to route all traces from a limited area of the central chiplet (e.g., a specific side or surface of the chiplet).

Brief Description of the Drawings

[0002] [Figure 1A] FIG. 13 is a block diagram of an exemplary chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 1B] FIG. 14 is a diagram of an exemplary chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 2A] FIG. 15 is a diagram of a stage of a manufacturing process of a chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 2B] FIG. 16 is a diagram of a stage of a manufacturing process of a chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 2C] FIG. 17 is a diagram of a stage of a manufacturing process of a chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 2D] FIG. 18 is a diagram of a stage of a manufacturing process of a chip for hybrid bridge fan-out chiplet connection according to some embodiments. [Figure 3]This is a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, according to several embodiments. [Figure 4] This is a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, according to several embodiments. [Figure 5] This is a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, according to several embodiments. [Figure 6] This is a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, according to several embodiments. [Figure 7] This is a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, according to several embodiments. [Modes for carrying out the invention]

[0003] According to various embodiments of the present disclosure, a hybrid bridge fan-out chiplet connection includes coupling one or more first chiplets to a central chiplet of a chip using a plurality of fan-out traces. Such a hybrid bridge fan-out chiplet connection includes coupling one or more second chiplets to a central chiplet using one or more interconnect dies (ICDs).

[0004] In some embodiments, each of the one or more second chiplets is positioned closer to the central chiplet than to the one or more first chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, bonding one or more first chiplets to the central chiplet includes laminating a plurality of fan-out trace layers on a wafer comprising the central chiplet, one or more first chiplets, and one or more second chiplets. In some embodiments, bonding one or more second chiplets to the central chiplet includes bonding one or more interconnect dies to the chip layer. In some embodiments, the method further includes forming one or more conductive pillars within the chip layer. In some embodiments, the method further includes capping one or more conductive pillars and one or more interconnect dies. In some embodiments, one or more second chiplets comprise a plurality of second chiplets, one or more interconnect dies comprise a plurality of interconnect dies, and each of the plurality of second chiplets is coupled to a central chiplet using each of the interconnect dies of the plurality of interconnect dies.

[0005] In some embodiments, the chip for hybrid bridge fanout chiplet connection includes a central chiplet, one or more first chiplets, each coupled to the central chiplet using multiple fanout traces, and one or more second chiplets, each coupled to the central chiplet using one or more interconnect dies (ICDs).

[0006] In some embodiments, each of the one or more second chiplets is positioned closer to the central chiplet than to the one or more first chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, the one or more first chiplets are coupled to the central chiplet by a plurality of fan-out trace layers stacked on a wafer including the central chiplet, the one or more first chiplets, and the one or more second chiplets. In some embodiments, one or more interconnect dies are bonded to a layer of chips stacked on a plurality of fan-out trace layers. In some embodiments, the chip further includes one or more conductive pillars. In some embodiments, the chip further includes a plurality of caps for one or more conductive pillars and one or more interconnect dies (ICDs). In some embodiments, one or more second chiplets comprise a plurality of second chiplets, one or more interconnect dies comprise a plurality of interconnect dies, and each of the plurality of second chiplets is coupled to a central chiplet using each of the interconnect dies of the plurality of interconnect dies.

[0007] In some embodiments, the device for hybrid bridge fan-out chiplet connection includes one or more components, at least one of which is operably coupled to a chip, the chip including a central chiplet, one or more first chiplets, each coupled to the central chiplet using multiple fan-out traces, and one or more second chiplets, each coupled to the central chiplet using one or more interconnect dies (ICDs).

[0008] In some embodiments, each of the one or more second chiplets is positioned closer to the central chiplet than to the one or more first chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, the one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. In some embodiments, the one or more first chiplets are coupled to the central chiplet by a plurality of fan-out trace layers stacked on a wafer including the central chiplet, the one or more first chiplets, and the one or more second chiplets. In some embodiments, one or more interconnect dies are bonded to a layer of chips stacked on a plurality of fan-out trace layers. In some embodiments, the chip further includes one or more conductive pillars. In some embodiments, the chip further includes a plurality of caps for one or more conductive pillars and one or more interconnect dies (ICDs). In some embodiments, one or more second chiplets comprise a plurality of second chiplets, one or more interconnect dies comprise a plurality of interconnect dies, and each of the plurality of second chiplets is coupled to a central chiplet using each of the interconnect dies of the plurality of interconnect dies.

[0009] Figure 1A is a block diagram of a non-exclusive, exemplary chip 100. The exemplary chip 100 can be implemented in a variety of computing devices, including mobile devices, personal computers, peripheral hardware components, game devices, set-top boxes, etc. Chip 100 includes several chiplets 102a-102n and 104a-104m. Each of the chiplets 102a-102n and 104a-104m is a functional circuit block designed to integrate with the other chiplets 102a-102n and 104a-104m. Chip 100 also includes a central chiplet 106. The central chiplet 106 is distinguished from the other chiplets 102a-102n and 104a-104m by the fact that each of the other chiplets 102a-102n and 104a-104m is coupled to the central chiplet 106 (e.g., communicatively coupled, conductively coupled). Each of the chiplets 102a-102n, 104a-104m, and the central chiplet 106 are placed on an organic substrate. The organic substrate is composed of small organic molecules or polymers containing polycyclic aromatic compounds such as pentacene, anthracene, and rubrene. Each of the chiplets 102a-102n, 104a-104m, and the central chiplet 106 are placed within a layer of a molded material such as epoxy. The molded material serves to fix the chiplets 102a-102n, 104a-104m, and the central chiplet 106 in place. The molded material layer is coplanar with the chiplets 102a-102n, 104a-104m, and the central chiplet 106, allowing an additional redistribution layer to be applied on top of the chiplets 102a-102n, 104a-104m, and the central chiplet 106.

[0010] Communication connections between the central chiplet 106 and chiplets 102a-102n, 104a-104m are used to perform input / output communication between chip components. One existing solution for connecting multiple chiplets 102a-102n, 104a-104m to the central chiplet 106 involves using an interconnect die (ICD) or active bridge die that can be used to connect the central chiplet 106 to adjacent chiplets 104a-104m. However, such an active bridge die covering multiple chiplets may affect power and ground connections to the chiplet die. An alternative existing embodiment for connecting chiplets to the central chiplet involves using fan-out traces (e.g., embedded in a redistribution layer) to connect chiplets m to the central chiplet. However, even with a high-density fan-out routing layer, routing all necessary traces from a limited area of ​​the central chiplet to many different other chiplets is often impossible and does not scale as the number of chiplets that need to be connected to the central chiplet increases.

[0011] However, the exemplary chip 100 in Figure 1A implements interconnect dies (ICDs) (e.g., bridge dies) 108a-108m to connect the central chiplet 106 to the chiplet closest to the central chiplet 106 among chiplets 102a-102n and 104a-104m, and multiple fanout traces 110 to connect the central chiplet 106 to chiplets 102a-102n and 104a-104m that are not connected to the central chiplet, using interconnect dies 108a-108m. In this configuration, the connection that implements both fanout and ICD to connect multiple chiplets to the central chiplet is called a hybrid bridge fanout interconnect. In this way, the power and ground connections are not affected by the ICD, and designs utilizing such hybrid bridge fanout interconnects allow the number of chiplets connected to the central chiplet to be scalable.

[0012] The interconnecting dies 108a-108m are silicon dies that provide connection coupling between two chiplets. For example, both the central chiplet 106 and the chiplets 104a-104m include multiple input / output (I / O) connection points made of metal or other conductive material. The interconnecting dies 108a-108m include conductive paths that terminate at the I / O connection points. By aligning the I / O connection points of the interconnecting dies 108a-108m with the I / O connection points of the central chiplet 106 and the chiplets 104a-104m, and then joining the interconnecting dies 108a-108m to the central chiplet 106 and the chiplets 104a-104m, a conductive path is formed between the central chiplet 106 and the chiplets 104a-104m through the interconnecting dies 108a-108m. In the illustrated example, chip 100 contains two rows of chiplets, with the row of chiplets 104a-104m being closest to the central chiplet 106. Therefore, interconnect dies 108a-108m connect the chiplets 104a-104m to the central chiplet 106. For example, in some embodiments, each chiplet 104a-104m connected using interconnect dies 108a-108m is connected to the central chiplet 106 using its own dedicated interconnect dies 108a-108m. In other words, m interconnect dies 108a-108m are used to connect m chiplets 108a-108m to the central chiplet 106. In other embodiments, a single interconnect die 108a-108m is used to connect multiple chiplets 104a-104m in the same row to the central chiplet 106.

[0013] Furthermore, the chip 100 uses multiple fan-out traces 110 to connect the central chiplet 106, and interconnect dies 108a to 108m to connect chiplets 102a to 102n and 104a to 104m that are not connected to the central chiplet 106. In the exemplary chip 100, chiplets 102a to 102n are connected to the central chiplet 106 using fan-out traces 110. The fan-out traces 110 are traces of conductive material such as carbon, silver, or aluminum, traced within a layer of dielectric material such as polyimide. The fan-out traces 110 are traced in multiple layers of dielectric material, hereafter referred to as fan-out trace layers. The fan-out trace layers into which the fan-out traces 110 are embedded are redistribution layers. Redistribution layers are generally extra metal layers on the chip that make the I / O pads of the integrated circuit available elsewhere on the chip for better access to the pads when needed. Each fan-out trace 110 connects the central chiplet 106 to chiplets 102a-102n via one or more conductive interconnects in the intermediate layer of the chip 100 (e.g., the intermediate fan-out trace layer or other redistribution layer). For example, the fan-out trace 110 provides a conductive link from the I / O connection points of chiplets 102a-102n to the I / O connection points of the central chiplet 106.

[0014] In some embodiments, each fan-out trace layer includes fan-out traces 110 from the central chiplet 106 to each chiplet 102a-102n connected using fan-out traces 110. For example, the first fan-out trace layer includes a first fan-out trace 110 from the central chiplet 106 to each chiplet 102a-102n, the second fan-out trace layer includes a second trace 110 from the central chiplet 106 to each chiplet 102a-102n, and so on. Thus, assuming x fan-out trace layers, each chiplet 102a-102n has x fan-out traces 110 to the central chiplet. Those skilled in the art will understand that other combinations or arrangements of fan-out traces 110 in the fan-out trace layers are possible.

[0015] The chip 100 in Figure 1A implements both interconnecting dies 108a-108m for coupling the central chiplet 106 to the nearest chiplets 104a-104m, and a fan-out trace 110 for connecting the central chiplet 106 to other further chiplets 102a-102n. Those skilled in the art will understand that the arrangement of chiplets 102a-102n, 104a-104m and the central chiplet 106 is illustrative and other arrangements are possible. For example, in some embodiments, the chip 100 includes an additional column of chiplets that share a row with the chiplets 102a-102n, 104a-104m. In some embodiments, the additional column of chiplets is arranged adjacent to the opposing face of the central chiplet 106 (e.g., the right face of the central chiplet 106 opposite the left face of the central chiplet 106). In such embodiments, this additional column of chiplets is positioned adjacent to the central chiplet 106, and is connected to the central chiplet 106 using an interconnect die. In some embodiments, further columns of chiplets are positioned adjacent to this additional column of chiplets and are connected using additional fan-out traces.

[0016] Those skilled in the art will also understand that the use of “rows” or “columns” of chiplets as used herein relates to which face of the central chiplet 106 a particular group of chiplets is positioned relative to. For example, the above example illustrates connecting the rows of chiplets 104a to 104m closest to the left face of the central chiplet 106 using interconnect dies 108a to 108m, but in some embodiments, rows of chiplets closest to the top or bottom face of the central chiplet 106 are connected using interconnect dies. In this example, additional rows of chiplets further away from the top or bottom face of the central chiplet 106 are also connected using fan-out traces 110.

[0017] Figure 1B is a diagram of an exemplary chip for hybrid bridge fan-out chiplet connection according to several embodiments. For example, Figure 1B shows a cross-sectional view of chip 100 of Figure 1A. Figure 1B shows chiplets 102a, 104a, and central chiplet 106 in layers of molded product 120. Molded product 120 comprises epoxy or another material that fixes chiplets 102a, 104a, and central chiplet 106 in place on a substrate (not shown). Such a substrate includes, for example, an organic substrate composed of small organic molecules or polymers containing polycyclic aromatic compounds such as pentacene, anthracene, and rubrene.

[0018] The redistribution layer 122 is deposited on a layer of molded product 120 containing chiplets 102a, 104a, and a central chiplet 106. The redistribution layer 122 is made of a dielectric material such as polyimide or another insulating material. The redistribution layer 122 includes a conductive interconnect 124 made of copper or another conductive material. The conductive interconnect 124 provides input / output point connection points for chiplets 102a, 104a, and the central chiplet 106. Thus, signals between any of the chiplets 102a, 104a, and the central chiplet 106 use a conductive path having the conductive interconnect 124 as an endpoint.

[0019] A plurality of fan-out trace layers 126 are stacked on the redistribution layer 122. Each fan-out trace layer 126 is a redistribution layer (e.g., a layer of a dielectric material such as polyimide or another insulating material) that accommodates one or more fan-out traces 110. The fan-out traces 110 of each fan-out trace layer 126 form a signal path between the central chiplet 106 and the chiplet 102. Each fan-out trace layer 126 also includes conductive interconnects 128. The conductive interconnects 124 provide input / output point connection points for the chiplets 102a, 104a, and the central chiplet 106, while the conductive interconnects 128 provide a conductive path between the fan-out trace layers 126. Thus, signals proceed between adjacent fan-out trace layers 126, the redistribution layer 122, or the redistribution layer 130 described below using the conductive interconnects 128.

[0020] Another redistribution layer <130> is stacked on the fan-out trace layers <126>. The redistribution layer <130> accommodates conductive pillars <132> of copper or another conductive material. The conductive pillars <132> provide a conductive path between the cap <134> and the chiplets <102a>, <104a>, and the central chiplet <106> through the intervening conductive interconnects <124>, <128>. The cap <132> is composed of a tin-silver alloy or another material suitable for solderable connections. The interconnect die <108a> is also accommodated within the redistribution layer <{130}>. The interconnect die <108a> forms a signal path between the central chiplet <106> and <104a using the conductive interconnects <128> of the intervening redistribution layer <126>. The conductive pillars <130> and the interconnect die <108a> are further accommodated within another layer <134> of the molding to hold the conductive pillars <130> and the interconnect die <108a> in place.

[0021] It should be noted that in the original text, there seems to be a bit of confusion in the numbering and naming consistency in the description of "導電性ピラー130" in the translated part of ID=4. It should probably be "導電性ピラー132" for better coherence, but I translated it as is according to the original text.Figures 2A to 2D show exemplary side views of the manufacturing stages of a chip for hybrid bridge fan-out chiplet connections according to several embodiments. As shown in Figure 2A, the silicon die for the central chiplet 106, chiplet 102a, and chiplet 104a is reconfigured on a carrier (not shown). Reconfiguring the central chiplet 106, chiplet 102a, and chiplet 104a includes positioning the central chiplet 106, chiplet 102a, and chiplet 104a on the carrier and applying a molded material 202 around the central chiplet 106, chiplet 102a, and chiplet 104a to fix their positions within the chip 100. In some embodiments, the molded material 202 includes epoxy or another material. The front aluminum layer 204 is exposed to allow conductive connectivity of the central chiplet 106, chiplet 102a, and chiplet 104a. Although Figure 2A describes the aluminum layer 204, it will be understood that other conductive materials can be used instead of or in addition to aluminum.

[0022] As shown in Figure 2B, a fan-out trace layer 206 is applied to the chip 100 on the aluminum layer 204. Each fan-out trace layer 206 is a redistribution layer containing one or more fan-out traces 110 made of copper or another conductive material. In this example, the fan-out traces 110 provide connections between the central chiplet 106 and the chiplet 102a. Each fan-out trace layer 206 also includes a conductive interconnect 208 that provides conductive paths between the fan-out trace layers 206. The conductive interconnect 208 is made of copper or another conductive material. The fan-out trace layer 206 is also made of a dielectric material such as polyimide or another insulating material. Thus, the dielectric material of the fan-out trace layer 206 accommodates the fan-out traces 110 and the conductive interconnect 208. By applying multiple fan-out trace layers 206, the multiple connection paths of the fan-out traces 110 connect the central chiplet 106 to the chiplet 102a. Furthermore, the conductive interconnect 208 enables signal transmission between the fan-out trace layer 206 and the central chiplet 106 to the chiplets 102a and 104a. For example, as will be described in more detail below, signal paths from the chiplets 102a, 104a and the central chiplet 106 are formed through the conductive interconnect 208 of the fan-out trace layer 206 and terminate at solderable connection points on the surface of the chip.

[0023] As shown in FIG. 2C, another layer of dielectric material (e.g., another redistribution layer) is applied over the top of the fan-out trace layer 208. Conductive pillars 210 are formed within this applied redistribution layer. In some embodiments, forming the conductive pillars 210 includes inserting pre-formed conductive pillars within the dielectric material forming the redistribution layer. In other embodiments, forming the conductive pillars includes extruding a conductive material to form the conductive pillars within the redistribution layer. The conductive pillars 210 are composed of copper or another conductive material. An interconnect die 108a is disposed within this redistribution layer to provide a connection coupling between the central chiplet 106 and the chiplet 104a via the conductive interconnects 208 included in the intermediate fan-out trace layer 206. For example, the interconnect die 108a includes a conductive path that contacts or is bonded to the conductive interconnect 208 coupled to the central chiplet 106 on one end and contacts or is bonded to the conductive interconnect 208 coupled to the chiplet 104a on another end. In some embodiments, the interconnect die 108a includes one or more through-silicon vias that provide a conductive path through the interconnect die 108a from one side of the interconnect die 108a to the opposite side of the interconnect die 108a (e.g., from the top of the interconnect die 108a to the opposite side surface in the redistribution layer). Thus, solderable connections can be formed using the interconnect die 108a into the underlying fan-out trace layer 206 and other components through the through-silicon vias.

[0024] As shown in FIG. 2D, an additional molding 212 is applied to the chip 百. Next, the molding 212 is partially ground to expose the conductive pillars 210 and, if any, the through-silicon vias of the interconnect die 108. Thus, the ground molding 212 is flush with the exposed conductive pillars 210 and through-silicon vias. A cap 214 is applied to the exposed conductive pillars 210 and the through-silicon vias of the interconnect die 108a. The cap 214 is composed of a tin-silver alloy or another material suitable for solderable connections.

[0025] Figures 2A to 2D show a manufacturing process (e.g., a “die-first” manufacturing process) in which the constituent layers are applied to the chiplets 102a, 104a, and central chiplet 106. However, it will be understood that in some embodiments, the chip 100 is manufactured using a “die-last” manufacturing process. For example, the chiplets 102a, 104a, and central chiplet 106 are applied as part of the last layer applied to the chip 100.

[0026] For further explanation, Figure 3 shows a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection, which involves connecting one or more first chiplets 102a-102n to a central chiplet 106 using multiple fan-out traces 110 (e.g., within chip 100). In some embodiments, one or more first chiplets 102a-102n are located in the same column of multiple columns of chiplets 102a-102n, 104a-104m. In such embodiments, the multiple first chiplets 102a-102n are chiplets from chiplets 102a-102n, 104a-104m that are not adjacent to the central chiplet 106 (e.g., separated from the central chiplet 106 by one or more other columns of chiplets). In some embodiments, one or more first chiplets 102a-102n are located in the same row among multiple rows of chiplets 102a-102n, 104a-104m. In such embodiments, the multiple first chiplets 102a-102n are chiplets among chiplets 102a-102n, 104a-104m that are not adjacent to the central chiplet 106 (for example, separated from the central chiplet 106 by one or more other rows of chiplets). The fan-out trace 110 is a trace of copper or another conductive material etched within a layer of dielectric material or applied to a layer of dielectric material. In some embodiments, the fan-out trace 110 connects the central chiplet 106 and the chiplets 102a-102n via one or more intermediate layers through one or more conductive interconnects 208.

[0027] Furthermore, the method in Figure 3 includes coupling one or more second chiplets 104a-104m to the central chiplet 106 using one or more interconnect dies 108a-108m. The interconnect dies 108a-108m are silicon dies that provide a connecting link between the chiplets 108a-108m and the central chiplet 106. In some embodiments, each chiplet 104a-104m that is coupled using the interconnect dies 108a-108m is connected to the central chiplet 106 using its own dedicated interconnect die 108a-108m. In other words, m interconnect dies 108a-108m are used to connect m chiplets 108a-108m to the central chiplet 106. In some embodiments, one or more second chiplets 104a-104m are located in the same column of multiple columns of chiplets 102a-102n, 104a-104m. In such embodiments, the multiple second chiplets 104a to 104m are chiplets in a row of chiplets 102a to 102n and 104a to 104m that are adjacent to or closest to the central chiplet 106. In some embodiments, one or more second chiplets 104a to 104m are included in the same row of multiple rows of chiplets 102a to 102n and 104a to 104m. In such embodiments, the multiple second chiplets 104a to 104m are chiplets in a row of chiplets 102a to 102n and 104a to 104m that are adjacent to or closest to the central chiplet 106.

[0028] For further explanation, Figure 4 shows a flowchart illustrating an exemplary method for hybrid bridge fan-out chiplet connection. The method in Figure 4 is similar to the method in Figure 3 in that it includes coupling one or more first chiplets 102a-102n to a central chiplet 106 using multiple fan-out traces 110 302, and coupling one or more second chiplets 104a-104m to the central chiplet 106 using one or more interconnect dies 108a-108m 304 (for example, within chip 100).

[0029] The method in Figure 4 differs from Figure 3 in that the method involves bonding one or more first chiplets 102a-102n to a central chiplet 106 using a plurality of fan-out traces 110 (e.g., within a chip 100), but also involves laminating a plurality of fan-out trace layers 206 onto a wafer comprising a central chiplet 106, one or more first chiplets 102a-102n, and one or more second chiplets 104a-104m. The wafer comprises a reconfigured central chiplet 106, one or more first chiplets 102a-102n, and one or more second chiplets 104a-104m, which are positioned and fixed in place using a molded product 202 (e.g., epoxy or another material). In some embodiments, multiple fan-out trace layers 206 are laminated on an exposed aluminum layer bonded or connected to a central chiplet 106, one or more first chiplets 102a to 102n, and one or more second chiplets 104a to 104m.

[0030] The fan-out trace 110 is a trace of a conductive material such as carbon, silver, or aluminum traced within a lattice of a dielectric material such as polyimide. The fan-out trace 110 is traced across multiple layers of the dielectric material. Each fan-out trace 110 connects the central chiplet 106 to the first chiplets 102a to 102n via one or more conductive interconnects in the intermediate layer of the chip 100 (e.g., the intermediate fan-out trace layer 206 or other layers).

[0031] In some embodiments, each fan-out trace layer 206 includes fan-out traces 110 from the central chiplet 106 to each second chiplet 102a-102n connected using fan-out traces 110. For example, the first fan-out trace layer includes a first fan-out trace 110 from the central chiplet 106 to each chiplet 102a-102n, the second fan-out trace layer includes a first second trace 110 from the central chiplet 106 to each chiplet 102a-102n, and so on. Thus, assuming x fan-out trace layers 206, each chiplet 102a-102n has x fan-out traces 110 to the central chiplet. Those skilled in the art will understand that other combinations or arrangements of fan-out traces 110 in the fan-out trace layers 206 are possible.

[0032] For further explanation, Figure 5 shows a flowchart illustrating an exemplary method for hybrid bridge fanout chiplet connection according to some embodiments of the present disclosure. The method in Figure 5 is similar to the method in Figure 3 in that it includes coupling one or more first chiplets 102a-102n to a central chiplet 106 using a plurality of fanout traces 110 302, and coupling one or more second chiplets 104a-104m to the central chiplet 106 using one or more interconnect dies 108a-108m 304 (for example, within chip 100).

[0033] The method in Figure 5 differs from that in Figure 3 in that bonding one or more second chiplets 104a-104m to a central chiplet 106 using one or more interconnect dies 108a-108m (for example, within chip 100) includes bonding one or more interconnect dies 108a-108m to a layer of the chip 502. In some embodiments, the layer to which one or more interconnect dies 108a-108m are bonded is laminated on top of one or more fan-out trace layers 206. Thus, in some embodiments, bonding one or more interconnect dies 108a-108m 502 includes bonding one or more interconnect dies 108a-108m to a conductive interconnect portion 208 in the fan-out trace layer 206, which provides a conductive connection to the central chiplet 106 and the corresponding second chiplets 104a-104m for a given interconnect dies 108a-108m. In some embodiments, the layer to which one or more interconnect dies 108a to 108m are joined includes a layer of dielectric material.

[0034] For further explanation, Figure 6 shows a flowchart illustrating another exemplary method for hybrid bridge fan-out chiplet connection according to embodiments of the present disclosure. The method of Figure 6 is similar to the method of Figure 3 in that it includes coupling one or more first chiplets 102a-102n to a central chiplet 106 using a plurality of fan-out traces 110 302, and coupling one or more second chiplets 104a-104m to the central chiplet 106 using one or more interconnect dies 108a-108m 304 (for example, within chip 100).

[0035] The method in Figure 6 differs from that in Figure 3 in that it includes forming one or more conductive pillars 210 within a layer of chip 100.602 In some embodiments, the layer of chip 100 on which the conductive pillars 210 are formed is the layer to which one or more interconnect dies 108a-108m are bonded. In some embodiments, the conductive pillars 210 are made of copper or another conductive material.302 In some embodiments, forming the conductive pillars 210 includes inserting pre-formed conductive pillars 210 into a dielectric material forming a layer.In other embodiments, forming the conductive pillars 210 includes extruding a conductive material to form the conductive pillars 210 within a layer.In some embodiments, a molded product 212 such as epoxy is applied around the conductive pillars 210.

[0036] For further explanation, Figure 7 shows a flowchart illustrating another exemplary method for hybrid bridge fan-out chiplet connection according to embodiments of the present disclosure. The method of Figure 7 is similar to the method of Figure 3 in that it includes coupling one or more first chiplets 102a-102n to a central chiplet 106 using a plurality of fan-out traces 110 302, coupling one or more second chiplets 104a-104m to the central chiplet 106 using one or more interconnect dies 108a-108m 304 (e.g., within chip 100), and forming one or more conductive pillars 210 within the layers of chip 100 602.

[0037] The method in Figure 7 differs from that in Figure 6 in that it includes capping 702 one or more conductive pillars 210 and one or more interconnect dies 108a-108m. Capping 702 one or more conductive pillars 210 and one or more interconnect dies 108a-108m involves applying a certain amount of capping material to one or more conductive pillars 210 and one or more interconnect dies 108a-108m to facilitate soldering or other connections. For example, in some embodiments, the capping material includes a tin-silver alloy or other material suitable for solderable connections.

[0038] Considering the above explanation, readers will recognize that the advantages of a hybrid bridge fanout chiplet connection include the following: • Improved computing system performance by providing low-latency, high-bandwidth connectivity between the central chiplet and other chiplets on the same chipset.

[0039] The flow diagrams and block diagrams in the figures illustrate the architecture, functionality, and operation of possible embodiments of the systems, methods, and computer program products according to various embodiments of the present disclosure. In this regard, each block in a flow diagram or block diagram may represent a module, segment, or portion of instructions containing one or more executable instructions for implementing a specified logical function. In some alternative embodiments, the functions described in a block may occur out of the order shown in the figure. For example, two blocks shown consecutively may actually be executed substantially simultaneously, and blocks may be executed in reverse order depending on the functions involved. It should also be noted that each block in a block diagram and / or flow diagram, and combinations of blocks in a block diagram and / or flow diagram, may be implemented by a dedicated hardware-based system that performs a specified function or action, or combines dedicated hardware with computer instructions.

[0040] It will be understood from the above description that modifications and changes can be made to various embodiments of this disclosure. The statements herein are for illustrative purposes only and should not be construed as restrictive. The scope of this disclosure is limited only by the following claims.

Claims

1. A chip for hybrid bridge fan-out chiplet connection, The aforementioned chip is The central tiplet and One or more first chiplets, each coupled to the central chiplet using multiple fan-out traces, and one or more first chiplets not coupled to the central chiplet using an interconnect die (ICD), It comprises one or more second chiplets, each coupled to the central chiplet using one or more ICDs, The one or more ICDs are located in layers that do not include the plurality of fan-out traces. Tip.

2. Each of the one or more second chiplets is positioned closer to the central chiplet than the one or more first chiplets. The chip according to claim 1.

3. The one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. The chip according to claim 1.

4. The one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. The chip according to claim 1.

5. The one or more first chiplets are coupled to the central chiplet by a plurality of fan-out trace layers stacked on a wafer comprising the central chiplet, the one or more first chiplets, and the one or more second chiplets. The chip according to claim 1.

6. The one or more interconnect dies are bonded to the layers of the chip stacked on the plurality of fan-out trace layers. The chip according to claim 5.

7. Further comprising one or more conductive pillars, The chip according to claim 1.

8. The system further comprises one or more conductive pillars and a plurality of caps for one or more interconnecting dies (ICDs), The chip according to claim 7.

9. The one or more second chiplets include a plurality of second chiplets, the one or more interconnecting dies include a plurality of interconnecting dies, and each of the plurality of second chiplets is coupled to the central chiplet using each of the interconnecting dies of the plurality of interconnecting dies. The chip according to claim 8.

10. A device for connecting hybrid bridge fan-out chiplets, The aforementioned device is The chip comprises one or more components operably coupled to it, The aforementioned chip is The central tiplet and One or more first chiplets, each coupled to the central chiplet using multiple fan-out traces, and one or more first chiplets not coupled to the central chiplet using an interconnect die (ICD), It comprises one or more second chiplets, each coupled to the central chiplet using one or more ICDs, The one or more ICDs are located in layers that do not include the plurality of fan-out traces. Device.

11. Each of the one or more second chiplets is positioned closer to the central chiplet than the one or more first chiplets. The apparatus according to claim 10.

12. The one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. The apparatus according to claim 10.

13. The one or more first chiplets are arranged in a first row of chiplets, and the one or more second chiplets are arranged in a second row of chiplets. The apparatus according to claim 10.

14. The one or more first chiplets are coupled to the central chiplet by a plurality of fan-out trace layers stacked on a wafer comprising the central chiplet, the one or more first chiplets, and the one or more second chiplets. The apparatus according to claim 10.

15. The one or more interconnect dies are bonded to the layers of the chip stacked on the plurality of fan-out trace layers. The apparatus according to claim 14.

Images