| Andre Przywara | 615b730 | 2012-09-04 08:28:07 +0000 | [diff] [blame] | 1 | Processor boosting control | 
|  | 2 |  | 
|  | 3 | - information for users - | 
|  | 4 |  | 
|  | 5 | Quick guide for the impatient: | 
|  | 6 | -------------------- | 
|  | 7 | /sys/devices/system/cpu/cpufreq/boost | 
|  | 8 | controls the boost setting for the whole system. You can read and write | 
|  | 9 | that file with either "0" (boosting disabled) or "1" (boosting allowed). | 
|  | 10 | Reading or writing 1 does not mean that the system is boosting at this | 
|  | 11 | very moment, but only that the CPU _may_ raise the frequency at it's | 
|  | 12 | discretion. | 
|  | 13 | -------------------- | 
|  | 14 |  | 
|  | 15 | Introduction | 
|  | 16 | ------------- | 
|  | 17 | Some CPUs support a functionality to raise the operating frequency of | 
|  | 18 | some cores in a multi-core package if certain conditions apply, mostly | 
|  | 19 | if the whole chip is not fully utilized and below it's intended thermal | 
|  | 20 | budget. This is done without operating system control by a combination | 
|  | 21 | of hardware and firmware. | 
|  | 22 | On Intel CPUs this is called "Turbo Boost", AMD calls it "Turbo-Core", | 
|  | 23 | in technical documentation "Core performance boost". In Linux we use | 
|  | 24 | the term "boost" for convenience. | 
|  | 25 |  | 
|  | 26 | Rationale for disable switch | 
|  | 27 | ---------------------------- | 
|  | 28 |  | 
|  | 29 | Though the idea is to just give better performance without any user | 
|  | 30 | intervention, sometimes the need arises to disable this functionality. | 
|  | 31 | Most systems offer a switch in the (BIOS) firmware to disable the | 
|  | 32 | functionality at all, but a more fine-grained and dynamic control would | 
|  | 33 | be desirable: | 
|  | 34 | 1. While running benchmarks, reproducible results are important. Since | 
|  | 35 | the boosting functionality depends on the load of the whole package, | 
|  | 36 | single thread performance can vary. By explicitly disabling the boost | 
|  | 37 | functionality at least for the benchmark's run-time the system will run | 
|  | 38 | at a fixed frequency and results are reproducible again. | 
|  | 39 | 2. To examine the impact of the boosting functionality it is helpful | 
|  | 40 | to do tests with and without boosting. | 
|  | 41 | 3. Boosting means overclocking the processor, though under controlled | 
|  | 42 | conditions. By raising the frequency and the voltage the processor | 
|  | 43 | will consume more power than without the boosting, which may be | 
|  | 44 | undesirable for instance for mobile users. Disabling boosting may | 
|  | 45 | save power here, though this depends on the workload. | 
|  | 46 |  | 
|  | 47 |  | 
|  | 48 | User controlled switch | 
|  | 49 | ---------------------- | 
|  | 50 |  | 
|  | 51 | To allow the user to toggle the boosting functionality, the acpi-cpufreq | 
|  | 52 | driver exports a sysfs knob to disable it. There is a file: | 
|  | 53 | /sys/devices/system/cpu/cpufreq/boost | 
|  | 54 | which can either read "0" (boosting disabled) or "1" (boosting enabled). | 
|  | 55 | Reading the file is always supported, even if the processor does not | 
|  | 56 | support boosting. In this case the file will be read-only and always | 
|  | 57 | reads as "0". Explicitly changing the permissions and writing to that | 
|  | 58 | file anyway will return EINVAL. | 
|  | 59 |  | 
|  | 60 | On supported CPUs one can write either a "0" or a "1" into this file. | 
|  | 61 | This will either disable the boost functionality on all cores in the | 
|  | 62 | whole system (0) or will allow the hardware to boost at will (1). | 
|  | 63 |  | 
|  | 64 | Writing a "1" does not explicitly boost the system, but just allows the | 
|  | 65 | CPU (and the firmware) to boost at their discretion. Some implementations | 
|  | 66 | take external factors like the chip's temperature into account, so | 
|  | 67 | boosting once does not necessarily mean that it will occur every time | 
|  | 68 | even using the exact same software setup. | 
|  | 69 |  | 
|  | 70 |  | 
|  | 71 | AMD legacy cpb switch | 
|  | 72 | --------------------- | 
|  | 73 | The AMD powernow-k8 driver used to support a very similar switch to | 
|  | 74 | disable or enable the "Core Performance Boost" feature of some AMD CPUs. | 
|  | 75 | This switch was instantiated in each CPU's cpufreq directory | 
|  | 76 | (/sys/devices/system/cpu[0-9]*/cpufreq) and was called "cpb". | 
|  | 77 | Though the per CPU existence hints at a more fine grained control, the | 
|  | 78 | actual implementation only supported a system-global switch semantics, | 
|  | 79 | which was simply reflected into each CPU's file. Writing a 0 or 1 into it | 
|  | 80 | would pull the other CPUs to the same state. | 
|  | 81 | For compatibility reasons this file and its behavior is still supported | 
|  | 82 | on AMD CPUs, though it is now protected by a config switch | 
|  | 83 | (X86_ACPI_CPUFREQ_CPB). On Intel CPUs this file will never be created, | 
|  | 84 | even with the config option set. | 
|  | 85 | This functionality is considered legacy and will be removed in some future | 
|  | 86 | kernel version. | 
|  | 87 |  | 
|  | 88 | More fine grained boosting control | 
|  | 89 | ---------------------------------- | 
|  | 90 |  | 
|  | 91 | Technically it is possible to switch the boosting functionality at least | 
|  | 92 | on a per package basis, for some CPUs even per core. Currently the driver | 
|  | 93 | does not support it, but this may be implemented in the future. |