Intel Turbo Boost

Intel Turbo Boost ( also Turbo Boost Technology) is a function for automatic overclocking of CPUs from Intel, which are used in some versions of the Nehalem microarchitecture use. Turbo Boost allows the on-demand dynamic increase of the processor clock.

Turbo Boost is activated automatically when the operating system requests the highest performance ( " P0 -state "). Since these "Performance States " ( abbreviated as P-states ) are governed by the ACPI table, which support any modern operating system, there are no additional drivers or additional software needed for the functionality of the technology. This function may be referred to as automatic or dynamic overclocking feature.

Operation

Since older programs are often not designed for the use of multiple cores, they also use to a modern multi-core processor only one thread or core, creating a higher clocked single-core processor can perform such an application faster than a quad-core processor with a lower clock, even if its total computing power is higher than that of the Einkernprozessors. Since the thermal design power ( TDP short, nor thermal budget ) for a processor, the maximum thermal through the cooling system power dissipation of the whole processor includes irritates a processor in which only one core is fully utilized, the TDP not completely. Requires the processor to be thermal budget due to inactive cores not complete, you can use a portion of the unspent budget, for example, a core without the cooling device would dissipate more heat than it, in a state in which all cores are fully active must be able to dissipate. Thus, a core can use the unused portion of the TDP, the temperature and the current and power consumption must be below certain limits with Turbo Boost though. The temperature is read by a sensor in the processor. The core voltage of the processor knows himself, because he transmitted it as the target value to the voltage converter on the motherboard. About an extra line to the processor from the voltage converter receives a although quite inaccurate, analog signal information on how much electricity it delivers just multiplied by the core voltage reaches this information for the processor but for anyway to estimate the approximate power consumption and check order to whether the TDP is already maxed out.

So that the part of the TDP that would normally be allocated to a processor core, as unused valid, the core must be at least in the ( deep) sleep mode ( in the ACPI table as a "C3" state hereinafter) dwell. This means that in this mode, neither the operating system using Hyper -Threading simulated parallel threads of a physical core can run a process. If this is the case and is required by the operating system the highest possible performance for at least a core multiplier is increased automatically by the processor in single steps, where each multiplier stage ( in Intel lingo, this turbo stages referred to as " speed bins' ) of clock is increased to 133.33 MHz. The maximum attainable multiplier stage depends on the number of active (C0 "Operating State " C1 " Halt State " or C1E "Enhanced Halt State " and C2 " Stop Grant " or C2E "Enhanced Stop Grant " ) or sleeping (≥ C3 -state ) from nuclei, with Intel active how many cores can be increased in the manufacture of the information to the number of stages in the multiplier, and the maximum achievable multiplier stage in the processor stores. If the processor exceeds one of the given heat, electricity or power limits, it goes into the same multiplier levels back down, where a processor can change up to 200 times per second, its frequency in this way.

Whether a processor (especially under full load when the operating system for all the cores, a 100% utilization reports ) its TDP actually fully exploited, depends not least on the program used. But is under a program, for example, only the integer unit allows that Gleitkommarechenwerke unused, means less heat than code that auslastet all arithmetic units alike. When Turbo Boost is not turned on despite lying below the limit temperature, current and power values ​​, this is sometimes at a BIOS setting that disables this feature completely. Another possibility is that in the BIOS, the power saving function ( " SpeedStep " or " EIST " ) has not been turned off, and thus the processor be denied the C-States, with which a core is always considered active and thus the assigned part of the TDP can be redistributed.

Example of Operation

A Core i7- 920XM Extreme Edition has a standard clock speed of 2,000 MHz. The highest possible turbo stage 2 with four and three active cores, 8 and 9 with two active nuclei with only one active core. With each raised Turbo setting the multiplier increases by one step, and thus the clock at 133.33 MHz. Provided that the temperature, current, and power limits have not been exceeded, the processor can thus obtain the following maximum clock frequencies:

• 3 or 4 cores active: 2.000 2 × 2.000 = 133.33 266.66 ≈ 2,267 *
• 2 cores active: 2000 8 × 133.33 = 2000 1066.64 ≈ 3,067 *
• 1 core active: 2000 9 × 133.33 = 2000 1199.97 ≈ 3,200 *

* All data (except the digit multiplier steps) are given in MHz

Since the Sandy Bridge microarchitecture is the reference clock 100 instead of 133.33 MHz.

Processors with Turbo Boost

Currently, Turbo Boost is only supported by the processor Core i5 families of the 600 and 700 series as well as the Core i7 processors in the 800 and 900 series. The introduced with Sandy Bridge version 2.0 of Turbo Boost currently only support the Core i5 and i7 processors 2xxx with a version number in the field. In all Pentium and Core i3 processors, this technology is not integrated.

History

In a white paper by Intel in November 2008 is reported via Turbo Boost, which as a new technology based on the Nehalem microarchitecture, published in the same month processors is to be used for the first time.

A similar technology called Turbo Boost Intel Dynamic Acceleration (IDA ) was included in some of the Core 2 or Core microarchitecture -based Centrino processors. This technology was not the same distribution and attention to part as she learns Turbo Boost. Thus, IDA has been disabled in the BIOS of most laptops. Another problem with this technology has been the difficulty to take the advantage of this technology, if the system has not started the Einkernmodus.

2011, Intel released version 2.0, which was used for the first time in the main processors of the Sandy Bridge family.

Disadvantages

While Turbo Boost allows the theoretical possibility programs that only use one thread to run faster, practically can be on a normal system, this advantage but use plain bad. Order of the maximum possible multiplier stage (with Core i5 and Core i7, these are for all processors, Turbo Boost support, 0-2 steps) get out at three and four active cores, at least two cores must be located in the C3 sleep mode. Modern operating systems distribute the work to be calculated and executed applications evenly across all available cores, enabling a processor core rarely enters the C3 mode, which he in turn shall not be deemed inactive and thus be TDP- share can not be redistributed. Another drawback occurs when Windows that processes that are not optimized for multiple threads ( single-threaded applications), eg in a quad-core processor divides and then on each core executes 25 % of the code rather than 100 % on one core to be calculated. Though it is single-threaded programs, and thus only a core is active, the other cores, the clock frequency need time to move in the C3 time mode and the active core in order to increase the turbine stages, whereby in this variant is lower than if the program would not switch between the different cores. Technologies such as Core Parking can remove these disadvantages, but first need to establish or interspersed in isolated operating system.

Differences with Turbo Core

Intel's Turbo Boost differs in several parts of AMD's Turbo Core. One of the big differences is that Turbo Boost the TDP proportion of inactive processor core can be equal to redistribute and not - unlike Turbo Core - at least half of the nuclei must be considered inactive, so that the processor ever turn on the turbo and thus a higher clock speed can achieve. While Turbo Core can increase the clock rate of the processor only at a fixed level ( depending on the processor 400 or 500 MHz ), some mobile quad-core processors from Intel can increase the clock multiplier up to 9 levels with only one active core. Since in modern operating systems, but is rarely the case that only one core of the processor is busy, the highest multiplier level cap is up but not rarely achieved in practice.

While already apply when Turbo Core processor cores in the C1 state as inactive with Turbo Boost a core shall not be deemed inactive if it remains in the C3 state. The disadvantage of this is due to the fact that the processor core switch faster in the C1 state and " wake up " from this again can be, for others it is more likely that the processor is in no processor load in C1 or in a deeper sleep state than in the ≥ C3 state. Turbo Boost needed - unlike Turbo Core - moreover an additional entry in the BIOS. AMD has further separate the P-states of the processor P-states of the operating system. Calls the operating system full power to (P0 -State), processor, however, first in the P1 rather than in the P0 state. The P1- State stands is not the same for the first stage of power saving, but for working with nominal clock, making the Turbo clock gets its own entry as the highest possible performance. So in the ACPI table is created a separate entry for the Turbo setting, what the processor, for example, may set a higher core voltage in the Turbo setting. Although caused the separation of the P0 state of the operating system and the P0 state of the processor in some versions of the Linux kernel problems, the kernel auslasen an incorrect number of cycles and the system up to 27 % slower ran in a row, the Kernel 2.6. However, 35 and patched versions of the old kernel versions do not have this problem.

Turbo Boost Technology 2.0

Turbo Boost 2.0 is the evolution of the first generation that was used for the first time in the processors with Sandy Bridge microarchitecture. As with the old and the new version is only available on systems with Core i5 and i7 processors with a version number in the range 2xxx. Core i3 and Pentium processors use Turbo Boost 2.0 still does not.

New feature of version 2.0 is that the processors under certain conditions all the cores for a short period overclock higher and may exceed the TDP, where for a short period, more loss of heat is produced can dissipate as the radiator on a permanent basis, in which case the thermal inertia of the CPU cooler is utilized. Depending on the situation can typically last up to 25 seconds of this condition. In addition to the dynamic overclocking of the main processor since Sandy Bridge and the GPU is overclocked at load, which was previously reserved for mobile processors. CPU and integrated GPU will share a TDP, which times the proportion of the CPU and the GPU times forms the larger part of the heat loss of power during operation depending on the software.

Since the reference clock is on the new motherboards with the chipsets of the 6- series only 100 instead of 133.33 MHz, the clock increases of various levels also fall lower. How many clock stages at how many simultaneously active cores each upshift is still vary depending on the processor.