Semiconductor equipment
The semiconductor device integrates error detection and correction within its layers, efficiently detecting and correcting errors using dual-core lockstep and error correction codes, reducing resource usage and power consumption.
Patent Information
- Application Number
- JP2025077900
- Authority / Receiving Office
- JP · JP
- Patent Type
- Applications
- Current Assignee / Owner
- Filing Date
- 2025-05-08
- Publication Date
- 2026-11-19
AI Technical Summary
Conventional semiconductor devices require additional components like a check memory to identify and correct memory errors, which is inefficient and resource-intensive.
A semiconductor device with a stacked structure that integrates error detection and correction units within the logic and semiconductor layers, allowing for in-line detection and correction of errors without additional components, using dual-core lockstep and error correction codes like Hamming codes.
Efficient and timely detection and correction of errors with reduced power consumption by identifying error blocks based on positional information and performing correction only where needed, eliminating the need for separate error detection components.
Smart Images

Figure 2026182016000001_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present disclosure relates to a semiconductor device. BACKGROUND ART
[0002] Patent Document 1 discloses a semiconductor device configured with a two-layer structure of a data memory and a check memory. Such a semiconductor device identifies a memory error occurrence block in the data memory based on error information from the check memory. PRIOR ART DOCUMENTS PATENT DOCUMENTS
[0003] PATENT DOCUMENT 1 Japanese Unexamined Patent Application Publication No. 11-284147 SUMMARY OF THE INVENTION PROBLEMS TO BE SOLVED BY THE INVENTION
[0004] In conventional semiconductor devices, to identify a memory error occurrence block, a check memory is added as a separate component from the semiconductor layer that is the target of error occurrence identification and correction. However, there is a demand for a technology capable of executing processing for identifying and correcting the memory error occurrence block in the data memory without adding such a separate component. This problem is common even when executing processing for correcting errors that may occur in a semiconductor layer in a semiconductor device including any semiconductor layer other than the data memory. MEANS FOR SOLVING THE PROBLEMS
[0005] The present disclosure can be implemented in the following modes.
[0006] According to one embodiment of the present disclosure, a semiconductor device (100, 100a, 100b, 100c, or 100d) having a stacked structure is provided. The semiconductor device comprises a logic layer (20) including a first logic region (21) on which a first logical operation is performed, and a semiconductor layer (40) including a semiconductor region (41) which is a region that includes at least one of a second logic region (42) on which a second logical operation is performed and a memory region (43) on which data is recorded. The logic layer has an error detection unit (25) that performs an error detection process to detect a first error (ER1) in the first logic region, and the semiconductor layer has an error correction unit (45) that performs an error correction process to correct a second error (ER2) that may occur in the semiconductor region. The error correction unit performs the error correction process on a target region (TA), which is at least a part of the semiconductor region, when the first error is detected in the first logic region.
[0007] In this type of semiconductor device, the logic layer has an error detection unit that performs error detection processing to detect a first error in the first logic region, so the error detection unit can detect a first error in the first logic region. Furthermore, the error correction unit performs error correction processing on a target region, which is at least a part of the semiconductor region, when a first error is detected in the first logic region. Therefore, error correction processing can be performed on the target region without adding any additional configurations for correcting errors that may occur in the semiconductor region, separate from the logic layer and the semiconductor layer.
[0008] Another embodiment of the present disclosure provides a semiconductor device (100, 100a, 100b, 100c, or 100d) having a stacked structure. This semiconductor device includes semiconductor layers (40, 40a) including a semiconductor region (41) which is a region that includes at least one of a logic region (42) on which logical operations are performed and a memory region (43) on which data is recorded, The logic layer comprises a logic layer (20, 20b) including an interposer region (22) electrically connected to the semiconductor layer, wherein the logic layer has an error detection unit (25) that performs error detection processing to detect a first error (ER1) in the interposer region. The semiconductor layer has an error correction unit (45, 45a) that performs an error correction process to correct a second error (ER2) that may occur in the semiconductor region, and the error correction unit performs the error correction process on a target region (TA), which is at least a part of the semiconductor region, when the first error is detected in the interposer region.
[0009] According to this type of semiconductor device, the logic layer including the interposer region has an error detection unit that performs error detection processing to detect a first error in the interposer region, so the error detection unit can detect a first error in the interposer region. [Brief explanation of the drawing]
[0010] [Figure 1] This is a block diagram showing the schematic configuration of a semiconductor device in the first embodiment. [Figure 2] This diagram schematically illustrates how an error occurs in the semiconductor device according to the first embodiment. [Figure 3] This is a flowchart showing the error detection and correction process in the first embodiment. [Figure 4] This block diagram shows the schematic configuration of a semiconductor device in the second embodiment. [Figure 5] This diagram schematically illustrates how an error occurs in the semiconductor device according to the second embodiment. [Figure 6] This is a flowchart showing the error detection and correction process in the second embodiment. [Figure 7] This is a block diagram showing the schematic configuration of the semiconductor device in the third embodiment. [Figure 8]This diagram schematically illustrates how an error occurs in the semiconductor device according to the third embodiment. [Figure 9] This is a flowchart showing the error detection and correction process in the third embodiment. [Figure 10] This diagram schematically illustrates how an error can occur in a semiconductor device in another embodiment. [Figure 11] This diagram schematically illustrates how an error can occur in a semiconductor device in another embodiment. [Modes for carrying out the invention]
[0011] A. First Embodiment: A1. Overall configuration of semiconductor device 100: The semiconductor device 100 shown in Figure 1 has a stacked structure of a logic layer 20 and a semiconductor layer 40. The semiconductor device 100 is formed by so-called heterogeneous integration, which integrates multiple semiconductor chips of different types and houses them in a single package. The logic layer 20 and the semiconductor layer 40 are electrically connected through a connector 90. The connector 90 is configured, for example, as a through-silicon via (TSV).
[0012] The logic layer 20 includes a first logic region 21 where logical operations are performed. The first logic region 21 is configured, for example, as a CPU (Central Processing Unit) and has multiple processor cores CR. In the first embodiment, the first logic region 21 has two processor cores CR. Each processor core CR performs logical operations in the first logic region 21. Hereinafter, the logical operations performed by each processor core CR in the first logic region 21 will be referred to as the "first logical operation." The first logical operation is performed as part of the main processing performed by the semiconductor device 100. In other words, the logic layer 20, which includes the first logic region 21 where the first logical operation is performed, is a necessary configuration for the semiconductor device 100 to function.
[0013] The semiconductor layer 40 includes a semiconductor region 41. In the first embodiment, the semiconductor region 41 includes a memory region 43 in which data is recorded. The memory region 43 includes an area configured as a RAM (Random Access Memory). The RAM is, for example, a DRAM (Dynamic RAM) or an SRAM (Static RAM).
[0014] As an example, the semiconductor device 100 is used in applications requiring high reliability, such as servers and in-vehicle devices. Therefore, in the semiconductor device 100, it is necessary to take appropriate measures against soft errors. A "soft error" is a phenomenon in which neutrons and the like contained in natural radiation NR such as cosmic rays collide with a semiconductor device, causing a temporary malfunction or failure in the semiconductor device. Soft errors are broadly classified into two types: SEU (Single Event Upset) and SET (Single Event Transient). SET is a phenomenon in which a pulse-like bit inversion occurs in the output of a logic gate. SEU is a phenomenon in which data held in a storage element such as a DRAM is inverted. In the description of soft errors to be given later, SET corresponds to a soft error that occurs in the first logic region 21, and SEU corresponds to a soft error that occurs in the memory region 43. Note that in the present disclosure, a soft error is simply referred to as an "error".
[0015] In order to appropriately cope with errors, the logic layer 20 constituting the semiconductor device 100 has an error detection unit 25 that executes error detection processing. Further, the semiconductor layer 40 constituting the semiconductor device 100 has an error correction unit 45 that executes error correction processing. Hereinafter, an error detected in the first logic region 21 is referred to as a "first error ER1", and an error that may occur in the semiconductor region 41 is referred to as a "second error ER2" to be distinguished from the first error ER1.
[0016] The error detection unit 25 executes error detection processing for detecting the first error ER1 in the first logic region 21. In the first embodiment, the first error ER1 is detected by so-called dual-core lockstep. In dual-core lockstep, first, each processor core CR executes the first logical operation processing, which is the same logical operation processing. Next, the error detection unit 5 executes error detection processing by comparing the processing results of the first logical operation processing by each processor core CR with the comparison circuit CP. Thereby, the first error ER1 in the first logic region 21 can be detected.
[0017] The error detection processing described above is executed for each predetermined block of the first logic region 21. Therefore, the error detection unit 25 can identify the error occurrence block in which the first error ER1 is detected. Thereby, the error detection unit 25 can acquire position-related information regarding the physical position where the first error ER1 is detected. The "position-related information" is information including information related to the physical position within the first logic region 21 where the first error ER1 is detected.
[0018] As shown in Figure 2, when natural radiation NR collides with the logic layer 20 located on the upper layer side relative to the semiconductor layer 40, the natural radiation NR may cause the first error ER1 in the logic layer 20. At this time, after colliding with the logic layer 20, the natural radiation NR passes through the logic layer 20, and may cause the second error ER2 in the semiconductor layer 40 located on the lower layer side relative to the logic layer 20. That is, at the timing when the first error ER1 occurs, the second error ER2 may also occur. Therefore, the error correction unit 45 executes error correction processing for correcting an error that may occur in the memory region 43, triggered by the detection of the first error ER1 in the first logic region 21. Note that in Figure 2, for the sake of simplifying the description, the package substrate, various wirings and the like constituting the semiconductor device 100 are omitted from the drawing.
[0019] As shown in Figure 2, the error correction process is performed on the target region TA, which is at least a portion of the memory region 43. The target region TA is described below. As mentioned above, natural radiation NR, after colliding with the logic layer 20, can penetrate the logic layer 20 and potentially cause a second error ER2 in the semiconductor layer 40. Therefore, the location where the second error ER2 occurs is likely to be close to the error-occurring block where the first error ER1 was detected. Accordingly, the error correction unit 45 identifies the target region TA based on position-related information regarding the physical location where the first error ER1 was detected in the first logic region 21. More specifically, the error correction unit 45 identifies a predetermined region of the memory region 43 as the target region TA, which includes a point that overlaps with the error-occurring block of the first error ER1 in the stacking direction.
[0020] The error correction unit 45 includes an ECC processing unit 46 and a rereading unit 47. The ECC processing unit 46 performs ECC (Error Check and Correct) processing on the target area TA when a first error ER1 is detected in the first logic area 21. More specifically, the ECC processing unit 46 detects and corrects a second error ER2 in the memory area 43 using an error correction code such as a Hamming code. The ECC processing is performed by operating an ECC arithmetic circuit (not shown). Various codes can be used as error correction codes, such as error-detecting codes (EDC), horizontal and vertical parity codes, Reed-Solomon codes, and BCH codes.
[0021] If the detected second error ER2 is a 1-bit error, the second error ER2 can be corrected by bit-inverting it. On the other hand, if the detected second error ER2 is a 2-bit or larger error, the ECC processing unit 46 cannot correct the second error ER2. Therefore, in this case, the re-read unit 47 performs a re-read process, which is the process of re-reading the data recorded in the memory area 43, as an error correction process. This allows the re-read unit 47 to correct the second error ER2. The re-read process is performed, for example, by re-reading the data stored in an auxiliary memory area (not shown) configured as ROM (Read Only Memory).
[0022] A2. Error detection and correction process: The error detection and correction process shown in Figure 3 refers to the process from the detection of the first error ER1 in the first logic region 21 until the error correction process is executed on the target region TA. The error detection and correction process is started when the power of the semiconductor device 100 is turned on and is then executed repeatedly.
[0023] In step S105, each processor core CR executes a first logical operation in the first logic region 21. Hereafter, "step S" will be simply referred to as "S". The first logical operation is repeatedly executed at predetermined time intervals based on the clock frequency of each processor core CR. The result of each operation is output to the comparison circuit CP of the error detection unit 25.
[0024] In S110, the error detection unit 25 determines whether the processing results of the first logical operation performed by each processor core CR are different. If the processing results are the same (S110: No), the process returns to S105. In other words, in this case, it can be determined that the first error ER1 has not been detected in the first logic area 21.
[0025] If the processing result differs (S110: Yes), it can be determined that the first error ER1 was detected in the first logic area 21. Therefore, in this case, in S115, the error correction unit 45 identifies the target area TA on which to perform error correction processing. The error correction unit 45 identifies the target area TA based on position-related information regarding the physical location where the first error ER1 was detected in the first logic area 21. More specifically, the error correction unit 45 identifies a predetermined area within the memory area 43 that includes a point that overlaps with the error-occurring block of the first error ER1 in the stacking direction as the target area TA.
[0026] In S120, the ECC processing unit 46 performs ECC processing on the target region TA. More specifically, the ECC processing unit 46 detects a second error ER2 in the memory region 43 using an error correction code such as a Hamming code. The ECC processing is performed by operating an ECC arithmetic circuit (not shown).
[0027] In S125, the ECC processing unit 46 determines whether or not a second error ER2 was detected in the target area TA. If no second error ER2 was detected (S125: No), the process returns to S105. In other words, in this case, it can be determined that no second error ER2 occurred in the memory area 43.
[0028] If a second error ER2 is detected in the target region TA (S125: Yes), in S130, the ECC processing unit 46 determines whether the detected second error ER2 is a 1-bit error. If the detected second error ER2 is a 1-bit error (S130: Yes), in S135, the ECC processing unit 46 performs an error correction process by bit-inverting the detected 1-bit second error ER2. This corrects the second error ER2.
[0029] On the other hand, if the detected second error ER2 is a 2-bit or greater error (S130: No), the ECC processing unit 46 cannot correct the second error ER2. Therefore, in this case, in S140, the re-read unit 47 executes a re-read process, which is the process of re-reading the data recorded in the memory area 43, as an error correction process. The re-read process is performed, for example, by reading the data stored in the auxiliary memory area configured as ROM. This corrects the second error ER2. In other words, in the first embodiment, the re-read unit 47 executes the re-read process on the condition that the second error ER2 detected by the ECC process is a 2-bit or greater error.
[0030] As described above, the semiconductor device 100 has an error detection unit 25 that performs error detection processing to detect a first error ER1 in the first logic region 21, so the error detection unit 25 can detect the first error ER1 in the first logic region 21. Furthermore, the error correction unit 45 performs error correction processing on a target region TA, which is at least a part of the semiconductor region 41, when the first error ER1 is detected in the first logic region 21. Therefore, error correction processing can be performed on the target region TA without adding a separate configuration for correcting a second error ER2 that may occur in the semiconductor region 41, apart from the logic layer 20 and the semiconductor layer 40. In addition, the error correction unit 45 can perform error correction processing at an appropriate timing triggered by the detection of the first error ER1 in the first logic region 21. For this reason, the power consumption of the error correction unit 45 can be reduced compared to a configuration in which error correction processing is performed repeatedly at regular intervals.
[0031] Furthermore, the error correction unit 45 identifies the target area TA based on position-related information regarding the physical location where the first error ER1 was detected in the first logic area 21. Therefore, it can identify the area in the memory area 43 that is highly likely to have a second error ER2 as the target area TA. In addition, since the error correction unit 45 performs error correction processing on the target area TA, compared to a configuration in which error correction processing is performed on the entire memory area 43, it is possible to detect and correct the second error ER2 in a shorter time and to reduce power consumption when detecting and correcting the second error ER2.
[0032] Furthermore, the error detection unit 25 performs error detection by comparing the processing results of the first logical operation performed by multiple processor cores CR, so the first error ER1 can be detected by a configuration using multiple processor cores CR. In addition, compared to a conventional configuration in which errors are detected by check memory, such a configuration may allow for the detection of the first error ER1 in a shorter time and may reduce power consumption when detecting the first error ER1.
[0033] Furthermore, the reread unit 47 performs the reread process, which is the process of rereading the data recorded in the memory area 43, as an error correction process. By performing the reread process as an error correction process, it is possible to appropriately correct the second error ER2 that may occur in the memory area 43.
[0034] Furthermore, if the ECC processing unit 46 detects a second error ER2 through ECC processing, it performs error correction processing on the detected second error ER2. By performing ECC processing as error correction processing, the second error ER2 that may occur in the memory area 43 can be appropriately corrected. In addition, in normal ECC processing, error correction processing is performed on the data to be used when data is read from the memory area 43. On the other hand, according to the semiconductor device 100, error correction processing can be performed on data recorded in the memory area 43 before it is read and used.
[0035] Furthermore, the rereading unit 47 executes the rereading process on the condition that the second error ER2 detected by the ECC process is 2 bits or more, so the second error ER2 that cannot be corrected by the ECC process can be appropriately corrected by the rereading process.
[0036] B. Second Embodiment: As shown in Figure 4, the semiconductor device 100a in the second embodiment differs from the first embodiment in that the semiconductor region 41 includes a second logic region 42 instead of a memory region 43. The same reference numerals are used for components and processes that are the same as in the first embodiment, and their detailed descriptions are omitted.
[0037] In the second embodiment, the semiconductor layer 40a includes a second logic region 42 in which logical operations are performed. The second logic region 42 is configured, for example, as a separate CPU from the first logic region 21.
[0038] The error correction unit 45a has a reprocessing unit 48. The reprocessing unit 48 is configured, for example, as a processor core that executes logical operations. Hereinafter, the logical operations executed by the reprocessing unit 48 in the second logic area 42 will be referred to as the "second logical operations" to distinguish them from the first logical operations. The reprocessing unit 48 executes an error correction process that re-executes at least a part of the second logical operations in the second logic area 42. Whether to re-execute all or part of the second logical operations may be predetermined, or it may be determined according to the processing status of the second logical operations by the reprocessing unit 48.
[0039] As shown in Figure 5, the reprocessing unit 48 performs error correction processing with the entire area of the second logic region 42 as the target region TA. In other words, the reprocessing unit 48 performs error correction processing regardless of the location-related information regarding the physical location where the first error ER1 was detected in the first logic region 21.
[0040] In the error detection and correction process shown in Figure 6, if the processing results of the first logical operation by each processor core CR differ (S110: Yes), in S112, the reprocessing unit 48 re-executes at least a portion of the second logical operation in the second logic area 42. This corrects the second error ER2 that may occur in the second logic area 42. The second error ER2 that may occur in the second logic area 42 is classified as SET, one of the SEU and SET mentioned above.
[0041] The semiconductor device 100a described above provides the same effects as the semiconductor device 100 of the first embodiment. Furthermore, since the reprocessing unit 48 re-executes at least a portion of the second logical operation processing in the second logic region 42, it can perform error correction processing to correct a second error ER2 that may occur in the second logic region 42, which is separate from the first logic region 21, triggered by the detection of a first error ER1 in the first logic region 21. In addition, the semiconductor device 100a eliminates the need to perform processing to identify the target region TA.
[0042] C. Third Embodiment: As shown in Figure 7, the semiconductor device 100b in the third embodiment differs from the first embodiment in its configuration for performing error detection processing. In addition, the semiconductor device 100b differs from the first embodiment in that the ECC processing unit 46 is omitted. Components and processing contents identical to those in the first embodiment are denoted by the same reference numerals, and their detailed descriptions are omitted.
[0043] In the third embodiment, the logic layer 20b includes an interposer region 22 in addition to the first logic region 21. The interposer region 22 is a region where various wirings (not shown) are formed for electrical connection with the semiconductor layer 40. The interposer region 22 is configured as a so-called silicon interposer, for example, using silicon as the material. The interposer region 22 has an error detection unit 25b. The error detection unit 25b has a comparator circuit CP in addition to a plurality of arithmetic circuits AC having the same configuration as each other. In the third embodiment, the number of arithmetic circuits AC in the error detection unit 25b is two.
[0044] In the third embodiment, the first error ER1 is detected by a circuit configuration known as DMR (Dual Modular Redundancy). In the DMR, first, the processor core CR in the first logic region 21 performs a first logical operation using each arithmetic circuit AC. Next, the error detection unit 25b performs an error detection process by comparing the calculation results using each arithmetic circuit AC.
[0045] As shown in Figure 7, the error correction unit 45 in the third embodiment has a reload unit 47, but unlike the first embodiment, the ECC processing unit 46 is omitted. As shown in Figure 8, the reload unit 47 performs error correction processing with the entire memory area 43 as the target area TA. That is, the reload unit 47 performs error correction processing regardless of the position-related information regarding the physical location where the first error ER1 was detected in the interposer area 22.
[0046] In the third embodiment, unlike the first and second embodiments, as shown in Figure 8, the semiconductor layer 40 is located above the logic layer 20b. Furthermore, the interposer region 22 of the logic layer 20b is located below the first logic region 21.
[0047] In the error detection and correction process shown in Figure 9, at S107, the processor core CR performs calculations using each arithmetic circuit AC, which has the same configuration as the others.
[0048] In S108, the error detection unit 25b compares the calculation results obtained using each calculation circuit AC as an error detection process and determines whether the calculation results differ from each other.
[0049] If the calculation results using each arithmetic circuit AC match (S108: No), the process returns to S107. In other words, in this case, it can be determined that the first error ER1 in the interposer area 22 has not been detected.
[0050] On the other hand, if the calculation results obtained using each arithmetic circuit AC differ (S108: Yes), in S140, the reread unit 47 performs a reread process of the memory area 43. This corrects the second error ER2 that may occur in the memory area 43.
[0051] The semiconductor device 100b described above provides the same effects as the semiconductor device 100 of the first embodiment. Furthermore, since the interposer region 22 has an error detection unit 25b, the first error ER1 in the logic layer 20b can be detected without providing an error detection unit 25b in the first logic region 21. Therefore, an area can be secured in the first logic region 21 for functions other than detecting the first error ER1.
[0052] Furthermore, the error detection unit 25b can detect the first error ER1 by comparing the calculation results of multiple arithmetic circuits AC by a single processor core CR in the first logic region 21, thus enabling the detection of the first error ER1 in the logic layer 20b without using multiple processor cores CR.
[0053] Furthermore, the error correction unit 45 has a reread unit 47, while the ECC processing unit 46 in the first embodiment is omitted, so error correction processing can be performed in a configuration that omits the ECC processing unit 46. In addition, in the semiconductor device 100b of the third embodiment, the processing for identifying the target region TA can be omitted.
[0054] D. Other embodiments: (D1) In the first to third embodiments, the logic layers 20 and 20b included a first logic region 21 in which logical arithmetic processing is performed, but the disclosure is not limited thereto. The logic layers 20 and 20b may include an interposer region 22 instead of the first logic region 21. Even in this case, in a configuration in which the interposer region 22 has an error detection unit 25b, the error detection unit 25b can perform error detection processing, for example, by comparing the calculation results using each arithmetic circuit AC.
[0055] (D2) In the first to third embodiments, the semiconductor devices 100, 100a, and 100b each comprised one semiconductor layer 40, but the disclosure is not limited thereto. For example, as shown in Figure 10, the semiconductor device 100c may include two or more semiconductor layers 40. In the example shown in Figure 10, each semiconductor layer 40 includes a memory area 43. In this case, for example, each memory area 43 has an error correction unit 45, and when a first error ER1 is detected in the first logic area 21, each error correction unit 45 may perform error correction processing on each memory area 43.
[0056] (D3) In the first to third embodiments, the semiconductor region 41 included either the second logic region 42 or the memory region 43, but the disclosure is not limited thereto. The semiconductor region 41 may include both the second logic region 42 and the memory region 43. In the example shown in Figure 11, the semiconductor device 100d has two systems, S1 and S2, which function independently within the semiconductor device 100d. In system S1, the semiconductor layer 40 located below the logic layer 20 includes the memory region 43. In system S2, there is a structure in which two semiconductor layers 40 are stacked. Of these two semiconductor layers 40, the upper semiconductor layer 40 includes the second logic region 42, and the lower semiconductor layer 40 includes the memory region 43. As an example, each of the three semiconductor layers 40 described above has an error correction unit 45. Even with this configuration, as shown in Figure 11, when a first error ER1 is detected in the first logic region 21, each error correction unit 45 can perform error correction processing on the second logic region 42 or memory region 43 in each semiconductor layer 40. In other words, in this disclosure, the semiconductor region 41 is a region that includes at least one of the second logic region 42 and the memory region 43.
[0057] (D4) In the first embodiment, the error correction unit 45 had a reread unit 47, but this disclosure may omit it. Even in this case, if the second error ER2 detected by the ECC process is a 1-bit error, the second error ER2 can be appropriately corrected.
[0058] (D5) In the first embodiment, the error correction unit 45 has an ECC processing unit 46 and a rereading unit 47, and detects or corrects the second error ER by ECC processing and rereading processing, but the disclosure is not limited thereto. Various configurations may be adopted as the configuration for detecting or correcting the second error ER2, such as a parity check with a configuration that adds a parity bit, a checksum, or a CRC (Cyclic Redundancy Check).
[0059] (D6) In the first and second embodiments, the number of processor cores CR in the first logic area 21 was two, but the disclosure is not limited thereto. The number of processor cores CR in the first logic area 21 may be three or more, in which case error detection processing may be performed by comparing the processing results of the first logical operations performed by the three or more processor cores CR.
[0060] (D7) In the second embodiment, the reprocessing unit 48 performed error correction processing with the entire area of the second logic area 42 as the target area TA, but the disclosure is not limited thereto. The reprocessing unit 48 may perform error correction processing by performing reload processing with a portion of the area of the second logic area 42, for example, a portion of the area constituting the processor core CR in the second logic area 42 as the target area TA.
[0061] (D8) In the third embodiment, the reload unit 47 performed error correction processing with the entire memory area 43 as the target area TA, but the disclosure is not limited thereto. The reload unit 47 may also perform error correction processing by reading some data from an auxiliary memory area (not shown) and reloading a portion of the memory area 43 as the target area TA.
[0062] (D9) In the third embodiment, the error detection unit 25b had two arithmetic circuits AC, but the disclosure is not limited thereto. The error detection unit 25b may have three or more arithmetic circuits AC. For example, if the error detection unit 25b has three arithmetic circuits AC, the first error ER1 may be detected using a circuit configuration of a majority voting circuit based on so-called TMR (Triple Modular Redundancy).
[0063] (D10) In the third embodiment, the interposer region 22 had an error detection unit 25b having a plurality of arithmetic circuits AC, but the disclosure is not limited thereto. Alternatively, the first logic region 21 of the logic layer 20b may have an error detection unit 25b having a plurality of arithmetic circuits AC. The logic layer 20b may also have a plurality of error detection units 25b. In this case, both the first logic region 21 and the interposer region 22 may each have an error detection unit 25b.
[0064] (D11) In the third embodiment, the interposer region 22 of the logic layer 20b was located on the lower side relative to the first logic region 21, but the disclosure is not limited thereto. The interposer region 22 may be located on the upper side relative to the first logic region 21. Alternatively, the semiconductor device 100b may have a layer structure in which the semiconductor region 41 constituting the semiconductor layer 40 is located between the first logic region 21 constituting the logic layer 20b and the interposer region 22, that is, a layer structure in which the first logic region 21, the semiconductor region 41, and the interposer region 22 are arranged in this order from top to bottom. Alternatively, the semiconductor device 100b may have a layer structure in which the first logic region 21 and the semiconductor region 41 are arranged on the upper side of the interposer region 22 without being stacked.
[0065] (D12) In the third embodiment, the interposer region 22 was configured as a so-called silicon interposer, but the disclosure is not limited thereto. The interposer region 22 may be an interposer made of various materials, such as a glass interposer using glass, or an organic interposer using organic materials.
[0066] This disclosure is not limited to the embodiments described above, and can be implemented in various configurations without departing from its spirit. For example, the technical features in the embodiments corresponding to the technical features in each form described in the summary of the invention can be replaced or combined as appropriate in order to solve some or all of the above-described problems, or to achieve some or all of the above-described effects. Furthermore, if a technical feature is not described as essential in this specification, it can be deleted as appropriate. [Explanation of Symbols]
[0067] 20, 20b…Logic layer, 21…First logic area, 22…Interposer area, 25, 25b…Error detection unit, 40, 40a…Semiconductor layer, 41…Semiconductor area, 42…Second logic area, 43…Memory area, 45, 45a…Error correction unit, 46…ECC processing unit, 47…Reload unit, 48…Reprocessing unit, 90…Connection unit, 100, 100a, 100b, 100c, 100d…Semiconductor device, AC…Arithmetic circuit, CP…Comparison circuit, ER1…First error, ER2…Second error, NR…Natural radiation, CR…Processor core, S1…System, S2…System, TA…Target area
Claims
1. A semiconductor device having a stacked structure (100, 100a, 100b, 100c, or 100d), A logic layer (20, 20b) including a first logic region (21) where a first logical operation is performed, A semiconductor layer (40, 40a) including a semiconductor region (41) which is a region that includes at least one of a second logic region (42) where a second logical operation is performed and a memory region (43) where data is recorded, Equipped with, The logic layer includes an error detection unit (25) that performs error detection processing to detect a first error (ER1) in the first logic region. The semiconductor layer has an error correction unit (45, 45a) that performs an error correction process to correct a second error (ER2) that may occur in the semiconductor region. The error correction unit performs the error correction process on a target region (TA), which is at least a part of the semiconductor region, when the first error is detected in the first logic region.
2. A semiconductor device according to claim 1, The error correction unit identifies the target region based on position-related information relating to the physical location where the first error was detected in the first logic region, and is a semiconductor device.
3. A semiconductor device according to claim 1, The first logic region has a plurality of processor cores (CRs) that perform the first logical operation processing, The error detection unit is a semiconductor device that performs the error detection process by comparing the processing results of the first logical operation processes performed by the plurality of processor cores.
4. A semiconductor device according to claim 1, The error detection unit has a plurality of arithmetic circuits (ACs) having the same configuration as each other, and performs the error detection process by comparing the calculation results obtained using the plurality of arithmetic circuits.
5. A semiconductor device according to claim 4, The logic layer further includes an interposer region (22) that is electrically connected to the semiconductor layer, The interposer region is a semiconductor device having the error detection unit.
6. A semiconductor device according to claim 1, The semiconductor region includes the memory region, The semiconductor device has a rereading unit (47) that performs a rereading process, which is a process of rereading the data recorded in the memory area, as the error correction process.
7. A semiconductor device according to claim 6, The error correction unit has an ECC processing unit (46) that performs ECC (Error Check and Correct) processing on the target area. The ECC processing unit, when the second error is detected by the ECC processing, executes the error correction process on the detected second error, in a semiconductor device.
8. A semiconductor device according to claim 7, The re-reading unit is a semiconductor device that executes the re-reading process on the condition that the second error detected by the ECC process is an error of 2 bits or more.
9. A semiconductor device according to claim 1, The semiconductor region includes the second logic region, The semiconductor device has an error correction unit which includes a reprocessing unit (48) that performs a process to re-execute at least a part of the second logical operation process in the second logic region as the error correction process.
10. A semiconductor device having a stacked structure (100, 100a, 100b, 100c, or 100d), A semiconductor layer (40, 40a) including a semiconductor region (41) which is a region that includes at least one of the logic region (42) where logical operations are performed and the memory region (43) where data is recorded, A logic layer (20, 20b) including an interposer region (22) electrically connected to the semiconductor layer, Equipped with, The logic layer includes an error detection unit (25) that performs error detection processing to detect a first error (ER1) in the interposer region. The semiconductor layer has an error correction unit (45, 45a) that performs an error correction process to correct a second error (ER2) that may occur in the semiconductor region. The error correction unit performs the error correction process on a target region (TA), which is at least a part of the semiconductor region, when the first error is detected in the interposer region.
Citation Information
Patent Citations
Memory circuit
JP1999284147A
