Graphics are very important in today’s computer age. However, despite how common it is to use computer graphics for GUIs, for video games, etc., the truth is that it is becoming more and more complex, especially when iGPUs, dGPUs, APUs and now tGPUs appeared . But don’t worry if all this confuses you, in this article you will learn what all this is and their differences…
Index of contents
- What is a GPU?
- What is an iGPU?
- The advantages of iGPU
- The disadvantages of iGPU
- iGPU and SoC
- What is tGPU or Tile-GPU
- What is a dGPU?
- What is an APU?
- Some common examples of APU
- Conclusion Which GPU design is right for me?
You may be interested in knowing:
- What are the best graphics cards on the market.
What is a GPU?
A GPU (Graphics Processing Unit) , or graphics processing unit, is a type of modern chip that is capable of 3D rendering for video games and other applications. For this, the data and instructions that arrive at this chip are processed to render, that is, to create the graphics.
In the past, graphics work was also overloaded on the CPU , but with the advancement of graphics quality, this put a huge overhead on this unit. That is why the first graphic accelerators emerged, to make way for what we currently know as graphics cards or GPUs.
A GPU is capable of performing a multitude of mathematical calculations per second to calculate the correct values for millions of pixels dozens or hundreds of times each second. All computing devices, from PCs to smartphones, have a GPU.
As a curiosity to say that the only computers that work without a GPU are usually some headless servers that operate remotely and only use CPUs. They don’t even need a graphical environment for their management, everything is done through the command console. But this is a different story…
What is an iGPU?
Many laptops, and also some desktop and AIO PCs, come with an integrated GPU, or iGPU . This type of unit is in the same packaging as the CPU, albeit on a separate chip as can be seen in the image above. These graphics fulfill their function, generating the necessary graphics, although they are not as powerful as dGPUs as we will see later.
In addition, another thing must be taken into account, and that is that when there is an iGPU, a dGPU could also coexist on the same computer . They are not incompatible, but can serve as a complement. But, there are several things to keep in mind:
- Many systems with iGPU, even if they have dGPU, the graphics have to finally go through the iGPU, which is the one that is connected to the video outputthat goes to the monitor, although some teams have started to implement solutions for this , since this can compromise performance.
- The two iGPUs will not be rendering the graphics at the same time, but one or the other can work, that is, they cannot add their capabilities to each other. Therefore, from the operating system, you can choose one or the other as appropriate.
- The iGPU does not have VRAM, instead it uses RAM for this purpose.
- You should not confuse an iGPU with the integrated GPU in an APU, since they are not exactly the same as we will see later.
The advantages of iGPU
As always, all technology has its advantages , and the iGPU was not going to be less. Some of the notable points are:
- Lower price.
- Since it is integrated into the CPU package, it does not require additional space, which is positive for ultrabooks.
- You also don’t need a separate cooling system, which saves space. Both CPU and iGPU will fit in the same TDP of the package.
- They usually have a lower consumption than dGPUs. They have good energy efficiency, so they will save battery, to prolong autonomy.
- If a dGPU exists alongside the iGPU, it can dynamically switch between them depending on the graphics power needed. For example, reduce consumption by using the iGPU when you are working with office automation or browsing, and use the dGPU when you need performance for gaming, etc.
The disadvantages of iGPU
Of course, not all are advantages in terms of iGPU, it also has its disadvantages compared to dedicated graphics. For example:
- The performance of the iGPU is quite limited, and for intense gaming or AAA titles it is not suitable. Neither for heavy rendering tasks.
- By not having a dedicated VRAM, it will make use of RAM, so CPU and iGPU will share the same memory, taking space from it for other applications.
- When you have a dGPU, it could even act as a bottleneck for it, since sometimes the graphics would have to finally go through the iGPU which is the one that is actually connected to the screen (although the heavy lifting is done on the graphics chip). the dGPU).
iGPU and SoC
Another important thing to consider is the iGPUs that are integrated into the SoC (System-on-a-Chip) . These SoCs not only integrate the CPU and GPU, they also include memory, controllers, and other system units. All on a single chip. Therefore, it is not the same as what we know as iGPU, since it is not on another chip within the same package, but rather everything is a single chip.
Being all on a monolithic chip, the integrated GPU can improve the interconnections with the rest of the units that are included, although generally the RAM memory is usually included outside the SoC, and in this case it will also be shared between the iGPU and CPU , and even other units such as a DSP, NPU, etc.
Currently, SoCs are very popular , since they are used for mobile devices, as well as being included in other Smart devices, as well as in the new Apple Macs, with SoCs M1, M2, etc., Qualcomm Snapdragon, Mediatek Helio, HiSilicon , Samsung Exynos, etc.
What is tGPU or Tile-GPU
Within the iGPU section I did not want to forget a term that Intel has recently introduced. And it is that now it also calls its new iGPUs as tGPU or Tile-GPU . But this should not confuse you, they are the same, only in this case a tile-type architecture is used, being able to add tiles (either through a chiplet or MCM design or through 3D packaging) as you want to scale the performance.
What is a dGPU?
A discrete GPU, dedicated GPU, or dGPU , is a graphics processing unit that is not part of the CPU, that is, it is implemented in separate packaging or on a separate PCB. Some dGPUs are simply soldered onto the motherboard itself, others are mounted on modules like MXMs . In either case, it will also be accompanied by VRAM memory chips. Therefore, you will not need to make use of the shared RAM.
Unlike a graphics card , which includes a more complex PCB and includes the cooling system as well, a dGPU is slightly different. Also, as I said before, a dGPU can also coexist with an iGPU.
The dGPUs that come built in to the motherboard are not easily replaceable , as it would involve desoldering the BGA from the GPU and replacing it with another GPU. But this is expensive and special equipment and laboratory are needed. On the other hand, those found in MXM modules for laptops can be easily replaced as you would do with a conventional graphics card from a desktop PC.
Another consideration is that dGPUs don’t typically need extra power from the PSU like high-end graphics cards do. In the case of the dGPU, when they are integrated into the motherboard, they get their power through it.
An IGP or Integrated Graphics Processor is a graphics chip built into the motherboard itself. This was very common in the past, on some lower-middle-range motherboards for users who didn’t need an external graphics card. These chips could be disabled from the BIOS in some cases. There were also some graphics integrated into the motherboard chipset itself, and when it broke, there was no choice but to replace the entire motherboard…
I also would not like to forget that the dGPU is not exactly the same as the GPU that comes with graphics cards. Between NVIDIA, AMD and Intel’s GPU models for the desktop and their chips for mobile (Mobile) there are slight differences, especially in the TGP which is lower, and also the performance is often somewhat lower. Keep in mind that they have to make the battery last…
What is an APU?
Just like Intel embraced iGPUs in its processors, AMD did something different by using the term APU (Accelerated Processing Unit) . In this case, the GPU can also be considered integrated, but unlike a conventional iGPU, in this case it is integrated on-chip, that is, within the same chip as the CPU. Therefore, everything is a monolithic chip.
In other words, the APU could be considered something intermediate between the iGPU and a SoC , since it integrates the GPU on the same chip as the CPU as the SoC does, but it would not be an SoC, since it lacks other additional units that if they are typical in SoCs.
The first APUs arrived in 2011. This technology was originally known by AMD as Fusion, and was intended to revolutionize the market with a chip with heterogeneous architecture , although it ultimately did not turn out as planned. Of course, currently there are still APUs to compete with the Intel line that includes iGPUs.
Being in the same die, it has benefits in terms of performance when the CPU must tell what to do in the GPU. However, as with the iGPU, in this case there will also be no dedicated VRAM, instead the CPU and GPU will have to access and share the same RAM .
Some advantages of the APU over the iGPU are:
- Better performance in communication between CPU-GPU.
- Lower consumption for communication than if they were on separate chips.
- Resources can be shared between the two easily.
- Eliminates the need to install a dGPU or dedicated graphics card, which is cheaper.
- Like the iGPU, it also supports that you can include a dGPU in the same computer and use one or the other, although with the same problems that we saw with the iGPU.
However, as usual, APUs are not without their drawbacks :
- Its graphics power is not as powerful as that of a dGPU.
- They add complexity to the monolithic chip and make it more expensive.
Some common examples of APU
AMD has used its APUs for many years, and continues to do so for notebook chips. Even APUs have made it to the Opteron, Athlon, A-Series, Ryzen, Threadripper, and EPYC series. Currently, consoles like the Sony PlayStation or Microsoft Xbox, and even Valve’s Steam Deck make use of these APUs.
Conclusion Which GPU design is right for me?
| TYPE | SIMPLE EXPLANATION |
| graphics card | We refer to a PCB board that includes the GPU chip, as well as VRAM memory chips, BIOS, and all the necessary auxiliary elements, such as ports, power socket, cooling system, etc. |
| GPUs | It refers only to the graphics processing unit, that is, the central chip. It manipulates and alters memory to create images in a frame buffer that can then be output as video to a screen. It requires other components to work. |
| iGPU | It is an integrated GPU in the same packaging as the CPU. The iGPU works the same way as a GPU, but all the cooling, ports, memory, etc. are taken from other components. |
| APU | You could say it’s like an AMD iGPU, but it’s on the same monolithic chip as the CPU. AMD APUs support Heterogeneous System Architecture (HSA), which allows the CPU and GPU to be on the same bus, thereby reducing communication latency. |
In conclusion, say that the following scenarios could occur where we recommend one type or another:
- If graphics performance isn’t important to you, since you’re not going to be gaming intensively or doing other heavy graphics tasks, then an iGPU or APU is best.
- If performance is very important to you, because you are going to be rendering or playing AAA titles, then a dGPU for laptops or a graphics card for desktop PCs is ideal.
- When consumption is critical, an APU or SoC is better than any other unit.
- If power consumption isn’t critical, you could use any kind of graphics unit.
- When power scaling is important, a graphics card with multi-GPU setups in mind is best.