A 3D semiconductor device, the device comprising: a first level comprising a first single crystal layer, said first level comprising first transistors, wherein each of said first transistors comprises a single crystal channel; first metal layers interconnecting at least said first transistors; a second metal layer overlaying said first metal layers; and a second level comprising a second single crystal layer, said second level comprising second transistors, wherein said second level overlays said first level, wherein at least one of said second transistors comprises a gate all around structure, wherein said second level is directly bonded to said first level, and wherein said bonded comprises direct oxide to oxide bonds.

A method includes providing a substrate having an epitaxial stack of layers including a plurality of semiconductor channel layers interposed by a plurality of dummy layers. The substrate includes a first device region and a second device region. An etch process is performed to remove a first portion of the epitaxial stack of layers from the second device region to form a trench in the second device region. The removed first portion of the epitaxial stack of layers includes at least one semiconductor channel layer of the plurality of semiconductor channel layers. An epitaxial layer is formed within the trench in the second device region and over the second portion of the epitaxial stack of layers. A top surface of the epitaxial layer in the second device region is substantially level with a top surface of the epitaxial stack of layers in the first device region.

An integrated circuit includes at least one first area including logic circuitry. The logic circuitry includes library blocks selected from a logic circuit library. A first one of the library blocks is provided with at least two symmetry mirror edges perpendicular to a height of the library blocks. Two adjacent ones of the library blocks are joined at a common symmetry mirror edge.

A SRAM device includes a substrate, at least one two-transistor static random access memory (2T-SRAM), an inner dielectric layer, a plurality of contacts, an inter-layer dielectric (ILD) layer, a plurality of vias, and a conductive line. The 2T-SRAM is disposed on the substrate, the inner dielectric layer covers the 2T-SRAM, and the contacts are disposed in the inner dielectric layer and coupled to the 2T-SRAM. The ILD layer covers the inner dielectric layer and the contacts, and the vias are disposed in the ILD layer and respectively coupled to the 2T-SRAM trough the corresponding contacts. The conductive line is disposed on the ILD layer and connects with the plurality of vias, wherein the thickness of the conductive line is less than or equal to one-tenth of the thickness of the via such that it can significantly reduce the coupling effect compared with the traditional bit line.

A semiconductor device includes a first transistor, a second transistor and a third transistor provided on a substrate, the first to third transistors respectively including source and drain regions spaced apart from each other, a gate structure extending in a first direction on the substrate and interposed between the source and drain regions, and a channel region connecting the source and drain regions to each other. A channel region of the second transistor and a channel region of the third transistor respectively include a plurality of channel portions, the plurality of channel portions spaced apart from each other in a second direction, perpendicular to an upper surface of the substrate, and connected to the source and drain regions, respectively. A width of a channel portion of the third transistor in the first direction is greater than a width of a channel portion of the second transistor in the first direction.

An apparatus is described. The apparatus includes a compute-in-memory (CIM) circuit for implementing a neural network disposed on a semiconductor chip. The CIM circuit includes a mathematical computation circuit coupled to a memory array. The memory array includes an embedded dynamic random access memory (eDRAM) memory array. Another apparatus is described. The apparatus includes a compute-in-memory (CIM) circuit for implementing a neural network disposed on a semiconductor chip. The CIM circuit includes a mathematical computation circuit coupled to a memory array. The mathematical computation circuit includes a switched capacitor circuit. The switched capacitor circuit includes a back-end-of-line (BEOL) capacitor coupled to a thin film transistor within the metal/dielectric layers of the semiconductor chip. Another apparatus is described. The apparatus includes a compute-in-memory (CIM) circuit for implementing a neural network disposed on a semiconductor chip. The CIM circuit includes a mathematical computation circuit coupled to a memory array. The mathematical computation circuit includes an accumulation circuit. The accumulation circuit includes a ferroelectric BEOL capacitor to store a value to be accumulated with other values stored by other ferroelectric BEOL capacitors.

An integrated circuit includes: a memory cell block including a plurality of bitcells; and an input and output (I/O) block including a plurality of gate-all-around (GAA) transistors connected to the bitcells, wherein the I/O block includes a plurality of active regions disposed separately from one another in a first direction, each of which extends in a second direction that is vertical to the first direction, and in which the GAA transistors are formed, a plurality of power rails disposed separately from one another in the first direction, and configured to provide power to the GAA transistors, and a plurality of signal lines disposed between the power rails, and configured to provide signals to the GAA transistors, a first number of bitcells among the bitcells are connected to the GAA transistors formed in a second number of active regions among the active regions, and the second number is twice the first number.

An integrated circuit includes: a cell array including a plurality of memory cells in a plurality of first columns and including a plurality of word line assist cells in at least one second column; a plurality of word lines respectively extending on a plurality of first rows of the cell array and connected to the plurality of memory cells and the plurality of word line assist cells; and a row driver configured to drive the plurality of word lines.

A semiconductor device in a multi-chip package (MCP) includes a controller, at least one non-volatile memory die including an array of non-volatile memory cells and connected to the controller through wire bonding, and at least one volatile memory die including an array of volatile memory cells and connected to the controller through wire bonding. The controller is configured to control operations of the at least one non-volatile memory die and the at least one volatile memory die.

An integrated circuit structure includes a Static Random Access Memory (SRAM) cell, which includes a read port and a write port. The write port includes a first pull-up Metal-Oxide Semiconductor (MOS) device and a second pull-up MOS device, and a first pull-down MOS device and a second pull-down MOS device forming cross-latched inverters with the first pull-up MOS device and the second pull-up MOS device. The integrated circuit structure further includes a first metal layer, with a bit-line, a CVdd line, and a first CVss line in the first metal layer, a second metal layer over the first metal layer, and a third metal layer over the second metal layer. A write word-line is in the second metal layer. A read word-line is in the third metal layer.