Heterogeneous Power Plan settings can be used to provide guidance to the Windows scheduler and core parking subsystems regarding thread scheduling and core parking preferences in systems with Hybrid Core Architecture, featuring at least to different types of cores Performant (P) and Efficiency (E) cores. Quick CPU offers quick access to these settings from the main application form for 12th+ generation Intel CPUs. These settings are also accessible at any time via Quick CPU's dedicated Power Plan management interface for all CPU platforms.

There are three diffrent sets of power settings that will be covered here.

• Heterogeneous Policy: This setting determines the policy used on systems with hybrid core architecture. It is related to the core parking engine, which makes scalability decisions about the workload and identifies the optimal set of cores for the workload. In platforms with hybrid core architecture, the core parking engine also decides the workload execution path, considering Performant (P) or Efficiency (E) cores. The Heterogeneous policy has several different setting values that determine the unparking strategy for the optimal set of compute cores.
  For our testing below, we will use the value "0," known as "Standard Core Parking," which prioritizes the most performant cores first for unparking. Note that our core parking index is set to 100%, effectively disabling core parking, so our system will not be affected by this policy setting.
• Heterogeneous Thread Scheduling Policy (Normal or Long Threads): This setting defines the thread scheduler policy for normal or long running threads in relation to Performance (P) or Efficiency (E) cores.

  The options include:

  1. All Processors (0): Utilize all available processor cores.
  2. Performant Processors (1): Schedule the thread using only P-Cores.
  3. Prefer Performant Processors (2): Prefer P-Cores for most tasks, but not exclusively.
  4. Efficient Processors (3): Schedule the thread using only E-Cores.
  5. Prefer Efficient Processors (4): Prefer E-Cores for most tasks, but not exclusively.
  6. Automatic (5): Automatically determine the optimal scheduling preference between P-Cores and E-Cores.
• Heterogeneous Short Running Thread Scheduling Policy: This policy has the same meanings and values as the policy for long threads but applies to short running threads.

The primary difference between long and short running threads is determined by the duration of the workload that needs to be scheduled.

Short Running Threads: These threads are designed to last for a very short period, handling miscellaneous tasks such as small asynchronous procedure calls, minor data calculations, and background activities. The duration of these tasks can typically be measured in processor time quantum units or milliseconds.

Long Running Threads: These tasks require more time to execute and are associated with activities that demand continuous and sustained computational resources. Examples include data processing, video editing, or complex calculations performed by an application. The duration of these tasks varies depending on the workload and can extend for as long as the lifetime of the application process that originated the thread.

Determining the Nature of Threads (Long vs. Short)

The operating system can determine the nature of a thread based on:

• Thread past behavior.
• Thread analysis, priority, type of task etc.
• Explicit hints from the application code.

Hybrid Core Assignment

Typically, Efficiency cores (E-Cores) are well-suited to execute short running threads, contributing to energy efficiency and reduced heat generation. In oposite, Performant cores (P-Cores) can leverage their higher performance and computational capacity to execute long running threads. It's important to note that there are no strict boundaries in assigning workloads to P or E cores. The actual scheduling decision can depend on the current system state and a variety of other factors.