SpeedStep and all-all-all

We all know that energy is not a cheap thing and is not commonly found. And, as they say, the farther, the more terrible - that is why all of humanity is concerned about its economy. The approaches to this economy are very diverse, here there are windmills, electric cars, and the universal accounting of everything and everything, and much more - the main thing is that the movement is everywhere. Manufacturers of digital devices are not lagging behind the trend, but in some ways they are even ahead of it, at least the results in this area are more substantial and real than in others. To verify this, let's talk about energy-saving technologies in Intel processors.

The word “energy saving” first appeared in Intel specifications at the end of the last century, when SpeedStep technology for Intel Pentium III mobile processors was announced, which provided for the possibility of lowering the processor frequency during its inactivity. Initially, in the first version of SpeedStep, there were only two operating frequencies - normal and low. Accordingly, there were two indicators of heat dissipation and power consumption.
Further, the evolution of SpeedStep went along the path of complicating the processor power circuits. The more advanced Enhanced Intel SpeedStep (EIST) technology already operated with several intermediate points between the maximum and minimum possible processor frequencies, in addition to the frequency, the voltage supplied to the processor also changed. The EIST work schedule for many Pentium M processors that are memorable can be seen in the figure below.


As you can see, the processor frequency was increased by 200 MHz quanta, while its TDP varied from 6 to 24.5 watts.
For the SpeedStep to work correctly, the technology must be supported by the motherboard (or, more precisely, its BIOS) and the operating system. Consumption modes are controlled from the BIOS or the corresponding OS settings menu. The fact that SpeedStep really works can be verified using utilities that show the actual current frequency of the processor, for example, CPU-Z.
Beginning with the Pentium M, there have been no fundamental changes in EIST, there is a constant adjustment for each new generation of processors, which has its own, different indicators of operating voltage, natural frequency and frequency of the FSB central bus. Say, the Intel Core 2 Duo processor can vary its frequency from 1.6 to 2.8 GHz with an increment of 266 MHz.

However, not a single SpeedStep provides energy savings in Intel processors. Other technologies are also used, for example, Ultra Fine Grained Power Control, which allows you to disable those processor modules that are currently idle. It is curious that even if the processor as such is very heavily loaded, it still has idle sections that can be turned off if desired.


The working moment of the processor. Idle areas are shown in green.

Another means of manipulating the processor frequency is Intel Turbo Boost Technology. The principle of its operation is directly opposite to SpeedStep: the processor frequency does not decrease, but increases beyond the standard frequencies within its capabilities, which are indicated in the specification. Each processor supporting Turbo Boost is marked with approximately the following entry: 7/7/9/10, where the dash indicates the number of maximum increment portions (for Sandy Bridge processors, for example, the increment is 100 MHz), which can be added with all four cores running , three two and one, respectively. Thus, if the base frequency of the processor is 2500 MHz, then the maximum, taking into account Turbo Boost, will be equal to 2500 + (10 × 100) = 2500 + 1000 = 3500 MHz.


BIOS Management for SpeedStep and Turbo Boost

At first glance, Turbo Boost does not look like an energy-saving system, but in fact it is just that, since an increase in consumption due to an increase in frequency is often more than compensated by a decrease in the time it takes to complete a task.

Another important front in the fight against excess calorie watts is the reduction in overall processor consumption. Significant progress has also been made along this path. In the article on heat dissipation , a good plate was given, which shows that Intel processors, having passed the maximum gluttony in the Prescott area, have since significantly reduced their TDP even in absolute terms, not to mention the conversion relative to gigahertz cores.


Process reduction + microarchitecture optimization = energy efficiency

A new milestone on the path to a bright, future-saving future will be the next Intel Haswell processor microarchitecture, which we will see embodied in silicon in 2013. The Haswell line of processors will consist, as usual, of the desktop and mobile versions, but they will also add a special option for ultrabooks (with the U index). New CPUs for PCs and laptops consume 20% less than the Sandy Bridge family. Of particular interest is Haswell U, which is a single-chip solution with an integrated I / O controller and voltage regulator on board, which will allow more accurate and faster control of this voltage. Intel emphasizes that in sleep mode (with the ultrabook closed) it consumes 20 times less energy than its predecessors.

As we can see, with each new generation of the Intel processor, more and more third-party elements are absorbed. From the point of view of energy efficiency, this is undoubtedly a benefit, since it reduces overall costs. What will we optimize when the processor reaches the extremely small technological process and collects all the electronics of the motherboard inside? Surely, some options will remain even in this case.