The present invention relates to a semiconductor module.
In the related art, a volatile memory (RAM) such as a dynamic random access memory (DRAM) has been known as a storage device. DRAMs are required to have high performance of an arithmetic unit (hereinafter referred to as a logic chip) and a large capacity capable of withstanding an increase in amount of data. Therefore, the capacity has been increased by miniaturizing a memory (memory cell array, memory chip) and increasing the number of cells in a plane. On the other hand, this type of increase in capacity has reached its limit due to the weakness to noise caused by the miniaturization, the increase in die area, and the like.
Therefore, in recent years, a technology has been developed that realizes a large capacity by laminating a plurality of planar memories to form a three-dimensional (3D) structure. In addition, there has been proposed a semiconductor module that reduces an installation area of the logic chip and the RAM by disposing the logic chip and the RAM by lamination (refer to, for example, Patent Documents 1 and 2). Patent Document 1: Japanese Unexamined Patent Application (Translation of PCT Application), Publication No. 2014-512691 Patent Document 2: Japanese Unexamined Patent Application, Publication No. 2010-232659
With the increase in performance of the logic chip and the increase in amount of data, an improvement in communication rate between the logic chip and the RAM is also required along with the increase in capacity. Therefore, it is preferable to provide a semiconductor module capable of improving a bandwidth between the logic chip and the RAM.
An object of the invention is to provide a semiconductor module capable of improving a bandwidth between a logic chip and a RAM.
According to the invention, there is provided a semiconductor module including: a logic chip; a RAM unit which is a lamination-type RAM module; a spacer which is disposed to be laminated over the RAM unit along a lamination direction; an interposer which is electrically connected to each of the logic chip and the RAM unit; and a connection unit which communicably connects the logic chip and the RAM unit. The logic chip and the spacer are disposed to be adjacent to each other in a direction intersecting the lamination direction of the RAM unit, wherein the RAM unit is placed on the interposer, and one end of the RAM unit is disposed to overlap one end of the logic chip in the lamination direction. The connection unit communicably connects the one end of the RAM unit and the one end of the logic chip.
In addition, it is preferable that a pair of the RAM unit and the spacer are provided with the logic chip interposed therebetween, and the connection unit is provided for each of the RAM units.
In addition, it is preferable that a thickness of the spacer is substantially equal to a thickness to the logic chip.
In addition, it is preferable that a thickness of the spacer is larger than a thickness of the logic chip.
In addition, it is preferable that, among ends of the spacer, an end opposite to a side facing the logic chip is disposed to protrude from the RAM unit in a direction intersecting the lamination direction of the RAM unit.
In addition, it is preferable that the semiconductor module further includes a plurality of pillars which communicably connect the interposer and the logic chip, each pillar being longer than a thickness of the RAM unit in the lamination direction.
In addition, it is preferable that a plurality of the logic chips are provided for one interposer, and a pair of the RAM units and a pair of the spacers are provided for each of the logic chips. Effects of the Invention
According to the present invention, it is possible to provide a semiconductor module capable of improving a bandwidth between a logic chip and a RAM.
Hereinafter, a semiconductor module according to one embodiment of the present invention will be described with reference to
As illustrated in
As illustrated in
The MPU 20 is a plate-like body having a rectangular shape in a plan view. As illustrated in
A plurality of pillars 30 are disposed. The pillars 30 communicably connect the interposer 10 and the MPU 20. Specifically, one end of the pillar 30 is connected to the interposer 10, and the other end is connected to the circuit surface of the MPU 20. Each of the pillars 30 is longer than the thickness in the lamination direction C of the RAM unit 40 described later.
The RAM unit 40 is configured to include a lamination-type RAM module having a rectangular shape in a plan view. The RAM unit 40 is placed on the upper surface of the interposer 10 as illustrated in
The RAM unit 40 is formed by laminating memory circuits (not illustrated). Specifically, the RAM unit 40 is formed by laminating dies (not illustrated) of a plate-like body having a rectangular shape in a plan view having a memory circuit on the upper surface in the lamination direction C. The die is an Si substrate in which circuits are formed therein, and each of the laminated dies is electrically connected to an adjacent die. A power supply terminal and a ground terminal connecting between the laminated dies are formed by, for example, bumpless TSV, and a signal line is formed by a TCI (ThruChip Interface). In addition, the phrase “electrically connected” is not limited to “directly connected”, but the phrase includes “indirectly connected” (for example, by using a magnetic field) like TCI.
The connection unit 50 is a communication interface that connects the MPU 20 and the RAM unit 40. The connection unit 50 is configured with, for example, a TCI, a Cu pad, or the like. The connection unit 50 communicably connects the MPU 20 and the RAM unit 40. The connection unit 50 is connected to one end of a surface (upper surface) of the RAM unit 40 opposite to the surface (lower surface) placed on the interposer 10. In addition, the connection unit 50 is connected to one end of a surface (lower surface) of the MPU 20 facing the interposer 10. Specifically, the connection unit 50 is connected to a portion of the upper surface of the RAM unit 40 facing the MPU 20 and a portion of the lower surface of the MPU 20 facing the RAM unit 40. The connection unit 50 is disposed in each of the RAM units 40. For example, in the present embodiment, the connection units 50 are disposed between the four RAM units 40 and the MPU 20, respectively. In addition, the connection unit 50 is not limited to a unit that physically connects the MPU 20 and the RAM unit 40, but the connection unit 50 includes a unit that communicably connects the MPU 20 and the RAM unit 40 in a wireless manner (for example, TCI).
The spacer 60 is placed on the upper surface of the RAM unit 40. The spacer 60 is configured to have, for example, a rectangular shape in a plan view. The spacer 60 is made of, for example, silicon. The thickness of the spacer 60 is configured to be substantially equal to the thickness of the MPU 20 or larger than the thickness of the MPU 20. More preferably, the height from the upper surface of the interposer 10 to the upper surface of the spacer 60 is substantially equal to or is equal to the height from the upper surface of the interposer 10 to the upper surface of the MPU 20 connected to the interposer 10 by the pillars 30. The spacer 60 is disposed to be adjacent to the MPU 20 in a direction intersecting the lamination direction C of the RAM unit 40. In the present embodiment, the spacer 60 is disposed to be adjacent to the MPU 20 so as to interpose the side surface of the MPU 20. Among the ends of the spacer 60, the end opposite to the side facing the MPU 20 is disposed to protrude from the RAM unit 40 in the direction intersecting the lamination direction C of the RAM unit 40. Specifically, among the ends of the spacer 60, the end opposite to the side facing the MPU 20 is disposed to protrude from the RAM unit 40 in the direction opposite to the side facing the MPU 20.
In addition, in a case where there is no need for a gap between the MPU 20 and the RAM unit 40, the thickness of the spacer 60 is substantially equal to the thickness of the MPU 20. In this case, the connection unit 50 is mounted on the RAM unit 40 and the MPU 20. For example, in a case where the MPU 20 and the RAM unit 40 are connected by TCI or Cu hybrid bonding technique, the connection unit 50 is mounted on the RAM unit 40 and the MPU 20 as a coil (not illustrated) disposed inside the RAM unit 40 and the MPU 20 or a Cu pad (not illustrated) exposed on the upper surface of the RAM unit 40 and the lower surface of the MPU 20.
The support 70 is made of, for example, silicon. The support 70 is formed to have, for example, a substantially rectangular shape in a plan view. The support 70 is placed on the upper surface of the spacer 60 and the upper surface of the MPU 20. The support 70 is formed with such a size that can cover the spacer 60 and the MPU 20 in a plan view.
Next, operations of the semiconductor module 1 will be described. First, a power is supplied from the interposer 10 to the MPU 20. In addition, a power is supplied from the interposer 10 to the RAM unit 40. In addition, the MPU 20 is ground-connected to the interposer 10. The RAM unit 40 is ground-connected to the interposer 10. In addition, a power and a ground may be supplied from the MPU 20 to the RAM unit 40 via the connection unit 50 therebetween.
In a case where a data is stored in the RAM unit 40, first, the data is transmitted from the interposer 10 to the MPU 20 via the pillar 30. The MPU 20 transmits a calculation result calculated on the basis of the transmitted data to the RAM unit as a store signal. That is, the store signal transmitted from the MPU 20 is transmitted to the RAM unit 40 through the circuit surface of the MPU 20 and the connection unit 50. The RAM unit 40 stores the data included in the store signal on the basis of the address included in the store signal.
On the other hand, in a case where a data is loaded from the RAM unit 40, first, a load signal is transmitted from the interposer 10 to the MPU 20 via the pillar 30. That is, the load signal transmitted from the MPU 20 is transmitted to the RAM unit 40 through the circuit surface of the MPU 20 and the connection unit 50.
The RAM unit 40 loads a data from a corresponding address on the basis of the address included in the load signal. The RAM unit 40 transmits the loaded data to the MPU 20 via the connection unit 50.
Next, the structure of the semiconductor module 1 will be described. First, as illustrated in
In addition, as illustrated in
Next, the support 70 is prepared as illustrated in
According to the semiconductor module 1 according to the embodiment as described above, the following effects can be obtained.
(1) A semiconductor module includes a logic chip, a RAM unit 40 which is a lamination-type RAM module, a spacer 60 which is disposed to be laminated along a lamination direction of the RAM unit 40, an interposer 10 which is electrically connected to each of the logic chip and the RAM unit 40, and a connection unit 50 which communicably connects the logic chip and the RAM unit 40. The logic chip and the spacer 60 are disposed to be adjacent in a direction intersecting the lamination direction of the RAM unit 40, the RAM unit 40 is placed on the interposer 10, one end of is disposed so as to overlap with one end of the logic chip in the lamination direction, and the connection unit 50 connects one end of the RAM unit 40 and one end of the logic chip. Accordingly, since the MPU 20 and each of the pair of RAM units 40 can be directly connected by the connection unit 50, it is possible to shorten the signal line (the length of the connection unit 50) between the MPU 20 and each of the pair of RAM units 40. Therefore, the bandwidth between the MPU 20 and the pair of RAM units 40 can be increased.
(2) In the semiconductor module, a pair of the RAM unit 40 and the spacer 60 are provided with the logic chip interposed therebetween, and the connection unit 50 is provided for each RAM unit 40. Accordingly, since each RAM unit 40 is individually connected to the MPU 20 by the connection unit 50, a plurality of RAM units 40 can be easily connected to the MPU 20, and the capacity of the RAM unit 40 can be easily increased.
(3) The thickness of the spacer 60 is substantially equal to or larger than the thickness of the logic chip. Accordingly, it is possible to stably dispose the support 70 while connecting the lower surface of the MPU 20 to the upper surface of the RAM unit 40.
(4) Among the ends of the spacer 60, the end opposite to the side facing the logic chip is disposed to protrude from the RAM unit 40 in a direction intersecting the lamination direction of the RAM unit 40. Accordingly, since the exposed area of the spacer 60 is increased in comparison with the case where the side surface of the spacer 60 is flush with the side surface of the RAM unit 40, it is possible to improve the heat dissipation of the heat generated in the RAM unit 40. In addition, since a paste or the like for lamination can be applied to the entire surface of the RAM unit 40, the structure is stabilized, so that the inclination of the RAM unit 40 can be prevented.
(5) The semiconductor module further includes a plurality of pillars 30 which communicably connect the interposer 10 and the logic chip, each of the plurality of pillars 30 being longer than the thickness of the RAM unit 40 in the lamination direction. Accordingly, it is possible to dispose the position of the MPU 20 at a distance from the upper surface of the interposer 10 by the length of the pillar 30. Therefore, it is possible to allow a portion of the upper surface of the RAM unit to face a portion of the lower surface of the MPU 20, and thus, it is possible to shorten the signal line (the length of the connection unit 50).
Although the preferred embodiment of the semiconductor module according to the present invention has been described above, the present invention is not limited to the above-described embodiment, and the present invention can be appropriately modified.
For example, as illustrated in
In addition, in the above-described embodiment, a pair of the RAM unit 40 and the spacer 60 are described to be provided so as to interpose the MPU 20, but the present invention is not limited thereto. For example, the RAM unit 40 and the spacer 60 may be disposed on only one side of the MPU 20. In addition, the RAM unit 40 and the spacer 60 may be disposed on three sides of the MPU 20 or may be disposed on four sides so as to surround the MPU 20.
In addition, the arithmetic unit is not limited to the MPU 20 and may be widely applied to all logic chips. The memory is not limited to the DRAM, but the memory may be widely applied to a random access memory (RAM) including a nonvolatile RAM (for example, an MRAM, a ReRAM, an FeRAM, and the like).
Filing Document | Filing Date | Country | Kind |
---|---|---|---|
PCT/JP2017/041887 | 11/21/2017 | WO | 00 |