Intel Turbo Boost

Intel Turbo Boost (also Turbo Boost Technology and Turbo Boost for short ) is a function for the automatic overclocking of main processors from Intel , which are used in some versions of the Nehalem micro-architecture . Turbo Boost allows the demand-oriented dynamic increase of the processor clock .

Turbo Boost is activated automatically when the operating system asks for the highest possible performance ("P0-State"). Since these “Performance States” (abbreviated to P-States ) are regulated in the ACPI table, which every modern operating system supports, no additional drivers or additional software are required for the functionality of this technology. This function is also known as automatic or dynamic overclocking function.

functionality

Since older programs are often not designed to use multiple processor cores, they only use one thread or core in a modern multi-core processor , which means that a single-core processor with a higher clock rate can execute such an application faster than a four-core processor with a lower clock rate, even if its overall computing power is higher than that of the single core processor. Since the thermal design power ( TDP for short ; also thermal budget ) for a processor includes the maximum power dissipation of the entire processor that can be dissipated by the cooling system, a processor in which only one core is fully utilized does not fully utilize the TDP. If the processor does not fully use its thermal budget due to inactive cores, a core can use part of the budget that has not yet been used without the cooling device having to dissipate more heat than it would in a state in which all cores are fully active, must be able to dissipate. In order for a core to be able to use the unused part of the TDP, the temperature and the current and power consumption must be below certain limits with Turbo Boost . The temperature is read out by a sensor in the processor. The processor itself knows the core voltage because it has transmitted it as a target value to the voltage converter on the motherboard . Via an extra line, the processor receives the information from the voltage converter via a rather imprecise, analog signal, how much current it is currently delivering, but multiplied by the core voltage, this information is still sufficient for the processor to estimate and thus check the approximate power consumption whether the TDP is already exhausted.

So that the part of the TDP that would normally be assigned to a processor core is considered unused, the core must at least remain in (deep) sleep mode (referred to as the “C3” state in the ACPI table). This means that in this mode neither of the two parallel threads of a physical core simulated to the operating system by means of Hyper-Threading are allowed to execute a process. If this is the case and the operating system requests the highest possible performance for at least one core, the processor automatically increases the multiplier in individual steps, with the clock rate per multiplier level (in Intel jargon these turbo levels are also referred to as "speed bins") is increased by 133.33 or 100  MHz . The maximum achievable multiplier level 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) cores, whereby Intel stores the information by how many multiplier levels may be increased with how many active cores, as well as the maximum achievable multiplier level in the processor. If the processor exceeds one of the given heat, power or power limit values, it goes down again in equal multiplier steps, whereby a processor can change its clock frequency up to 200 times per second in this way.

Whether a processor (especially under full load, when the operating system reports a load of 100% for all cores) actually fully utilizes its TDP depends not least on the program used. For example, if a program only uses the integer unit, but leaves the floating-point arithmetic units unused, less heat is generated than with code, which uses all arithmetic units equally. If Turbo Boost is not switched on despite the temperature, current and power values being below the limit, this is sometimes due to a BIOS setting that completely deactivates this function. Another possibility is that the power saving function (“ SpeedStep ” or “EIST”) has been switched off in the BIOS , thus denying the processor the C-States, which means that a core is always considered active and the part of the TDP assigned to it is not can be redistributed.

Example of how it works

A Core i7-920XM Extreme Edition has a standard clock frequency of 2,000 MHz. The highest possible turbo levels are 2 with four and three active cores, 8 with two active cores and 9 with only one active core. With each increased turbo level, the multiplier increases by one level and thus the clock rate by 133.33 MHz. Provided that the temperature, current and power limit values ​​have not yet been exceeded, the processor can achieve the following maximum clock frequencies:

3 or 4	Cores:	2,000 MHz	+	2 x 133.33 MHz	=	2,000 MHz	+	266.66 MHz	≈	2,267 MHz
2	Cores:	2,000 MHz	+	8 x 133.33 MHz	=	2,000 MHz	+	1,066.64 MHz	≈	3,067 MHz
1	Core:	2,000 MHz	+	9 x 133.33 MHz	=	2,000 MHz	+	1,199.97 MHz	≈	3,200 MHz

Since the Sandy Bridge micro-architecture, the reference clock has been 100 instead of 133.33 MHz.

A detailed frequency table of the individual CPUs can be found on the Intel support pages:

Processors with Turbo Boost

Turbo Boost is only supported by the Core i5 processor families and the Core i7 processors. Version 2.0 of Turbo Boost introduced with Sandy Bridge also only supports the Core i5 and i7 processors. This technology is not integrated in all Pentium and Core i3 processors.

history

A white paper from Intel from November 2008 reports on Turbo Boost , which is to be used for the first time as a new technology in the processors based on the Nehalem microarchitecture and released that same month.

A Turbo Boost- like technology called Intel Dynamic Acceleration (IDA) was already included in some Centrino processors based on the Core 2 or Core microarchitecture . However, this technology has not received the same level of popularity and attention that Turbo Boost has received. IDA has been disabled in the BIOS of most laptops. Another problem with this technology was the difficulty in taking advantage of this technology unless the system was started in single core mode.

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

disadvantage

Turbo Boost offers the theoretical possibility of running programs that only use one thread faster, but in practice this advantage is difficult to use on a normal system. In order to go beyond the maximum possible multiplier level (with Core i5 and Core i7, this is 0 to 2 levels for all processors that support Turbo Boost ) with three and four active cores, at least two cores must be in C3 sleep mode. Modern operating systems distribute the applications to be calculated and executed evenly across all available cores, which means that a processor core rarely changes to C3 mode, which in turn means that it is not considered inactive and its TDP component cannot be redistributed. Another disadvantage is found in Windows, which processes processes that are not optimized for multiple threads ( single thread applications ), e.g. B. a quad-core processor and then processes 25% of the code on each core instead of calculating 100% on one core. Although these are single-threaded programs and so only one core is active, the other cores need time to switch to C3 mode and the active core time to increase the turbo levels, which increases the clock frequency in this variant is lower than if the program did not switch between the different cores. Technologies such as core parking can eliminate these disadvantages, but must first establish themselves or establish themselves in the individual operating systems.

Differences to Turbo Core

Intel's Turbo Boost differs from AMD's Turbo Core in several ways . One of the big differences is that Turbo Boost can redistribute the TDP portion of an inactive processor core and not - unlike Turbo Core - at least half of the cores must be considered inactive for the processor to switch on the Turbo at all and thus a higher clock rate can reach. While Turbo Core can only increase the processor's clock rate by a fixed level (400 or 500 MHz depending on the processor), some mobile quad-core processors from Intel can increase the clock rate by up to 9 multiplier levels with only one active core. Since it is seldom the case with modern operating systems that only one core of the processor is loaded, the highest multiplier level is rarely or not at all achieved in practice.

While with Turbo Core processor cores in C1 state are already considered inactive, with Turbo Boost a core is only considered inactive when it remains in C3 state. The disadvantage of this is on the one hand that the processor core can switch to the C1 state more quickly and can "wake up" from it again; on the other hand, it is more likely that the processor is not in C1 with any processor load or is in a deeper sleep state than in ≥ C3 state. Unlike Turbo Core , Turbo Boost also requires an additional entry in the BIOS. AMD has also separated the P-states of the processor from the P-states of the operating system. However, if the operating system requests full performance (P0 state), the processor first goes to P1 instead of P0 state. The P1 state does not stand for the first power-saving level, as usual, but for working with the nominal cycle, which means that the turbo cycle has its own entry as the highest possible performance. This creates a separate entry for the turbo stage in the ACPI table, with which the processor can, for example, set a higher core voltage in the turbo stage. The separation of the P0 state of the operating system and the P0 state of the processor caused problems with some versions of the Linux kernel , whereby the kernels missed out an incorrect clock speed and as a result the system ran up to 27% slower, the kernels from 2.6. 35 and patched versions of the old kernel versions no longer have this problem.

Turbo Boost Technology 2.0

Turbo Boost 2.0 is the further development of the first generation, which was used for the first time in the processors with Sandy Bridge micro- architecture . As with the old version, 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 still do not use Turbo Boost 2.0.

What is new in version 2.0 is that, under certain conditions, the processors may overclock all cores higher for a short period of time and exceed the TDP, whereby more heat loss is produced for a short period than the cooler can dissipate in the long term, whereby the thermal inertia of the CPU Cooler is used. Depending on the situation, this state can typically last up to 25 seconds. In addition to the dynamic overclocking of the main processor, since Sandy Bridge the graphics processor has also been overclocked during load, which was previously reserved for mobile processors. The CPU and integrated GPU share the TDP, which means that, depending on the software, sometimes the share of the CPU and sometimes that of the GPU forms the greater part of the heat loss during operation.

Since the reference clock is only 100 instead of 133.33 MHz on the new motherboards with chipsets of the 6 series, the clock increases of the individual levels are also lower. How many clock levels are switched up with how many cores that are active at the same time still differs depending on the processor.

See also

• AMD Turbo Core

Web links

• The Basics of Intel Turbo Boost (Server Mile Technet)

Individual evidence

1. ↑ Intel® Core ™ i5 Desktop Processor Turbo boost frequency table
2. ↑ Intel® Core ™ i7 Desktop Processor Intel® Turbo Boost Technology frequency table
3. ↑ Intel® Turbo Boost Technology in Intel® Core ™ Microarchitecture (Nehalem) Based Processors
4. ↑ Intel Launches Fastest Processor on the Planet
5. ↑ Tech ARP - Intel Dynamic Acceleration ( Memento of the original from July 21, 2011 in the Internet Archive ) Info: The archive link was inserted automatically and has not yet been checked. Please check the original and archive link according to the instructions and then remove this notice.  
6. ↑ Intel's Turbo Boost: Lynnfield Gets Afterburners: Intel Core i5 And Core i7: Intel's Mainstream Magnum Opus
7. ↑ Core Parking in Windows Server 2008 R2 and Windows 7 - Dr. Dobb's and Intel Go Parallel Programming ( Memento from March 17, 2010 in the Internet Archive )