CPU core selection

With the ARM Snapdragon platform being commonly used for Surface devices and also several Intel Chipsets supporting e-cores, it has now become commonplace to have heterogeneous systems with efficiency cores and performance cores in use. Heterogeneous systems support a combination of e-cores and p-cores. p-cores are high power “performance” cores and e-cores are “efficiency” cores. p-cores generally run faster than e-cores at the expense of consuming more CPU power.

While efficiency cores are cost effective and can improve general responsiveness of applications while consuming less power, at low latency this can create performance problems due to load imbalances when mixing and matching e-cores with p-cores. As such, it's better to deprioritize use of e-cores for low latency applications.

There are two Aud.ini configuration variables named CoreSelectionMethod and ExcludeEfficiencyCores that allow Sonar more fine-grained control over these CPU resources. These variables can be specified in Edit > Preferences > Audio - Configuration File. Sonar will now favor p-cores over e-cores and first use p-cores when available.

CoreSelectionMethod

This Aud.ini variable specifies how Sonar chooses preferred cores:

0 = No preference

When this is selected, Sonar makes no attempt to specify which cores should be used for audio engine threads.

1 = Requests preferred core

Specifies preferred per thread cores. This is the default option and is similar to the now deprecated UseIdealProcessor variable. Note that this request is only a hint and as a result the core assignment may be overridden by the operating system.

2 = Restrict to preferred core

This is similar to 1, but it actually restricts operation of the engine threads to only run on the specified cores. In this case the OS will not change core assignment. Depending on your CPU this option may allow the engine to run more efficiently without the overhead of its threads being switched to other cores by Windows. This can lead to lower CPU use in general.

ExcludeEfficiencyCores

Setting this variable to True, instructs Sonar to aggressively exclude efficiency cores (e-cores) from audio processing and to only use p-cores. This can be beneficial since all cores will now have equal performance and as a result you may see better load balancing or better low latency performance. Note that if you have insufficient p-cores you may get worse performance by excluding e-cores completely.

 

See:

Audio - Configuration File (Advanced)

AUD.INI