First of all, we must clarify that this article is not a tutorial about how to increase the clock speed or overclock a specific CPU or GPU model, but we are going to explain how, through artificial intelligence algorithms, more processors Advanced are capable of dynamically adjusting the clock speed to the thermal and consumption situation of our PCs.
Let’s not forget that even if we talk about a form factor as widely used as a laptop, each model, and even within the same brand, has its own thermal, energy consumption and ventilation specifications.
Therefore, contrary to what happens with exclusive designs, such as a video game console, in the design of the different hardware components in PC, especially the CPU and GPU, they have to create mechanisms that allow the hardware to adjust a series of parameters automatically according to the thermal or consumption information at all times.
One of the peculiarities of artificial intelligence is its ability to make predictions through the data obtained at all times and we will have heard in the middle of the marketing of some other company of an advanced overclocking system based on AI that allows to adjust a lot the better the clock speed of a specific processor or graphics card. Well then, let’s demystify these pieces of hardware.
How does Artificial Intelligence use overclocking?
When we talk about artificial intelligence we are not talking about machines that are self-aware of themselves or of their own thinking, but about a combination of hardware and software that is used mainly for two types of tasks mainly:
- The classification of the data.
- Prediction of the evolution of the data.
Starting with the data classification, this is understood in that by labeling a set of information we teach the system to identify information and with it it can categorize it. How does this help with overclocking? Well, it is simple, each instruction of the CPU has a specific energy cost, in such a way that when it is executed the overclocking system can take notes and collect information about the consumption of each one of them. What is achieved with it? Well, that we can execute as many instructions as possible at the highest speed that allows a constant consumption of the processor, which translates into being able to accelerate the execution of the programs more efficiently.
As for the other aspect, that of prediction, we must remember first of all what a mathematical function is, which in a simplified way is defined as the relationship between at least two functions that are dependent on each other where the dependency is done in the form of a mathematical formula. Do not be alarmed, our goal is not to do an algebra class, but if your memory fails you we can reduce the explanation that a function is a mathematical equation where one half defines the value of the other half.
For example, suppose our function is X2, So if we give values to the unknown and we draw them on a graph, it would end up making a parabola like the one in the image that you have just above these lines, this allows us to make a prediction towards where the function will mathematically evolve.
Well, artificial intelligence prediction systems work the other way around. So they do not start with a formula and from there they take a set of data, but the path is the other way around and what they do is collect a large amount of information to create a prediction, which is a formula that they will derive of it. When this formula is applied, it is compared with the original data to see if the resulting data is the same. And what happens if the AI is wrong? He simply performs a movement called backtraking that allows him to polish the resulting formula.
Consumption and temperature data
Every machine learning system requires large amounts to work and while traditional overclocking systems use a series of internal tables in common with all models of a processor that indicates the relationships between clock speeds, temperatures, consumption and time in the which can be together as reference elements, the artificial intelligence-based overclock instead get their data on time of operation. The reason for this is very simple, for reasons that are beyond the control of the manufacturer in the creation of a new wafer, not all chips come out the same in it.
One way to get the thermal and consumption information is the use of electromechanical components that make advanced telemetry systems in key parts of the circuitry and give the information in real time. This allows the overclocking system to collect data that will use it to infer a power consumption formula that allows the system to have a stable overclock system. Obviously, since this system is based on AI, it has to have a supervision system, which acts as a security system that will automatically adjust the voltages and clock speed when critical speeds are reached.
Little by little, the AI overclock system will polish the resulting mathematical function, allowing a more energy efficient clock speed adjustment while allowing it to last longer.
Goodbye to traditional overclocking?
No, we do not believe that overclocking through artificial intelligence will make the traditional one disappear, mainly due to the fact that we see it as a complementary element that will make it safer to increase the speed of both our CPUs and our GPUs. In addition, overclocking is a rising value among hardware enthusiasts and there are users always willing to pay more to be able to fiddle with their hardware, if in addition the manufacturers give them a totally safe environment where the system adapts to the chosen cooling system. by the user as then the levels of customization and modding can reach new levels.
These new intelligent overclocking systems more advanced than the previous ones also allow not only a part of the system to adjust to consumption, but the whole system can do it in unison.