Abstract: Methods and apparatus for employing selective compression for addressing congestion control for Artificial Intelligence (AI) workloads. Multiple interconnected compute nodes are used for performing an AI workload in a distributed environment, such as training an AI model. Periodically, such as following an epoch for processing batches of training data in parallel, the compute nodes exchange Tensor data (e.g., local model gradients) with one another, which may lead to network/fabric congestion. Compute nodes and/or switches in the distributed environment are configured to detect current or projected network/fabric congestion and to selectively apply variable rate compression to packets containing the Tensor data to alleviate/avoid the congestion. Tensor data may be selectively applied at source compute nodes by computing a network pause time and comparing that time to a compression compute time.

Abstract: A computing device, system and method. The computing device includes a memory storing instructions, and a processing circuitry coupled to the memory. The processing circuitry is configured to execute the instructions to process a first control signal and a second control signal from respective first and second control pins of a computing platform. The processing circuitry is further to transition the computing platform, based on a combination of the first control signal and the second control signal and using at least one voltage pin on the platform, between a low power state and a retention power state without transitioning to an operational power state in between.

Abstract: Technologies of managing power during an activation cycle of a processor core or other compute domain include determining new operation limits for active processor cores or other compute domains during an activation cycle of a hibernating processor core or other hibernating compute domain to reduce the likelihood of a power surge during the activation of the hibernating processor core or other compute domain. The active processor cores or other compute domain are monitored until their operating points are at or below the new operating limits. Thereafter, the hibernating processor core or other hibernating compute domain is activated.

Abstract: A computing device, system and method. The computing device includes a memory storing instructions, and a processing circuitry coupled to the memory. The processing circuitry is configured to execute the instructions to process a first control signal and a second control signal from respective first and second control pins of a computing platform. The processing circuitry is further to transition the computing platform, based on a combination of the first control signal and the second control signal and using at least one voltage pin on the platform, between a low power state and a retention power state without transitioning to an operational power state in between.

Abstract: Technologies of managing power during an activation cycle of a processor core or other compute domain include determining new operation limits for active processor cores or other compute domains during an activation cycle of a hibernating processor core or other hibernating compute domain to reduce the likelihood of a power surge during the activation of the hibernating processor core or other compute domain. The active processor cores or other compute domain are monitored until their operating points are at or below the new operating limits. Thereafter, the hibernating processor core or other hibernating compute domain is activated.

Abstract: Technologies of managing power during an activation cycle of a processor core or other compute domain include determining new operation limits for active processor cores or other compute domains during an activation cycle of a hibernating processor core or other hibernating compute domain to reduce the likelihood of a power surge during the activation of the hibernating processor core or other compute domain. The active processor cores or other compute domain are monitored until their operating points are at or below the new operating limits. Thereafter, the hibernating processor core or other hibernating compute domain is activated.

Abstract: Technologies of managing power during an activation cycle of a processor core or other compute domain include determining new operation limits for active processor cores or other compute domains during an activation cycle of a hibernating processor core or other hibernating compute domain to reduce the likelihood of a power surge during the activation of the hibernating processor core or other compute domain. The active processor cores or other compute domain are monitored until their operating points are at or below the new operating limits. Thereafter, the hibernating processor core or other hibernating compute domain is activated.