A semiconductor device, the device including: a first level including control circuits, where the control circuits include a plurality of first transistors and a plurality of metal layers; and a memory level disposed on top of the first level, where the memory level includes an array of memory cells, where each of the memory cells includes at least one second transistor, where the control circuits control access to the array of memory cells, where the first level is bonded to the memory level, where the bonded includes oxide to oxide bonding regions and a plurality of metal to metal bonding regions, and where at least a portion of the array of memory cells is disposed directly above at least one of the plurality of metal to metal bonding regions.

A semiconductor device, the device including: a first level including control circuits, where the control circuits include a plurality of first transistors and a plurality of metal layers; a memory level disposed on top of the first level, where the memory level includes an array of memory cells, where each of the memory cells include at least one second transistor, where the control circuits control the array of memory cells, where the first level is bonded to the memory level, where the bonded includes oxide to oxide bonding regions and a plurality of metal to metal bonding regions, and where at least one of the memory cells is disposed directly above at least one of the plurality of metal to metal bonding regions.

A memory circuit includes a first memory array and a second memory array. The first memory array and the second memory array are independent. The first memory array includes a plurality of general bits and the second memory array includes a plurality of spare bits. An address of defective bit in the first memory array is stored in the second memory array, and the memory circuit repairs the defective bit by one of the spare bits according to the address.

A memory circuit configured to perform multiply-accumulate (MAC) operations for performance of an artificial neural network includes a series of synapse cells arranged in a cross-bar array. Each cell includes a memory transistor connected in series with a memristor. The memory circuit also includes input lines connected to the source terminal of the memory transistor in each cell, output lines connected to an output terminal of the memristor in each cell, and programming lines coupled to a gate terminal of the memory transistor in each cell. The memristor of each cell is configured to store a conductance value representative of a synaptic weight of a synapse connected to a neuron in the artificial neural network, and the memory transistor of each cell is configured to store a threshold voltage representative of a synaptic importance value of the synapse connected to the neuron in the artificial neural network.

A location of at least one fail bit to be repaired in a memory block of a memory is extracted from at least one memory test on the memory block. An available repair resource in the memory for repairing the memory block is obtained. It is checked, using machine learning, whether the at least one fail bit is unrepairable, according to the location of the at least one fail bit, and the available repair resource. When the checking indicates that the at least one fail bit is not unrepairable, it is determined whether a Constraint Satisfaction Problem (CSP) containing a plurality of constraints is solvable. The constraints correspond to the location of the at least one fail bit in the memory block, and the available repair resource. In response to determining that the CSP is not solvable, the memory block is marked as unrepairable or the memory is rejected.

A memory device includes a memory cell array including normal memory cells and redundant memory cells; first page buffers connected to the normal memory cells through first bit lines including a first bit line group and a second bit line group and arranged in a first area corresponding to the first bit lines in a line in a first direction; and second page buffers connected to the redundant memory cells through second bit lines including a third bit line group and a fourth bit line group and arranged in a second area corresponding to the second bit lines in a line in the first direction, wherein, when at least one normal memory cell connected to the first bit line group is determined as a defective cell, normal memory cells connected to the first bit line group are replaced with redundant memory cells connected to the third bit line group.

A memory circuit configured to perform multiply-accumulate (MAC) operations for performance of an artificial neural network includes a series of synapse cells arranged in a cross-bar array. Each cell includes a memory transistor connected in series with a memristor. The memory circuit also includes input lines connected to the source terminal of the memory transistor in each cell, output lines connected to an output terminal of the memristor in each cell, and programming lines coupled to a gate terminal of the memory transistor in each cell. The memristor of each cell is configured to store a conductance value representative of a synaptic weight of a synapse connected to a neuron in the artificial neural network, and the memory transistor of each cell is configured to store a threshold voltage representative of a synaptic importance value of the synapse connected to the neuron in the artificial neural network.

Representative implementations of devices and techniques provide correction for a defective die in a wafer-to-wafer stack or a die stack. A correction die is coupled to a die of the stack with the defective die. The correction die electrically replaces the defective die. Optionally, a dummy die can be coupled to other die stacks of a wafer-to-wafer stack to adjust a height of the stacks.

Embodiments provide a method and an apparatus for repairing a fail location. When repairing a fail location of a wafer, a fail bit in a wafer to be processed may be first determined, and a target potential fail bit associated with the fail bit may be determined based on a potential mining rule included in a mining rule library.

A memory device includes a memory cell array including normal memory cells and redundant memory cells; first page buffers connected to the normal memory cells through first bit lines including a first bit line group and a second bit line group and arranged in a first area corresponding to the first bit lines in a line in a first direction; and second page buffers connected to the redundant memory cells through second bit lines including a third bit line group and a fourth bit line group and arranged in a second area corresponding to the second bit lines in a line in the first direction, wherein, when at least one normal memory cell connected to the first bit line group is determined as a defective cell, normal memory cells connected to the first bit line group are replaced with redundant memory cells connected to the third bit line group.