One of the key points today in terms of semiconductor design is all about the power consumption and the color transmitted by the components. Because excessive energy consumption generates excessive heat which can shorten the life of the component or, failing that, render it unusable forever.
Another reason is about power consumption, many designs are used are tricks such as separating the energy domain from its different parts, so that when a component is not in use, the power supply is cut off and it stops working. While others are based on decreasing the clock speed if the workload is low and increasing it when it is high.
But for processors to adapt, they need real-time information that marks the temperatures and voltage of the various components, to adapt their clock speeds and to activate and deactivate the various parts of the hardware, either at the SoC level. , or at the level. multiple components on a board.
What are telemetry systems and where are they located?
In reality, telemetry systems are nothing more than small chips which are nothing more than digital thermometers and / or voltmeters, which are responsible for making continuous measurements on the equipment to which they are connected and to transfer this information to a series of microcontrollers. telemetry obtained by surveillance systems, they manage clock speeds, voltage and are even capable of shutting down parts of the processor.
As for their location, it depends, for example they can be found in the same chip as in the form of external components, depending on the specifications and usefulness of each type of processor. In fact, most of today’s SoCs have multiple hardware monitoring systems that send the telemetry data to different microcontrollers.
These are extremely important in SoCs, where the proximity of the components produces what we call thermal throttling, which prevents the different parts of their tight integration from being able to achieve the same clock speeds as separately, so it is essential. that the voltage and the systems are inside the SoC.
What is a microcontroller?
A microcontroller is itself a computer on a chip, with a much higher level of integration than an SoC because they integrate both the processing units and the RAM memory on the same chip, having no communication with the ‘outside only through a series of I / O pins which are used to load the program to be executed recursively.
Microcontrollers began to be used in PCs from the 1983 IBM PC XT in which the Intel 8048 handled the 8086, over time they became more complex and took care of various back-end tasks. -plan such as management of the power and temperature of processors.
The reason that microcontrollers are used and not microprocessors is that by not sharing RAM with the processor, not only access conflicts are avoided, but also malicious code accesses them. However, firmware updates are loaded from certain addresses in the system RAM before being copied to the RAM of each microcontroller during startup.
An example of a microcontroller for telemetry: the AMD SMUIn many diagrams of AMD’s SoCs, CPUs, and GPUs, you will have seen a room called SMU, which you shrug your shoulders on for not knowing what it is and what its features are. If we read the official AMD documentation on what SMU is, we can find the following statement:
The System Management Unit, or SMU in English, is a sub-component of the northbridge which is responsible for various power management tasks during power up of the PC and while the PC is running, which includes a microcontroller to help (in said task).
It must be taken into account that since the appearance of the first x86-64 by AMD what we call northbridge, which is the hardware in charge of the communication of the CPU with the system RAM, is inside the processor, therefore the SMU unit or units are located in the processor itself.
The SMU is not only used by AMD in its processors but also in its GPUs and it is a Lattice LM32 microcontroller, which is licensed by AMD and is responsible for managing everything related to the consumption of energy at any time, the difference is that over time AMD has changed it and there are multiple SMUs for different cores.
For example, in the Ryzen 5000 for laptops, AMD placed a system management unit to manage the power consumption of each of the Zen 3 cores in this processor, so each of them has its own energy domain. and can fluctuate in clock speed and voltage. in a synchronized or independent manner with respect to the rest of the hearts.
The counterpart in the case of Intel is the so-called Management Engine, the function of which is exactly the same. AMD and Intel ME both have the distinction of having a privilege level above the processor itself, having the ability to shut down the CPU in its tracks and the rest of the components if a dangerous situation arises for the PC .