AMD has submitted a new Linux kernel patch series designed to improve how the operating system distributes workloads across processors containing different types of CPU cores. The proposed changes introduce Capacity Aware Scheduling (CAS) support through AMD’s amd-pstate driver, allowing Linux to distinguish between higher-performance and lower-power cores when assigning tasks.
AMD engineer Vishal Badole submitted the two-patch series to the Linux kernel mailing list on November 14, 2026. The work targets heterogeneous processors that combine performance and efficiency cores, addressing a limitation in how Linux currently handles differences in their processing capabilities.
According to AMD’s technical explanation, the existing amd-pstate driver does not provide the Linux scheduler with individual CPU capacity information. Consequently, the scheduler can treat cores with different performance characteristics as equally capable, potentially assigning demanding tasks to slower cores while more powerful cores remain available.

The proposed implementation addresses this problem by publishing asymmetric per-CPU capacity values through the kernel’s hybrid capacity scaling interface. This would give Linux additional information when deciding where workloads should run, potentially improving task distribution, processing throughput and energy efficiency on supported hardware.
The first patch modifies amd-pstate to calculate each CPU’s relative capacity using information about its maximum operating frequency or effective highest-performance capabilities. It uses CPU 0 as a reference, assuming that it represents a high-performance core on supported systems.
When suitable maximum-frequency information is available, the driver uses those values to determine relative CPU capacity. Otherwise, it falls back to performance information obtained through AMD’s Collaborative Processor Performance Control (CPPC) interface.
These capacity values are then supplied to the kernel’s scheduling infrastructure, enabling it to account for differences between individual cores rather than assuming they all offer identical processing capabilities.
AMD intends to enable Capacity Aware Scheduling by default on compatible hardware. Users would retain the option to disable the feature through the amd_cas=disable kernel boot parameter. The driver would also expose the feature’s effective state through a sysfs attribute named cas.
The second patch, authored by AMD engineer Krishnamoorthi M, addresses potential conflicts with AMD’s Hardware Feedback Interface (amd-hfi) driver.
Both mechanisms can influence scheduler information used to distinguish CPU capabilities. Allowing them to manage the same scheduling state simultaneously could produce conflicting settings and unpredictable task placement.
To prevent this, the proposed changes stop amd-hfi from loading when amd-pstate Capacity Aware Scheduling is active, ensuring that only one mechanism manages the relevant capacity information.
The development is particularly interesting in the context of AMD’s upcoming processor architectures. AMD’s Zen 6 architecture introduces another generation of CPU improvements, while reports have suggested that future Zen 6 products could incorporate lower-power core variants.
However, the submitted patches describe support for heterogeneous AMD processors generally. They do not explicitly confirm a particular Zen 6 consumer processor configuration, establish which Ryzen models will support the feature, or provide a release schedule.
For Linux PC users, more accurate scheduling could eventually help CPU-intensive applications make better use of hybrid processors. Games that depend heavily on CPU performance might also benefit from improved workload placement, although AMD has not supplied gaming benchmarks or demonstrated specific frame-rate improvements.
The changes remain proposed Linux kernel patches rather than an established feature in a released kernel. Their final implementation, compatibility and availability will depend on the kernel review and integration process.
If accepted, the patches would give Linux a more detailed understanding of the performance differences between AMD CPU cores, helping prepare the operating system for heterogeneous processor designs without requiring the scheduler to treat every core as equally capable.
