Abstract: Logic (apparatus and/or software) is provided that separates read and restore operations. When a read is completed, the read data is stored in a restore buffer allowing other latency critical operations such as reads to be serviced before the restore. Deferring restore operations minimizes latency and burst bandwidth for reads and minimizes the performance impact of the non-critical restore operations.

Abstract: An adder with first and second majority gates. For a 1-bit adder, output from a 3-input majority gate is inverted and input two times to a 5-input majority gate. Other inputs to the 5-input majority gate are same as those of the 3-input majority gate. The output of the 5-input majority gate is a sum while the output of the 3-input majority gate is the carry. Multiple 1-bit adders are concatenated to form an N-bit adder. The input signals are driven to first terminals of non-ferroelectric capacitors while the second terminals are coupled to form a majority node. Majority function of the input signals occurs on this node. The majority node is then coupled to a first terminal of a non-linear polar capacitor. The second terminal of the capacitor provides the output of the logic gate. A reset mechanism initializes the non-linear polar capacitor before addition function is performed.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor which in turn comprises a polar layer comprising a base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen. The dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV. The capacitor stack additionally comprises first and second crystalline conductive oxide electrodes on opposing sides of the polar layer.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor which in turn comprises a polar layer comprising a base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen. The dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV. The capacitor stack additionally comprises first and second crystalline conductive oxide electrodes on opposing sides of the polar layer.

Abstract: A low power sequential circuit (e.g., latch) uses a non-linear polar capacitor to retain charge with fewer transistors than traditional CMOS sequential circuits. In one example, a sequential circuit includes pass-gates and inverters, but without a feedback mechanism or memory element. In another example, a sequential uses load capacitors (e.g., capacitors coupled to a storage node and a reference supply). The load capacitors are implemented using ferroelectric material, paraelectric material, or linear dielectric. In one example, a sequential uses minority, majority, or threshold gates with ferroelectric or paraelectric capacitors. In one example, a sequential circuit uses minority, majority, or threshold gates configured as NAND gates.

Abstract: Endurance mechanisms are introduced for memories such as non-volatile memories for broad usage including caches, last-level cache(s), embedded memory, embedded cache, scratchpads, main memory, and storage devices. Here, non-volatile memories (NVMs) include magnetic random-access memory (MRAM), resistive RAM (ReRAM), ferroelectric RAM (FeRAM), phase-change memory (PCM), etc. In some cases, features of endurance mechanisms (e.g., randomizing mechanisms) are applicable to volatile memories such as static random-access memory (SRAM), and dynamic random-access memory (DRAM). The endurance mechanisms include a wear leveling scheme that uses index rotation, outlier compensation to handle weak bits, and random swap injection to mitigate wear out attacks.

Abstract: An adder with first and second majority gates. For a 1-bit adder, output from a 3-input majority gate is inverted and input two times to a 5-input majority gate. Other inputs to the 5-input majority gate are same as those of the 3-input majority gate. The output of the 5-input majority gate is a sum while the output of the 3-input majority gate is the carry. Multiple 1-bit adders are concatenated to form an N-bit adder. The input signals are driven to first terminals of non-ferroelectric capacitors while the second terminals are coupled to form a majority node. Majority function of the input signals occurs on this node. The majority node is then coupled to a first terminal of a non-linear polar capacitor. The second terminal of the capacitor provides the output of the logic gate. A reset mechanism initializes the non-linear polar capacitor before addition function is performed.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor which in turn comprises a polar layer comprising a base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen. The dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV. The capacitor stack additionally comprises first and second crystalline conductive oxide electrodes on opposing sides of the polar layer.

Abstract: A new class of logic gates are presented that use non-linear polar material. The logic gates include multi-input majority gates. Input signals in the form of digital signals are driven to non-linear input capacitors on their respective first terminals. The second terminals of the non-linear input capacitors are coupled a summing node which provides a majority function of the inputs. The majority node is then coupled driver circuitry which can be any suitable logic gate such as a buffer, inverter, NAND gate, NOR gate, etc. In the multi-input majority or minority gates, the non-linear charge response from the non-linear input capacitors results in output voltages close to or at rail-to-rail voltage levels. Bringing the majority output close to rail-to-rail voltage eliminates the high leakage problem faced from majority gates formed using linear input capacitors.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor which in turn comprises a polar layer comprising a base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen. The dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV. The capacitor stack additionally comprises first and second crystalline conductive oxide electrodes on opposing sides of the polar layer.

Abstract: A new class of logic gates are presented that use non-linear polar material. The logic gates include multi-input majority gates. Input signals in the form of digital signals are driven to non-linear input capacitors on their respective first terminals. The second terminals of the non-linear input capacitors are coupled a summing node which provides a majority function of the inputs. The majority node is then coupled driver circuitry which can be any suitable logic gate such as a buffer, inverter, NAND gate, NOR gate, etc. In the multi-input majority or minority gates, the non-linear charge response from the non-linear input capacitors results in output voltages close to or at rail-to-rail voltage levels. Bringing the majority output close to rail-to-rail voltage eliminates the high leakage problem faced from majority gates formed using linear input capacitors.

Abstract: Approaches for integrating FE memory arrays into a processor, and the resulting structures are described. Simultaneous integrations of regions with ferroelectric (FE) cells and regions with standard interconnects are also described. FE cells include FE capacitors that include a FE stack of layers, which is encapsulated with a protection material. The protection material protects the FE stack of layers as structures for regular logic are fabricated in the same die.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor, which in turn comprises a polar layer comprising a crystalline base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen, wherein the dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV.

Abstract: An adder with first and second majority gates is provided. For a 1-bit adder, output from a 3-input majority gate is inverted and input two times to a 5-input majority gate. Other inputs to the 5-input majority gate are same as those of the 3-input majority gate. The output of the 5-input majority gate is a sum while the output of the 3-input majority gate is the carry. Multiple 1-bit adders are concatenated to form an N-bit adder. The input signals are driven to first terminals of non-ferroelectric capacitors while the second terminals are coupled to form a majority node. Majority function of the input signals occurs on this node. The majority node is then coupled to a first terminal of a non-linear polar capacitor. The second terminal of the capacitor provides the output of the logic gate. A reset mechanism initializes the non-linear polar capacitor before addition function is performed.

Abstract: A new class of logic gates are presented that use non-linear polar material. The logic gates include multi-input majority gates. Input signals in the form of digital signals are driven to non-linear input capacitors on their respective first terminals. The second terminals of the non-linear input capacitors are coupled a summing node which provides a majority function of the inputs. The majority node is then coupled driver circuitry which can be any suitable logic gate such as a buffer, inverter, NAND gate, NOR gate, etc. In the multi-input majority or minority gates, the non-linear charge response from the non-linear input capacitors results in output voltages close to or at rail-to-rail voltage levels. Bringing the majority output close to rail-to-rail voltage eliminates the high leakage problem faced from majority gates formed using linear input capacitors.

Abstract: Described is a packaging technology to improve performance of an AI processing system. An IC package is provided which comprises: a substrate; a first die on the substrate, and a second die stacked over the first die. The first die includes memory and the second die includes computational logic. The first die comprises a ferroelectric RAM (FeRAM) having bit-cells. Each bit-cell comprises an access transistor and a capacitor including ferroelectric material. The access transistor is coupled to the ferroelectric material. The FeRAM can be FeDRAM or FeSRAM. The memory of the first die may store input data and weight factors. The computational logic of the second die is coupled to the memory of the first die. The second die is an inference die that applies fixed weights for a trained model to an input data to generate an output. In one example, the second die is a training die that enables learning of the weights.

Abstract: A low power sequential circuit (e.g., latch) uses a non-linear polar capacitor to retain charge with fewer transistors than traditional CMOS sequential circuits. The sequential circuit includes a 3-input majority gate having first, second, and third inputs, and a first output. The sequential circuit includes a driver coupled to the first output, wherein the driver is to generate a second output. The sequential circuit further includes an exclusive-OR (XOR) gate to receive a clock and the second output, wherein the XOR gate is to generate a third output which couples to the second input, where the first input is to receive a data, and wherein the third input is to receive the second output.

Abstract: The disclosed technology generally relates to ferroelectric materials and semiconductor devices, and more particularly to semiconductor memory devices incorporating doped polar materials. In one aspect, a semiconductor device comprises a capacitor which in turn comprises a polar layer comprising a base polar material doped with a dopant. The base polar material includes one or more metal elements and one or both of oxygen or nitrogen. The dopant comprises a metal element that is different from the one or more metal elements and is present at a concentration such that a ferroelectric switching voltage of the capacitor is different from that of the capacitor having the base polar material without being doped with the dopant by more than about 100 mV. The capacitor stack additionally comprises first and second crystalline conductive oxide electrodes on opposing sides of the polar layer.

Abstract: Described is a packaging technology to improve performance of an AI processing system. An IC package is provided which comprises: a substrate; a first die on the substrate, and a second die stacked over the first die. The first die includes memory and the second die includes computational logic. The first die comprises DRAM having bit-cells. The memory of the first die may store input data and weight factors. The computational logic of the second die is coupled to the memory of the first die. In one example, the second die is an inference die that applies fixed weights for a trained model to an input data to generate an output. In one example, the second die is a training die that enables learning of the weights. Ultra high-bandwidth is changed by placing the first die below the second die. The two dies are wafer-to-wafer bonded or coupled via micro-bumps.

Abstract: Described is a packaging technology to improve performance of an AI processing system. An IC package is provided which comprises: a substrate; a first die on the substrate, and a second die stacked over the first die. The first die includes memory and the second die includes computational logic. The first die comprises a ferroelectric RAM (FeRAM) having bit-cells. Each bit-cell comprises an access transistor and a capacitor including ferroelectric material. The access transistor is coupled to the ferroelectric material. The FeRAM can be FeDRAM or FeSRAM. The memory of the first die may store input data and weight factors. The computational logic of the second die is coupled to the memory of the first die. The second die is an inference die that applies fixed weights for a trained model to an input data to generate an output. In one example, the second die is a training die that enables learning of the weights.