INTELLIGENT RESOURCE AND THERMAL MANAGEMENT USING ARTIFICIAL INTELLIGENCE FOR PERFORMANCE OPTIMIZATION

Smart overview of the Invention

The patent application discusses systems and methods for optimizing resource and thermal management in computing systems using artificial intelligence (AI). It focuses on dynamically adjusting resource allocation and thermal conditions in systems with multiple processing units, such as CPUs and GPUs. Traditional static configurations often lead to inefficiencies and increased energy consumption, whereas the proposed approach leverages AI to predict operational states and adapt system parameters in real-time, enhancing performance and reducing energy use.

AI-Driven Predictions

The system utilizes AI and machine learning models, including variational autoencoders and neural networks, to predict future states of processing units based on real-time metrics. These predictions allow the system to dynamically update resource management, such as task scheduling and thermal control, to accommodate workload variations and maintain optimal performance. By forecasting metrics like workload spikes and thermal conditions, the system can preemptively adjust parameters such as clock speeds and cooling settings.

Real-Time Adaptation

The proposed system can adapt to changing workloads in real-time without relying on predefined thresholds. This adaptability is achieved through real-time monitoring and communication with kernel components, enabling synchronous updates to system parameters. The AI models process metrics related to performance, thermal conditions, and energy consumption to output predicted states, which guide the resource management adjustments.

Implementation Details

The system includes one or more processors that monitor processing units to obtain relevant metrics. Using AI models, the processors predict the state of these units and update static data structures to optimize resource management. This involves allocating tasks between integrated and discrete GPUs, performing thermal management, and executing power management tasks. The system also updates AI models based on performance feedback to refine predictions.

System Components

The system's components include circuits that monitor processing units and communicate with kernel components through real-time channels. These circuits use communication protocols to transmit parameters between user and kernel spaces, ensuring timely updates before thresholds are reached. By performing tasks such as thermal and power management based on predicted states, the system enhances efficiency and performance across various operational conditions.