The performance of an integrated GPU is inferior, in most cases, to that of a dedicated GPU, that does not admit discussion. However, over the years the situation has noticeably improvedSo much so that nowadays we can enjoy a good experience even in games that are not too demanding with this type of graphic solutions.
The first thing that determines the performance of an integrated GPU is, of course, its architecture and internal configuration. For example, a next-generation solution like the Radeon Vega 11, which has 704 shaders, will perform better than a Radeon Vega 3, equipped with 192 shaders. The same applies to Intel’s integrated solutions, the HD family and, within it, the Iris Pro line, where a model with more execution units will perform better than another with fewer execution units, provided they both use the same architecture. .
It is important to be clear about this because it gives us the basis we need to understand the peculiarities of an integrated GPU. As with dedicated GPUs the graphics core configuration determines, together with the architecture and the working frequencies, the performance that it will be able to offer. Models with higher shader counts will be more powerful as long as the working frequencies and architecture match.
However, unlike with dedicated GPUs, integrated solutions they lack their own memory subsystem. For example, a GTX 1050 has a 128-bit bus and mounts 2GB of 7GHz GDDR5 memory, which you can access directly. This memory is exclusive to the dedicated GPU.
When we use a dedicated GPU the situation changes completely. These types of solutions do not have that memory subsystem, and do not have a fixed amount of integrated graphics memory, forcing them to draw on the system’s RAM and borrow a few gigabytes. This has very important consequences:
Bandwidth is determined by system memory bus and RAM speed.
A part of the RAM is dedicated to exclusive use of the integrated GPU, which can cause some applications to not work properly even though, in theory, we have enough RAM. For example, if we have 8 GB of RAM and run a game that requires that number, but we use an integrated GPU that consumes 2 GB of RAM as graphic memory. In that case only we would have 6 GB of RAM free, which would be used as system memory, that is, we would be below those 8 GB.
RAM memory it’s slower than conventional graphics memory, which can significantly limit the performance of the integrated GPU. For example, even with high-performance DDR4 modules we would only go as far as 3,200 MHz without having to do a huge expense, while GDDR5 memory runs smoothly at 7 or 8 GHz.
All this means that the RAM memory configuration that we have in our equipment can considerably affect the real performance of the integrated GPU. To this are added other keys, such as the power mode that we use, the drivers and other things that we are going to see below, and that, applied together, will help you improve the performance of your integrated GPU.
Please note that if we start from a very weak integrated GPU we should not expect any miracle, but in some cases the difference can be very large, especially in AMD’s Radeon Vega models and in the Intel Iris Pro series.
1.-Integrated GPU and RAM memory
We already know why RAM configuration is so important in maximizing the performance of an integrated GPU. So that it can develop its full potential and prevent it from being weighed down at any time we must have:
Dual channel RAM: Using two dual-channel RAM modules doubles the data bus as it goes from 64-bit to 128-bit. The difference is huge, so much so that a Radeon Vega 11 can double its gaming performance.
High working frequency: Having faster memory increases bandwidth and improves performance, although after a certain threshold the difference in performance and the price increase is no longer compensated. We should move into the 2,666 MHz-3,200 MHz band.
More memory than you need: It is important because, as we have said, a part of the RAM will be used as graphic memory when running intensive applications. If we are going to play with an integrated GPU, having 8 GB will not be enough for an optimal experience for what we have explained in the previous example. We should have at least 12 GB of memory.
In Intel integrated GPUs the impact of the memory configuration is also very large, but being less powerful solutions it is not always enough to go from a “non-playable” level to a playable one.
2.-The importance of drivers
I know it is a classic, but its impact has proven to be so great that we cannot help but remember it. Both Intel and AMD regularly release driver updates that improve the performance of its main GPUs with specific games, and its integrated solutions are not usually left out of those improvements.
The optimizations and adjustments that are introduced at the driver level can be complemented by other minor adjustments that allow, in some cases, “Scratch” a few FPS at no cost. Obvious to say that when we move in the 20-30 FPS band a few frames per second can make the difference between a slideshow and an acceptable experience.
In the case of AMD, the improvements that its GCN architecture has received through the successive drivers that have arrived in recent months have been so great that not installing them is equivalent to give up a free performance boost. Never underestimate the improvements that a new driver can bring.
Intel has also greatly improved its drivers, and has shaped some interesting optimizations in specific titlesSo if you have not updated your integrated GPU drivers for a long time, you know, you have an appointment with the manufacturer’s official website to download the latest compatible version.
3.-The energy mode used
It has a much bigger impact than we can think with the naked eye. As many of our readers will already know the power mode used in Windows 10 it can vary between three type models: low consumption, balanced and high performance.
When we use an integrated GPU with low power mode activated their working frequencies can be reduced to a large extent, which equates to a significant performance loss. To prevent this from happening we just have to select the high performance plan.
This plan will affect energy consumption and can drastically reduce autonomy in portable computers, but the integrated GPU, and the rest of the PC components, will not be constrained by energy limitations and they will be able to work at full power.
If you have doubts about how to enable the maximum performance power plan in Windows 10 you just have to follow the steps we gave you in this tutorial.
4.-Overclock within safe levels
Another way to get a little more performance is through overclocking. In the case of the integrated GPU the limitations are usually more marked than in the case of the dedicated GPUs, but This does not mean that it is impossible, and the upgrade may be enough, in some cases, to make the difference between playable and non-playable.
Well, we have two possibilities that we should use together, overclock the integrated GPU and also the RAM, whenever the latter is possible. I understand that it is something that can intimidate less expert users, that’s why I have left this tip for last, but the truth is that thanks to the Ryzen Master tool it is easier and safer than ever.
This tool allows us adjust in a few clicks the speed of the integrated GPU and its voltage. Thanks to its intuitive interface we can gradually increase the working frequency and do stability and temperature tests to see when we have reached the ceiling of our equipment.
For example, we can make 25 MHz increments and test stability and temperatures of work. When we reach a level where we begin to notice the slightest problem, we reduce the overclock a little and that’s it.
In the case of Intel integrated GPUs we can also overclock without too much effort, thanks to the tool Intel Extreme Tuning. The pattern to follow should be the same as we have indicated, although if you screw up and enter a too high and unstable value, do not worry, the equipment will restart and everything will return to the default configuration.
It is a complex question, since it implies know the own limitations of our integrated GPU and drivers from each manufacturer, but it can help us improve performance at the cost of certain sacrifices in graphics quality.
Both Intel and AMD integrate control panels of the drivers for its integrated GPUs sections where we can customize settings of 3D applications. NVIDIA also has such a control panel, but the company has not marketed integrated GPUs for a long time, and so we are omitting it in this article.
In the case of Intel we can find, for example, interesting options, such as the “optimized” mode, which activates improvements in specific games to boost performance, and generic adjustments that allow us to select between four different modes: performance, balanced, image quality and “custom”.
AMD also offers multiple options in control panel of your Radeon Software, where we can find various adjustments related to the quality of edge smoothing, anisotropic filtering and more. Reducing them will affect graphic quality, but will improve performance.