In one embodiment, a method for increasing power efficiency for an information handling system includes: monitoring, by a host service, one or more performance metrics associated with a memory device of the information handling system, the memory device including a power controller communicably coupled to a management device via a side-band bus; predicting, by the host service, an energy requirement for the memory device based on the one or more performance metrics; generating, by the host service, a power configuration profile based on the energy requirement, the power configuration profile indicating one or more power controller parameters associated with the power controller; sending, by the host service, the power configuration profile to the management device; receiving, by the management device, the power configuration profile; and modifying, by the management device and via the side-band bus, the one or more power controller parameters based on the power configuration profile.

A system may include a chassis comprising a plurality of bays, each of the plurality of bays configured to receive a respective modular information handling system, a shared infrastructure comprising a plurality of components which are shared by modular information handling systems received in the plurality of bays, and a chassis management controller communicatively coupled to the plurality of bays and configured to determine a power consumption for each respective modular information handling system and if the power consumption for a respective modular information handling system exceeds a maximum sled connector current limit for the respective modular information handling system, communicate a message to the respective modular information handling system to power down the respective modular information handling system.

Software-defined fail-safe power draw control is provided for rack power distribution units (PDUs). A PDU power manager can be used to map devices to a PDU and its sockets. The PDU power manager can also identify the power rating of the PDU and its sockets. With this information, the PDU power manager can generate and apply priority-based and dynamic power caps to the devices connected to the PDU. In this way, the PDU power manager can ensure that the power ratings of the PDU and of each of its sockets are not exceeded while ensuring that power delivery is prioritized based on the workloads that the devices may experience.

Systems and methods are provided for designing approximate, low-power deep learning accelerator chips that have little to no accuracy loss when executing a deep learning model. A set of approximate systolic arrays may be generated. The performance of each approximate systolic array in the set of approximate systolic arrays processing a deep neural network (DNN) may be determined. Each layer in the DNN may be mapped to an approximate systolic array in the set of approximate systolic arrays. A subset of the set of approximate systolic arrays may be selected for inclusion in the inference chip design based on the mapping and the performance of each approximate systolic array in the set of approximate systolic arrays.

A data storage device includes a non-volatile memory device including a memory block having a number of arrays, a number of power regulators, and a controller coupled to the non-volatile memory device. The arrays include a number of memory devices. A first array is determined to be in a non-responsive condition and a power regulator supplying power to the first array is instructed to cycle power to the first array. After the power to the first array has been cycled, a determination is made as to whether the first array is in a responsive condition.

The disclosure provides a method for processing data, and an electronic device. The method includes: obtaining first attribute information of input data and second attribute information of a computing device corresponding to the input data; selecting a target operator implementation mode from a plurality of candidate operator implementation modes based on the first attribute information and the second attribute information; determining a plurality of sub-operators included in an operator required for the input data from an operator library based on the target operator implementation mode, to generate the operator; and obtaining an operation result by performing an operation on the input data by the computing device based on the operator.

A data center infrastructure monitoring system and method that includes a plurality of geographically distributed data centers, each data center comprising a plurality of mechanical assets. A data center gateway system communicatively coupled to one or more data centers of the plurality of data centers is configured to determine a subset of the plurality of mechanical assets of the data center that provide cooling to equipment of a customer of the data center, and output, for display, a graphical depiction of the subset of the plurality of mechanical assets of the data center that provide cooling to equipment of the customer in the data center.

A storage system comprises one or more storage devices, power supplies supplying power to the storage device, a processor that performs in response to determining that the total power consumption of the one or more storage devices is less than a first percentage threshold of a load of the active power supplies, deactivating one or more of the active power supplies until the total power consumption is equal to or greater than the first percentage threshold of a load of each of the active power supplies, and in response to determining that the total power consumption is equal to or greater than a second percentage threshold of a load of each of the active power supplies, activating one or more of the deactivated ones of the power supplies until the total power consumption is less than the second percentage threshold of the load of each of the active power supplies.

Deploying energy-rated deep neural network models on energy-scored edge devices is provided. An overall energy efficiency rating is assigned to a deep neural network model based on utilizing software optimization and hardware accelerators during training of the deep neural network model. Energy scores are assigned to respective edge devices in an edge inference computing environment based on properties of each respective edge device. Particular edge devices are selected that have a corresponding energy score within a defined edge device energy score range for the overall energy efficiency rating that corresponds to the deep neural network model. The deep neural network model is deployed to the particular edge devices that have a corresponding energy score within the defined edge device energy score range for the overall energy efficiency rating that corresponds to the deep neural network model.

Described herein is an integrated data center that provides for efficient cooling, as well as efficient wire routing.