Hay muchas ocasiones, y depende del juego, en las que a pesar de tener all los adjuses optimizados por algún motivo, la GPU no te está dando todo su potencial, y en las que el rendimiento en el juego no sea malo, pero podría ser better. The first thing we recommend is to remove any kind of synchronization of the monitor with the graphics card, be it V-SYNC, FreeSync or G-SYNC, all in any of their terms. We do this to determine the specific problem and not have an FPS rate limit blocked by these technologies and thus be able to appreciate the framerate increases, but above all it will allow us to observe the behavior of the CPU and the GPU without any limit and understand the problems.
Why isn’t the GPU fully utilized in all games?
Besides your graphics card, many factors affect game performance: CPU, RAM, storage, temperature, the graphics API used by the game and, of course, its optimization. For this reason, if you want to know why the GPU is not working at 100% in some games, you must start by investigating the rest of the components, as well as of course making sure that you have the latest version of the graphics drivers and obviously the latest version of the operating system as well as graphics APIs.
If you have a monitoring tool like the one we showed you above (it’s Riva Tuner via MSI Afterburner) you will be able to see, in addition to FPS and GPU load, the temperatures and the load of the processor. This will help us understand the problem, because if the GPU load isn’t at 100%, but the CPU is, then you already have the answer: your CPU is bottlenecked. In this example, the causes would need to be determined, as it could be that another process is in the background consuming resources, or it is the game that is pushing the CPU to the limit.
In other words, a monitoring tool can show you the total load of all CPU cores (in our example it’s only 27%), but you need to check if that’s the case for each core individual.
For example, on an 8-core processor, you might have two at 100% and the rest nearly idle; the total would be 25% utilization so apparently the CPU wouldn’t be an issue but that would be because the game only supports CPU processing on two cores and they are maxed out but that doesn’t mean that. And having two or more cores loaded without rendering them all can also mean that the game’s graphics engine doesn’t parallelize or isn’t well-designed from a software standpoint.
This should be clear, because one game does not represent all or a large part of them. When a CPU is bottlenecked, its load is very high in one or more cores, regardless of the game you are playing (we always talk about “well” optimized games) and we explain ourselves.
There are titles that, in terms of requirements, are very affordable for any computer and, on the other hand, the processor is always under high load on one or more cores. It’s not a bottleneck, it’s bad optimization, like CS:GO or the newer WarZone, which had a high CPU usage issue due to a bug in a patch.
The Problems of Poor Software Optimization
In this case, it would be the fault of a poor multithreaded optimization gaming and there was not much you could do except upgrade your CPU to one with higher performance per core, but not much else. Hopefully you can reduce the graphics load which, while increasing the FPS the GPU is capable of, will reduce the CPU call requirements of RAM and graphics, as well as the SSD.
RAM performance can also prevent the GPU from running at 100% in some games, especially if they have high requirements. You can check this by lowering some in-game graphics settings, especially those related to focal length and texture quality, and see if easing them in addition to getting more FPS increases the load on the GPU.
If so, check that the RAM is running at its correct speed and not base, and make sure you are running it in dual or quad channel mode (depending on your system). To check it, you can use both Aida64 as CPU-Zfor example.
Another possible issue with RAM (games are becoming increasingly dependent on it due to importing and exporting textures from the SSD) is its stability. That the configuration is correct does not imply that the memory is physically correct. We recommend taking specific stress tests such as Runmemtest Pro or Karhu RAMTest and checking after 400% success that neither they nor BMI fail logically.
SSD issues limiting GPU to 100%
There is a problem that is little talked about and that affects more and more users: the performance of PCIe SSDs. We’ve gone from just 500MB/s with technology well above those speeds to hitting the technical limit of those speeds with the new PCIe 4.0 and soon to be 5.0 SSDs.
What happens to them? Well, there are two key factors, one of which is more controlled than the other: temperature and cache or controller collapse. The first is one that should be within optimal ranges, either because of the motherboard’s own heatsink or because we purchase a specific heatsink for it.
But as a general rule, an SSD should always be at 50% capacity and below 70º C if its maximum performance is to be desired. We have already explained this in other articles, so instead of repeating it, we will simply say that it is inexorable and linked to the NAND Flash technology itself, there is no workaround and it must be done. to fill.
The second handicap is the SSD cache or controller. Keep in mind that almost all manufacturers leave to Windows and its AHCI drivers the control and management of the inputs and outputs that this component has with the rest of the PC, as well as the management and control of the PCIe.
This means that there are situations where data is not being accessed and read in the correct cycles due to a series of cyclic redundancy errors that the controller algorithm itself and Windows TRIM need to resolve, but don’t. don’t. This creates a delay which in some cases leads to excessive usage of the SSD itself, but in most cases it produces high random times which reduce read, write and IOPS performance, especially the last two.
TRIM management and gaming performance
It’s always been said that defragmenting an SSD isn’t necessary and that it hurts its lifespan, and that’s not true at all. The algorithm of TRIM management in some controllers and with generic windows drivers it doesn’t work as it should and what defragmenting really does is force TRIM to work solving SSD performance issues and impacting correct handling of the processor for this and system RAM, lowering its usage from 100% to optimal levels, which could be 98% if bottlenecked or 2x% if left in that section, everything will be the team object that we have in hand.
There is a command that helps optimize TRIM that could be the solution to SSD performance and therefore CPU or GPU performance and is worth trying. It may not solve the problems, or maybe. In any case, it must be run from Windows PowerShell as administrator:
Optimize-Volume -DriveLetter C -ReTrim -Verbose
Once you are done optimizing the TRIM, we can close PowerShell and check if we have better performance and CPU or GPU below 100%. We hope that with these tips and tricks