A memory system having memory components, a remote direct memory access (RDMA) network interface card (RNIC), and a host system, and configured to: allocate a page of virtual memory for an application; map the page of virtual memory to a page of physical memory in the memory components; instruct the RNIC to perform an RDMA operation; perform, during the RDMA operation, a data transfer between the page of physical memory in the plurality of memory components and a remote device that is connected via a computer network to the remote direct memory access network interface card; and at least for a duration of the data transfer, lock a mapping between the page of virtual memory and the page of physical memory in the memory components.

Disclosed herein are techniques for implementing high-throughput low-latency hybrid memory modules with improved data backup and restore throughput, enhanced non-volatile memory controller (NVC) resource access, and enhanced mode register setting programmability. Embodiments comprise a command replicator to generate sequences of one or more DRAM read and/or write and/or other commands to be executed in response to certain local commands from a non-volatile memory controller (NVC) during data backup and data restore operations. Other embodiments comprise an access engine to enable an NVC in a host control mode to trigger entry into a special mode and issue commands to access a protected register space. Some embodiments comprise a mode register controller to capture and store the data comprising mode register setting commands issued during a host control mode, such that an NVC can program the DRAM mode registers in an NVC control mode.

A dynamic random access memory (DRAM) component (e.g., module or integrated circuit) can be configured to have multiple rows in the same bank open concurrently. The controller of the component divides the address space of the banks into segments based on row address ranges. These row address ranges do not necessarily correspond to row address ranges of the bank's subarrays (a.k.a. memory array tiles—MATs). When a command is sent to open a row, the controller marks a plurality of the segments as blocked. The controller thereby tracks address ranges in a bank where it will not open a second row unless and until the first row is closed. The memory component may store information about which, and how many, segments should be blocked in response to opening a row. This information may be read by the controller during initialization.

A DRAM-based storage device includes a DRAM and a control circuit. The DRAM includes a buffering area and a host accessing area. A data is stored in the host accessing area. The control circuit is electrically connected with the DRAM. The control circuit copies a portion of the data from the host accessing area to the buffering area at a predetermined time interval counted by the control circuit. Before the portion of the data is written to the buffering area, a first ECC decoding operation is performed on the portion of the data to correct error bits contained therein. If the portion of the data is corrected, the control circuit rewrites the corrected portion of the data into the host accessing area.

Technologies for allocating resources of managed nodes to workloads to balance multiple resource allocation objectives include an orchestrator server to receive resource allocation objective data indicative of multiple resource allocation objectives to be satisfied. The orchestrator server is additionally to determine an initial assignment of a set of workloads among the managed nodes and receive telemetry data from the managed nodes. The orchestrator server is further to determine, as a function of the telemetry data and the resource allocation objective data, an adjustment to the assignment of the workloads to increase an achievement of at least one of the resource allocation objectives without decreasing an achievement of another of the resource allocation objectives, and apply the adjustments to the assignments of the workloads among the managed nodes as the workloads are performed. Other embodiments are also described and claimed.

Embodiments may be generally direct to apparatuses, systems, method, and techniques to determine a configuration for a plurality of connectors, the configuration to associate a first interconnect protocol with a first subset of the plurality of connectors and a second interconnect protocol with a second subset of the plurality of connectors, the first interconnect protocol and the second interconnect protocol are different interconnect protocols and each comprising one of a serial link protocol, a coherent link protocol, and an accelerator link protocol, cause processing of data for communication via the first subset of the plurality of connectors in accordance with the first interconnect protocol, and cause processing of data for communication via the second subset of the plurality of connector in accordance with the second interconnect protocol.

Technologies for dynamically managing resources in disaggregated accelerators include an accelerator. The accelerator includes acceleration circuitry with multiple logic portions, each capable of executing a different workload. Additionally, the accelerator includes communication circuitry to receive a workload to be executed by a logic portion of the accelerator and a dynamic resource allocation logic unit to identify a resource utilization threshold associated with one or more shared resources of the accelerator to be used by a logic portion in the execution of the workload, limit, as a function of the resource utilization threshold, the utilization of the one or more shared resources by the logic portion as the logic portion executes the workload, and subsequently adjust the resource utilization threshold as the workload is executed. Other embodiments are also described and claimed.

Embodiments of an improved memory architecture for processing data inside of a device are described. In some embodiments, the device can store neural network layers, such as a systolic flow engine, in non-volatile memory and/or a separate first memory. A processor of a host system can delegate the execution of a neural network to the device. Advantageously, neural network processing in the device can be scalable, with the ability to process large amounts of data.

To provide a novel semiconductor device. The semiconductor device includes cell arrays and peripheral circuits; the cell arrays include memory cells; the peripheral circuits includes a first driver circuit, a second driver circuit, a first amplifier circuit, a second amplifier circuit, a third amplifier circuit, and a fourth amplifier circuit; the first driver circuit and the second driver circuit have a function of supplying a selection signal to the cell array; the first amplifier circuit and the second amplifier circuit have a function of amplifying a potential input from the cell array; the third amplifier circuit and the fourth amplifier circuit have a function of amplifying a potential input from the first amplifier circuit or the second amplifier circuit; the first driver circuit, the second driver circuit, the first amplifier circuit, the second amplifier circuit, the third amplifier circuit, and the fourth amplifier circuit include a region overlapping with the cell array; and the memory cells include a metal oxide in a channel formation region.

Disclosed herein are techniques for implementing high-throughput low-latency hybrid memory modules with improved data backup and restore throughput, enhanced non-volatile memory controller (NVC) resource access, and enhanced mode register setting programmability. Embodiments comprise a command replicator to generate sequences of one or more DRAM read and/or write and/or other commands to be executed in response to certain local commands from a non-volatile memory controller (NVC) during data backup and data restore operations. Other embodiments comprise an access engine to enable an NVC in a host control mode to trigger entry into a special mode and issue commands to access a protected register space. Some embodiments comprise a mode register controller to capture and store the data comprising mode register setting commands issued during a host control mode, such that an NVC can program the DRAM mode registers in an NVC control mode.