There is a recurring theme in consumer computing (much less on servers) where it is customary that some CPUs consume less than others, also with very large differences. The comparison between Intel and AMD here is a must, but although many already know where the shots will go, others will have no idea. We are talking about AMD Package Power Tracking or PPT, a feature that completely distorts the consumption of Ryzen CPUs and that it is convenient to deal with.
Why is it important that a CPU consumes more or less power? The big question is the million and where copper is currently beating in all the large processor design companies, of course also Intel, AMD and Apple among others. Why is this question relevant? For many terms that we are going to touch lightly as an introduction to the parameter that AMD introduced years ago and that has many people confused.
Efficiency over performance
Many will not believe it, but currently efficiency is ahead of performance from the point and time when the wafer manufacturing node is conceived. Achieving overall performance is much easier than managing to maintain a higher watt / GHz / IPC ratio than the previous node, and this is something that also applies to CPU architectures and therefore processors.
Densities of millions of transistors per mm2 apart, it is important to understand that AMD and Intel seek to provide processors with maximum performance but within a fixed limit of power consumption, which is more and more specific. This leaves a narrow margin for maneuver where the processor has to move with the operating system looking at what, when and how any process does as well as the load it receives.
The problem is the triangle (love if you want to be romantic in hardware) between energy, frequency and temperature. You need a balance that allows you to scale in performance without compromising them and this is a daunting task that also has to be simple for the operating system. That’s where the P states and frequency hops were born, so from now on we will focus on AMD to understand a little more how Package Power Tracking or PPT works.
Why is energy so important now?
For the same reason we just explained, for efficiency, but let’s give it another twist to understand the energy issue. In the past and although today it can also be done in processors, if the processor overheated it was synonymous with a high voltage or a high frequency, so by adjusting one or both parameters downwards we could reduce the temperature.
But this was an overly simplistic approach, since the performance drain is much greater thus limiting the total power consumption of the CPU. Are they not linked? Yes, but they are not the same nor does it work the same way. When you are competing for every point in a benchmark or for every watt in a CPU to equally maximize performance, you cannot afford to drop frequency, and the undervolt only affects the CPU and its voltage.
The solution is to directly limit energy consumption and this is where our protagonist comes in.
What is Package Power Tracking or PPT on AMD CPU?
The more or less literal translation of the term would be something like Tracking the power of the packet (of the socket). The feature is almost described by its own name and it is nothing more than the fact that AMD places a limit on the power a CPU can consume through the motherboard‘s rails and power phases.
Here you have to bear in mind that as a good rule it has limits. The first of them refers to processors with TDP of up to 105 watts or higher if necessary, the second does the same but with 65 watts. The function is to limit the power delivery for the first ones in up to 142 watts being this limitation of 88 watts for the second ones (+ -1% of margin or GAP according to the system of phases of the plate).
This is important depending on the processor to be analyzed, since those that are really close to the limit values in full will be able to scale their frequency more due to the PBO algorithm, which is automatic, while those that exceed 142 watts and 88 watts respectively will see their limited frequencies and they will touch those consumptions constantly, but they will not be able to go up from there.
Why is AMD limiting its CPUs with PPT?
Because it aims to offer a higher temperature, performance and efficiency ratio. As we have said, everything revolves around efficiency and leaving free the consumed watts or setting very high limits or for a short time what they do is be less efficient and shoot the temperature.
As expected, if we want we can break that PPT limit and let the CPU consume and go up more MHz, with more voltage and gain more performance, but the temperature will also rise above the 70 ºC that AMD implements. Therefore, processors such as the Ryzen 9 3950X or the 5950X always mark values between 140 and 144 watts on average or at peak (it depends on the stress and the ASIC they have) where in no case will they exceed that value if it is not by work and our grace to break the limit suggested by AMD.
So once the PPT limit is broken in BIOS, AMD CPUs have no more limits and can go up until the cooling allows it? No, there are safety limits that force the lowering to prevent disaster for the sake of processor safety, but this will depend on the resulting temperature, since if we use LN2 this limit is very high and we can simply take the silicon to its physical limit .
Interestingly, the AMD PPT limit is not locked down, that is, in post of the reduction of the energy consumed, which is perfect to maintain the highest performance by reducing the consumption and the temperature, since the jumps in the SKU CPUs are more progressive and not as abrupt as a drop in frequency and voltage, so the PBO algorithm is allowed to scale more frequently due to the drop in temperature, maintaining similar performance ranges.
This is what the Ryzen Master software does with its eco mode, only that instead of in BIOS with a program and with only 3 clicks and without leaving Windows, but of course it is more advisable to do it in the UEFI, it will always be more efficient in these terms.
