"cpupower frequency-set" provides a simplified mechanism to adjust processor frequencies when cpu frequency scaling is enabled in the OS. See the cpupower-frequency-set man page for details.Here is a brief description of options used in the config file. By default, settings are applied to all logical cpus in the system.Frequencies can be passed in Hz, kHz (default), MHz, GHz, or THz by postfixing the value with the desired unit name, without any space. Available frequencies and governors can be determined with "cpupower frequency-info".
The 'tuned' provides a number of predefined profiles for typical use cases. The 'tuned-adm' command is used to change settings of the tuned daemon. The tuned-adm command can query current settings, list available profiles, recommend a tuning profile for the system, change profiles directly, or turn off tuning.
throughput-performance: For typical throughput performance tuning. Disables power saving mechanisms and enables sysctl settings that improve the throughput performance of disk and network I/O. CPU governor is set to performance and CPU energy performance bias is set to performance. Disk readahead values are increased.
latency-performance: For low latency performance tuning. Disables power saving mechanisms. CPU governor is set to performance and locked to the low C states. CPU energy performance bias to performance.
balanced: Default profile provides balanced power saving and performance. It enables CPU and disk plugins of tuned and makes the conservative governor is active and also sets the CPU energy performance bias to normal. It also enables power saving on audio and graphics card.
powersave: Maximal power saving for whole system. It sets the CPU governor to ondemand governor and energy performance bias to powersave. It also enable power saving on USB, SATA, audio and graphics card.
accelerator-performance: Throughput performance based tuning with disabled higher latency STOP states.
desktop: Optimize for the desktop use-case.
hpc-compute: Optimize for HPC compute workloads.
intel-sst: Configure for Intel Speed Select Base Frequency.
network-latency: Optimize for deterministic performance at the cost of increased power consumption, focused on low latency network performance.
network-throughput: Optimize for streaming network throughput, generally only necessary on older CPUs or 40G+ networks.
optimize-serial-console: Optimize for serial console use.
virtual-guest: Optimize for running inside a virtual guest.
virtual-host: Optimize for running KVM guests.
L2 RFO (Read For Ownership) Prefetch is a prefetch mechanism that anticipates cache line reads for ownership (i.e., data that will be modified). When enabled, the processor intelligently prefetches adjacent data during read operations, improving cache hit rates and overall system performance for memory-intensive workloads.
Values for this BIOS option can be:
Enabled: Enable the L2 RFO prefetch function. The processor will speculatively prefetch adjacent cache lines when a read-for-ownership operation occurs, reducing latency for subsequent writes.
Disabled(Default): Disable the L2 RFO prefetch function. The processor does not perform speculative prefetching for RFO operations.
Intel Virtualization Technology for Directed I/O (VT-d) is a hardware-assisted virtualization feature that provides I/O device assignment to virtual machines via DMA remapping and interrupt remapping. It enhances performance, security, and isolation for virtualized environments by allowing direct device pass-through.
Values for this BIOS option can be:
Enabled(Default): Enable Intel VT-d. The processor supports I/O virtualization, enabling VMMs to assign physical devices to guest operating systems and perform address translation for DMA.
Disabled: Disable Intel VT-d. The processor does not provide I/O virtualization support.
KTI (Keizer Technology Interconnect) Prefetch, also known as Intel UPI (Ultra Path Interconnect) prefetch, is a mechanism that prefetches data deemed relevant into the L1 cache. This allows memory read requests to be issued earlier on the DDR bus, effectively reducing memory access latency.
Values for this BIOS option can be:
Enabled: Force-ble the KTI prefetcher. The processor will actively prefetch correlated data into the L1 cache to minimise read latency.
Auto(Default): The KTI prefetcher operates automatically, enabling prefetching when the hardware determines it is beneficial.
Disabled: Disable the KTI prefetcher. No speculative prefetching is performed over the KTI/UPI links.
Sub NUMA Clustering (SNC) divides the last level cache (LLC) into multiple disjoint clusters based on address ranges, each tied to a subset of the system's memory controllers. This reduces average LLC access latency and alleviates memory bandwidth contention. SNC2 enables two clusters with 1-way IMC interleaving.
Values for this BIOS option can be:
Enable SNC2 (2-clusters): Enable Sub NUMA Clustering with two LLC clusters and 1-way interleave across memory controllers. The LLC is split into two independent portions, improving locality and reducing cross-cluster accesses.
Disabled(Default): Disable Sub NUMA Clustering. The LLC is treated as a single unified cluster with default interleaving.
This setting allows the system administrator to select a predefined power and performance policy that tunes the processor, memory, and platform behaviour to best match the workload characteristics. Each profile adjusts parameters such as CPU frequency scaling, C-states, turbo boost, memory latency, and I/O responsiveness to achieve the desired balance between performance and energy efficiency.
Values for this BIOS option can be:
High Performance Compute: Maximises compute throughput for CPU-intensive, parallel workloads. Disables most power-saving features, enables maximum turbo frequencies, and favours performance over energy efficiency.
Balance/Energy Efficiency: Provides a balanced trade-off between performance and power consumption. Suitable for general-purpose server workloads where both metrics are important.
Power Saving: Aggressively reduces power consumption by limiting CPU frequency, enabling deeper C-states, and throttling performance where possible. Ideal for low-utilisation or energy-constrained environments.
Low Latency: Optimises for minimal response time, reducing interrupt coalescing, disabling deeper sleep states, and favouring quick wake-up. Best suited for real-time or interactive applications.
OLTP: Tuned for online transaction processing (database) workloads, with a focus on high I/O throughput, moderate CPU frequency, and low memory latency to support many concurrent short transactions.
Virtualization-Performance: Prioritises virtual machine performance by reducing virtualization overhead, enabling hardware assist features, and maintaining high CPU and memory bandwidth for guest workloads.
High Performance Compute Flexible: Similar to High Performance Compute but allows some power-saving mechanisms to engage during idle periods, offering a compromise between peak performance and idle energy use.
Static Base Frequency: Locks the CPU to a fixed base frequency (non-turbo), disabling dynamic frequency scaling. Provides predictable performance and power consumption, useful for certification or deterministic environments.
Single Core High Performance: Boosts a single core to its maximum turbo frequency while limiting others to lower states. Designed for workloads that are not multi-threaded and benefit from the highest single-thread performance.
Virtualization-Balance: A balanced policy for virtualized environments, offering reasonable performance for mixed virtual machine workloads while keeping power consumption moderate.
I/O Sensitive: Prioritises I/O throughput and latency over CPU performance, adjusting interrupt handling and storage path parameters to benefit disk or network-intensive applications.
Comprehensive Performance: A holistic performance-focused profile that tunes all subsystems (CPU, memory, I/O) for maximum overall throughput, often used for high-end enterprise applications.
Memory Low Latency-Balance: Tunes memory controller and sub-system parameters to reduce memory access latency while maintaining a balanced power profile. Suitable for memory-sensitive but not purely performance-driven workloads.
Memory Low Latency-Performance: Aggressively optimises memory timing and prefetching to achieve the lowest possible memory latency, at the cost of higher power consumption. Ideal for high-frequency trading or in-memory databases.
SpecPower: Aligns with SPECpower benchmark recommendations, aiming for a good performance-per-watt ratio as measured by standardised benchmarks.
SpecPower-Performance_test: A variant of SpecPower that slightly favours performance over absolute efficiency, often used during performance testing phases.
Balance/Energy Efficiency-00B: A specialised variant of the balanced profile with additional tuning parameters (identified by firmware revision 00B) for specific platform optimisations.
Custom: Allows the administrator to manually fine-tune individual power and performance parameters instead of using a predefined profile.
Note: The default selection may vary by platform and system configuration. Not all profiles are available on every processor or chipset; refer to your platform documentation for the supported options.