Abstract: A method for three-dimensional memory stacking may include providing a logic die including a circuit and a memory, providing a memory die including an additional memory having a same footprint as the circuit and memory in the logic die, and stacking the logic die and the memory die three-dimensionally with die-to-die data communication between the circuit and the additional memory by face-to-face hybrid bonds. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A method for three-dimensionally stacking systems on chip with face to face hybrid bonding may include providing a first die including a driver gate driving a first via ladder coupled to a first top metal layer. The method may additionally include providing a second die including a load gate coupled to a second via ladder coupled to a second top metal layer. The method may also include stacking the first die and the second die three-dimensionally using face-to-face hybrid bonds to couple the first top metal layer to the second top metal layer. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A method for non-uniform memory access on three-dimensionally-stacked hybrid memory may include providing a logic die including a circuit and a memory. The method may additionally include providing a plurality of memory dies including an additional memory. The method may also include stacking the logic die and the plurality of memory dies three-dimensionally using face-to-face hybrid bonds that provide non-uniform access to the additional memory by the circuit. Various other methods, systems, and computer-readable media are also disclosed.

Abstract: A device for reducing the effects of variation in inter-die communication in 3D-stacked systems may include a die-to-die interconnect that includes a first module configured to convert data from a first synchronous domain to a dual-rail quasi-delay-insensitive format and a second module configured to convert the data from the dual-rail quasi-delay-insensitive format to a second synchronous domain. Various other devices, systems, and methods of manufacture are also disclosed.

Abstract: An SRAM controller for performing sequential accesses using internal ports that operate concurrently on different rows. Each internal port includes a row address strobe (RAS) timer that generates clock signals controlling the timing of operations during a RAS phase in which word line decoding is performed once for a group of bit cells being accessed. The RAS phase can involve additional conditioning operations, such as precharging of local bits lines associated with the group of bit cells. The RAS phase is followed by an input/output (IO) phase in which individual bit cells are accessed in sequential address order using a column select signal generated by an IO timer. The RAS phase of a first internal port can be at least partially overlapped by the IO phase of a second internal port to hide the RAS latency of the first internal port. The IO timer can be shared among internal ports.

Abstract: Methods and corresponding systems and apparatuses for saving power through selectively disabling clock signals in a systolic array are described. In some embodiments, a clock gate controller is operable to output a gated clock signal from which local clock signals of processing elements in the systolic array are derived. The gated clock signal corresponds to a root clock signal that is distributed through a clock distribution network or clock tree. The clock gate controller is located along one branch of the clock distribution network. The branch can be associated with processing elements that form a column within the systolic array. Disabling the gated clock signal disables the local clock signals along the entire branch, preventing any components that are clocked by those local clock signals from consuming power. Additional clock gate controllers can similarly be provided for other branches, including a branch associated with another column.

Abstract: In some embodiments, a system comprises a static random access memory (SRAM) device and a controller. The SRAM device comprises a bit cell array comprising a plurality of bit cells arranged in a plurality of rows and a plurality of columns, each column operatively coupled to a pair of bit lines, wherein the plurality of columns is arranged as a plurality of column groups each comprising a plurality of local columns. The SRAM device further comprises a plurality of column decoders, each associated with a column group of the plurality of column groups. In some embodiments, the controller may be configured to read the local columns included in the column group by, for a given local column, sensing a voltage difference on a corresponding pair of bit lines, in a rearranged sequential order that is different from a physical sequential order of the plurality of local columns.

Abstract: An SRAM controller for performing sequential accesses using internal ports that operate concurrently on different rows. Each internal port includes a row address strobe (RAS) timer that generates clock signals controlling the timing of operations during a RAS phase in which word line decoding is performed once for a group of bit cells being accessed. The RAS phase can involve additional conditioning operations, such as precharging of local bits lines associated with the group of bit cells. The RAS phase is followed by an input/output (IO) phase in which individual bit cells are accessed in sequential address order using a column select signal generated by an IO timer. The RAS phase of a first internal port can be at least partially overlapped by the IO phase of a second internal port to hide the RAS latency of the first internal port. The IO timer can be shared among internal ports.

Abstract: In some embodiments, a system comprises a static random access memory (SRAM) device and a controller. The SRAM device comprises a bit cell array comprising a plurality of bit cells arranged in a plurality of rows and a plurality of columns, each column operatively coupled to a pair of bit lines, wherein the plurality of columns is arranged as a plurality of column groups each comprising a plurality of local columns. The SRAM device further comprises a plurality of column decoders, each associated with a column group of the plurality of column groups. In some embodiments, the controller may be configured to read the local columns included in the column group by, for a given local column, sensing a voltage difference on a corresponding pair of bit lines, in a rearranged sequential order that is different from a physical sequential order of the plurality of local columns.

Abstract: In some embodiments, an apparatus comprises: a static random access memory (SRAM) device. The SRAM device may have a bit cell array comprising a plurality of bit cells, the plurality of bit cells arranged in a plurality of rows and a plurality of columns, each column of the plurality of columns operatively coupled to a pair of bit lines. The apparatus may comprise a controller configured to: assert a word line associated with a row; perform a sequence of write operations while the word line remains asserted, each write operation corresponding to a bit cell associated with a different column of the plurality of columns and the row, wherein the word line has an elevated voltage relative to a non-elevated voltage during at least a portion of the sequence of write operations; and de-assert the word line after the sequence of write operations are performed.

Abstract: A neural network accelerator includes a first memory device, a controller connected to the first memory device through a high-bandwidth (e.g., three-dimensional) interconnect, a configurable processing element (PE) array connected to the first memory device through a first data bus and including a two-dimensional (2D) array of PEs, a local memory connected to the controller and connected, through a second data bus, to the configurable PE array. The controller is configured to, during execution of a neural network (NN), dynamically configure the neural network accelerator for executing each NN layer of a plurality of NN layers of the neural network by selecting either weights of a weight tensor or input data of an input tensor of a tensor operation of the NN layer to store into the local memory, and configuring input and output connections of PEs in the 2D array of PEs for performing the tensor operation.