The approaches described in this section are approaches that could be pursued, but not necessarily approaches that have been previously conceived or pursued. Therefore, unless otherwise indicated, it should not be assumed that any of the approaches described in this section qualify as prior art merely by virtue of their inclusion in this section. Further, it should not be assumed that any of the approaches described in this section are well-understood, routine, or conventional merely by virtue of their inclusion in this section.
As the computing power of semiconductor devices increases, power delivery becomes more challenging. It is not uncommon for high-end server and other datacenter components to require several hundred Watts of power, making the placement of power components, such as voltage regulators, more critical.
One solution is to incorporate power components into semiconductor device packages closer to where power is needed, but with this approach power must still be delivered across the package substrate to the power consuming dies, which adds impedance to the power distribution network. This approach also consumes valuable routing resources within the package.
Another solution is to collocate power components on the silicon die with semiconductor devices, but this can be expensive for System-on-a-Chip (SoC) applications that are implemented with advanced process node silicon that is more costly. Also, the power components compete with other elements for die area, leading to undesirable tradeoffs in area, power, performance and/or cost.
In view of the foregoing, there is a need for an approach for implementing power components closer to semiconductor devices that avoids the limitations and costs of other solutions.
Implementations are depicted by way of example, and not by way of limitation, in the figures of the accompanying drawings and in which like reference numerals refer to similar elements.
In the following description, for the purposes of explanation, numerous specific details are set forth in order to provide a thorough understanding of the implementations. It will be apparent, however, to one skilled in the art that the implementations may be practiced without these specific details. In other instances, well-known structures and devices are shown in block diagram form in order to avoid unnecessarily obscuring the implementations.
I. Overview
II. Architecture
A semiconductor device includes one or more active devices disposed between a processor die and a package substrate. According to an implementation, the semiconductor device includes a first layer with a processor die, a second layer with one or more active devices, and a third layer with a package substrate, where the second layer is disposed between the first and third layers. The one or more active devices are semiconductor-based devices that perform one or more functions. According to an implementation, the one or more active devices include power components, such as one or more voltage regulators, power management circuits, charge pumps, power rectifiers, power diodes, thyristors, switched-mode power supplies, etc., that participate in supplying power to the processor die and are electrically connected to the processor die using various connection configurations.
The implementations described herein provide the technical benefits of short path lengths for improved performance with a compact structure that avoids the use of edge wiring or interposers without occupying processor die space. Implementations include the use of through-silicon vias (TSVs) to provide short path lengths while reducing the number of connection resources used by the one or more power components.
A. Overview
The processor die 112 is a die for any type of processor, such as a Central Processing Unit (CPU), Graphics Processing Unit (GPU), Accelerated Processing Unit (APU), Application Specific Integrated Circuit (ASIC), Field Programmable Gate Array (FPGA), etc., and the memory die 114 is a die for any type of memory, such as High Bandwidth Memory (HBM), DRAM, SRAM, etc. The processor die 112 has a top surface 120a and a bottom surface 120b. The interconnect die 116 provides electrical connections between the processor die 112 and the memory die 114, for example, to carry commands and data between the computing elements formed on the processor die 112 and memory elements formed on the memory die 114. The interconnect die 116 is optional and alternatively an interposer is used to provide electrical connections between the processor die 112 and the memory die 114.
Connections 122 electrically connect the processor die 112 and the memory die 114 to the interconnect die 116 via a redistribution layer 124 and the connections 122 are implemented, for example, by metal pillars. The redistribution layer 124 is comprised of any number of metal routing layers. According to an implementation, the redistribution layer 124 is comprised of a polymer and acts as a stress buffer and/or an isolation film, while enabling redistribution layer routing. Connections 126a, 126b, 128a, 128b electrically connect, respectively, the processor die 112 and the memory die 114 to the package substrate 170. Connections 126a, 128a are comprised of metal pillars and connections 126b, 128b provide electrical connectivity to the package substrate 170 via, for example, C4 bumps or similar structures.
The VR 118 is any type of voltage regulator that is capable of converting an input voltage to one or more regulated output voltages. Example implementations of the VR 118 include, without limitation, an IC linear voltage regulator, an IC switching regulator, a DC/DC converter chip, etc. According to an implementation, the VR 118 is a silicon-based device. The VR 118 provides one or more regulated voltages to one or more components in the chip module 110 including, for example, circuitry on the processor die 112 and/or on the memory die 114. According to an implementation, the VR 118 provides one or more regulated voltages to multiple components, such as multiple processor dies, multiple memory dies, one or more processor dies and one or more memory dies, etc.
According to an implementation, the VR 118 is separately fabricated and then placed into the chip module 110 using a placement process, such as the process described in U.S. Pat. No. 10,510,721, the contents of which are hereby incorporated by reference in its entirety for all purposes. The processor die 112, the memory die 114, the interconnect die 116 and the VR 118 are held in position by a mold compound 130, such as an epoxy material, filler, etc.
Implementations are applicable to different physical sizes, shapes, and placements of the VR 118. In the example of
The package substrate 170 is a semiconductor device package substrate that includes any number of layers, such as a substrate, upper layers, and lower layers, which vary depending on a particular implementation.
The VR 118 has a top surface 132a and a bottom surface 132b. The physical orientation of the VR 118 varies depending upon a particular implementation. In the example of
Connections 136a, 136b and 138a, 138b electrically connect the redistribution layer 124 to the package substrate 170. The connections 136a, 136b and 138a, 138b are implemented by metal pillars, similar to connections 122. Implementations are not limited to the exact numbers of connections 134, 136a, 136b, 138a, 138b depicted in
According to the “face up” implementation, an input voltage is provided from the package substrate 170 to the VR 118 via the connections 136a, 136b and 138a, 138b, the redistribution layer 124 and the portion of the connections 134 that connect the redistribution layer 124 to the VR 118. The VR 118 provides a regulated output voltage to the processor die 112 via the portion of the connections 134 that connect the VR 118 to the processor die 112, and not the portion of the connections that are used to supply the input voltage to the VR 118. According to an implementation, the input voltage provided by the package substrate 170 is higher than the regulated output voltage that is provided to the processor die 112.
According to a “face down” implementation, the VR 118 is oriented so that the top surface 132a with the circuitry for performing voltage regulation, electrical connections, etc., faces downward towards the package substrate 170.
In the “face down” implementation depicted in
According to an implementation, Through-Silicon Vias (TSVs) are used to provide a compact semiconductor device structure with reduced use of metal connections.
In this example, the VR 118 is oriented “Face Down” so that the top surface 132a with the electrical connections and circuitry, such as metal traces, etc., faces downward towards the package substrate 170. In this implementation, an input voltage is provided by the package substrate 170 to the VR 118 via one or both of the connections 140a, 140b. The VR 118 generates a regulated voltage and provides the regulated voltage through one or more TSVs 142 to the connections 134 on the bottom surface 132b, and then to the processor die 112 via the redistribution layer 124, metal traces, etc. This provides the technical benefits of short path lengths while using relatively few connection resources, such as connections 136a, 136b, 138a, 138b (
According to an implementation, TSVs are used with semiconductor device structures with a power component in a “Face Up” orientation.
The input voltage is provided to the circuitry on the top surface 132a by the TSVs 142. The VR 118 generates a regulated voltage and provides the regulated voltage through the connections 134 and the redistribution layer 124, metal traces, etc., to the processor die 112. This provides the technical benefits of short path lengths while using relatively few connection resources, such as connections 136a, 136b, 138a, 138b, but in this implementation the TSVs 142 carry the input voltage instead of the regulated voltage as in
In step 206, the power components and other components are added to the chip module. For example, the interconnect die 116 and the VR 118 are added to the chip module 110 on the package substrate 170 using HDCL technology and secured in place with the mold compound 130. According to an implementation, this includes adding the connections 126b, 128b, 136b and 138b to the bottom of the chip module 110.
In step 208, the fabrication of the semiconductor device is completed, for example, by assembling the top layer components, e.g., the processor die 112 and the memory die 114, with the components from step 206, e.g., the VR 118 and the interconnect die 116, by adding or otherwise connecting the vertical connections at the interface, e.g., connections 122, 134 and connections 126a, 136a, 138a, 128a. Not all of the steps of