{"id":261260,"date":"2025-10-06T12:00:29","date_gmt":"2025-10-06T10:00:29","guid":{"rendered":"https:\/\/www.hwcooling.net\/?p=261260\/"},"modified":"2026-08-01T02:16:13","modified_gmt":"2026-08-01T00:16:13","slug":"how-a-graphics-card-works-dedicated-to-secondary-schools","status":"publish","type":"post","link":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/","title":{"rendered":"How a graphics card works. Intended for secondary schools"},"content":{"rendered":"<p><!--nextpage-->A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated in the processor (or earlier, in the chipset). Here, however, we will focus on discrete graphics cards.<!--more--><\/p>\n<p>The term \u201cgraphics card\u201d comes from the era when it was merely an add-in card providing video output to a monitor, including displaying raster graphics on its screen. In the 1990s, graphics cards (also called video cards) were combined with previously separate 3D accelerators, which initially existed as standalone components. These accelerators contained chips designed for demanding game graphics calculations and professional 3D visualization. Around the turn of the millennium, the concept of the GPU (Graphics Processing Unit) with expanded functions became widespread. In a narrower sense, the term GPU may also refer specifically to the chip that forms the core of the graphics card.<\/p>\n<p>Modern discrete GPUs almost always come in the form of an expansion card for a PCI Express \u00d716 slot. Historically, other interfaces such as ISA, VESA Local Bus, PCI, and AGP were used. You may also encounter GPUs incorporated directly on the motherboard, which is common in laptops. Especially in the past, swappable MXM modules were used\u2014they served a similar role in laptops as PCI Express cards in desktops.<\/p>\n<figure id=\"attachment_40117\" aria-describedby=\"caption-attachment-40117\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-08.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"noborder wp-image-40117 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-08-1024x600.jpg\" alt=\"\" width=\"640\" height=\"375\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-08-1024x600.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-08-300x176.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-08-768x450.jpg 768w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-40117\" class=\"wp-caption-text\">Classic desktop graphics card. Rear view of the PCB<\/figcaption><\/figure>\n<p>A graphics card can also be external. In this case, the component is housed in an enclosure connected via USB4 or Thunderbolt. However, the dominant standard remains the PCI Express card installed inside a desktop PC, i.e., a discrete graphics card in the true sense of the word.<\/p>\n<figure id=\"attachment_260744\" aria-describedby=\"caption-attachment-260744\" style=\"width: 670px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Externi-grafika-OneXGPU-zalozena-na-Radeonu-RX-7600M-XT.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260744 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Externi-grafika-OneXGPU-zalozena-na-Radeonu-RX-7600M-XT.jpg\" alt=\"\" width=\"670\" height=\"377\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Externi-grafika-OneXGPU-zalozena-na-Radeonu-RX-7600M-XT.jpg 670w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Externi-grafika-OneXGPU-zalozena-na-Radeonu-RX-7600M-XT-300x169.jpg 300w\" sizes=\"auto, (max-width: 670px) 100vw, 670px\" \/><\/a><figcaption id=\"caption-attachment-260744\" class=\"wp-caption-text\">External OneXGPU graphics unit based on Radeon RX 7600M XT (source: OneXGPU)<\/figcaption><\/figure>\n<p>Cards for PCI Express slots come in different sizes. The most powerful gaming models require robust cooling, which takes up more than one slot in a PC case. Single-slot graphics cards do exist but are rare today, as this design constraint makes cooling considerably more difficult.<\/p>\n<figure id=\"attachment_260745\" aria-describedby=\"caption-attachment-260745\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260745 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem-1024x576.jpg\" alt=\"\" width=\"640\" height=\"360\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem-1024x576.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem-300x169.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem-768x432.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem-1536x864.jpg 1536w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Galax-GeForce-RTX-4060-Ti-Max-s-jednoslotovym-chladicem.jpg 1600w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-260745\" class=\"wp-caption-text\">Galax GeForce RTX 4060 Ti Max with single-slot cooler (source: Galax)<\/figcaption><\/figure>\n<p>Most graphics cards are of \u201cfull-height\u201d design, with a standard-sized PCB. However, there are also so-called low-profile graphics cards. These have a reduced-size PCB with a maximum height of 6.44 cm, which the cooler also does not exceed. They may still take up more than one slot in thickness, though. Such cards are intended for cases that provide less vertical clearance for expansion cards. In these cases, a shorter expansion slot bracket is used. Low-profile cards can also be installed in standard PC cases, but a full-height bracket must then be used. Many cards include both bracket types in the box, allowing the original to be swapped after removing the screws.<\/p>\n<figure id=\"attachment_244641\" aria-describedby=\"caption-attachment-244641\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-scaled.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-244641 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-1024x746.jpg\" alt=\"\" width=\"640\" height=\"466\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-1024x746.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-300x218.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-768x559.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-1536x1119.jpg 1536w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/05\/DSC_0431-2048x1492.jpg 2048w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-244641\" class=\"wp-caption-text\">Low-profile graphics card: GeForce RTX 5060 OC Low Profile 8G. This card has the standard-sized bracket fitted which showcases the difference from a card with a full-sized PCB<\/figcaption><\/figure>\n<p>In recent years, the size of high-performance graphics cards has grown, which places greater demands on PC cases. Older case models may not allow the installation of exceptionally long graphics cards that are often manufactured today (over 30 cm in length is not uncommon). When building a PC, it\u2019s important to choose components with compatibility in mind to ensure the graphics card fits inside the case. It\u2019s also advisable to leave some clearance\u2014for example, aditional space is needed for routing power cables.<\/p>\n<p>Besides size, weight is also a concern. Powerful graphics cards are heavy enough that they often sag when mounted in the case. This can lead to damage to the card or the motherboard, and it\u2019s wise to prevent this with various brackets or supports, which are often included with the graphics card or even with the <a href=\"https:\/\/www.hwcooling.net\/en\/how-a-pc-case-works-dedicated-to-secondary-schools\/\">PC case<\/a>.<\/p>\n<p><em><strong>Please note:<\/strong> The article continues in the following chapters.<\/em><\/p>\n<script async src=\"\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js\"><\/script>\n<!-- responsive -->\n<ins class=\"adsbygoogle\"\n     style=\"display:block;background-color:transparent\"\n     data-ad-client=\"ca-pub-8150419924824893\"\n     data-ad-slot=\"6522017574\"\n     data-ad-format=\"auto\"><\/ins>\n<script>\n(adsbygoogle = window.adsbygoogle || []).push({});\n<\/script>\n<p><!--nextpage-->A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated in the processor (or earlier, in the chipset). Here, however, we will focus on discrete graphics cards.<!--more--><\/p>\n<h4>GPU<\/h4>\n<p>The most important component of a graphics card is the GPU (Graphics Processing Unit). It handles all of the card\u2019s compute and graphics processing functions, much like the CPU (processor) does for the rest of the computer. We will cover its role in more detail in the next chapter. GPUs are manufactured in BGA (Ball Grid Array) packages. That means the substrate of the chip has solder balls on the underside, which permanently attach it to the PCB that carries and interconnects all the graphics card components. They are therefore not mounted in any kind of socket, and the GPU cannot simply be replaced or upgraded on the card.<\/p>\n<p>Modern GPUs are among the most advanced and complex chips produced. Their size can reach the maximum silicon die area currently achievable with modern manufacturing lines\u2014around 600\u2013800 mm<sup>2<\/sup> (the limit is often referred to as the reticle limit).<\/p>\n<figure id=\"attachment_260741\" aria-describedby=\"caption-attachment-260741\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-scaled.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260741 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-1024x698.jpg\" alt=\"\" width=\"640\" height=\"436\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-1024x698.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-300x204.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-768x523.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-1536x1047.jpg 1536w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/GPU-Nvidia-GB203-2048x1396.jpg 2048w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-260741\" class=\"wp-caption-text\">GPU Nvidia GB203. The eight gray packages around the GPU are GDDR7 memory modules<\/figcaption><\/figure>\n<p><strong>Memory or VRAM<\/strong><\/p>\n<p>The GPU requires dedicated memory for its operation (integrated GPUs instead use part of the system\u2019s main memory). It&#8217;s often referred to as VRAM (video RAM), or graphics memory. Graphics cards use a special type of memory called GDDR\u2014currently GDDR7, alongside the previous GDDR6 generation. These memory chips operate at higher transfer rates (per pin of the memory bus) than standard system RAM technology such as DDR5 chips. GPUs demand very high memory bandwidth to perform well\u2014this allows them to read large amounts of texture and object data per second.<\/p>\n<p>For the same reason, the overall width of the memory bus (often referred to as the memory controller width) is also higher than the width of the system RAM. While system RAM typically uses a 128-bit bus, GPUs commonly use 192-bit, 256-bit, or even 384-bit interfaces. The most powerful GPUs even use 512-bit buses today. Memory bandwidth is the product of bus width and transfer rate per bit (sometimes also referred to simply as \u201ceffective clock speed\u201d). GDDR-type memory packages are usually placed closely around the GPU\u2014short length of the traces is critical for achieving the high data transfer rates required between GPU and graphics memory. The PCB traces connecting GPU and GDDR-type memory must meet strict signal integrity requirements.<\/p>\n<p>In rare cases, GPUs may use high-tech HBM (High Bandwidth Memory), such as HBM2, HBM3, or HBM3E. These are multi-layers memory packages that are integrated onto the GPU package using advanced 2.5D packaging, which is costly. This is why, with few exceptions in the past, HBM has not been used in gaming graphics cards. Instead, it is found in compute GPUs designed for datacenters and AI acceleration.<\/p>\n<figure id=\"attachment_201043\" aria-describedby=\"caption-attachment-201043\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-201043 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz-1024x683.jpg\" alt=\"\" width=\"640\" height=\"427\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz-1024x683.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz-253x168.jpg 253w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz-300x200.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz-768x512.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2024\/08\/AMD-Vega-10-upoutavka-foto-Fritzchens-Fritz.jpg 1920w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-201043\" class=\"wp-caption-text\">AMD Vega 10 GPU with HBM2 memory. The two packages to the left of the large GPU die are HBM2 modules. An interposer\u2014a large silicon chip under both the GPU and memory\u2014is required for interconnection (photo: <a href=\"https:\/\/www.flickr.com\/photos\/130561288@N04\/39427639455\/in\/photostream\/\">Fritzchens Fritz<\/a>)<\/figcaption><\/figure>\n<p>On the graphics card, there is also a non-volatile Flash memory that stores the card\u2019s BIOS. This functions similarly to the BIOS on a motherboard: it contains code to initialize the graphics card during system boot as well as other firmware-level functions. Permanent settings such as voltages, clock speeds, or power limits can also be stored there. Some graphics cards even feature two BIOS chips, selectable via a switch on the card. Typically, one BIOS is tuned for higher cooling performance (for lower temperatures during load) while the other favors quieter operation through lower fan speeds.<\/p>\n<p>The GPU is connected to the computer\u2019s CPU through the PCI Express bus, and the CPU controls the GPU like other peripherals. Commands and data are sent to the GPU over PCIe, and from the GPU the video outputs are provided to connect a monitor or other display device.<\/p>\n<h4>Voltage regulators<\/h4>\n<p>The GPU and its memory have relatively high power demands, so a graphics card requires dedicated power delivery. A significant portion of the PCB is therefore occupied by voltage regulators (sometimes referred to as the power delivery subsystem or VRMs\u2014Voltage Regulator Modules), which operate on the principle of a switching power supply (switching regulator). The VRM consists of MOSFETs for switching, as well as inductors and capacitors for smoothing voltage. These components step down the 12 V supply from the PSU to lower voltages (typically around 1 V) required by the GPU. Modern cards need several separate voltage rails for the GPU, memory, and possibly some sub-blocks of the chip.<\/p>\n<figure id=\"attachment_260742\" aria-describedby=\"caption-attachment-260742\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-scaled.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260742 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-1024x523.jpg\" alt=\"\" width=\"640\" height=\"327\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-1024x523.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-300x153.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-768x393.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-1536x785.jpg 1536w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Napajeci-kaskada-na-graficke-karte-GeForce-RTX-5080-Suprim-X-od-MSI-2048x1047.jpg 2048w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-260742\" class=\"wp-caption-text\">Power delivery on the MSI GeForce RTX 5080 Suprim SOC graphics card<\/figcaption><\/figure>\n<p>The VRM receives the input 12 V from the PSU via cables. On newer Nvidia cards, this is done through 12V-2&#215;6 connectors with 12 power and 4 signal pins, while AMD, Intel, and older Nvidia models typically use 8-pin PCI Express power connectors. These cables deliver 12 V to the card. Low-power graphics cards may instead draw all their power directly from the PCI Express \u00d716 slot, which per specification can supply up to 75 W (combined across the 3.3 V, 5 V, and 12 V rails).<\/p>\n<figure id=\"attachment_219268\" aria-describedby=\"caption-attachment-219268\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-219268 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-1024x680.jpg\" alt=\"\" width=\"640\" height=\"425\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-1024x680.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-253x168.jpg 253w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-300x199.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-768x510.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-600x398.jpg 600w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/124pinov\u00fd-konektor-12V-2x6-u-GeForce-RTX-5090-Founders-Edition-upoutavka-872x579.jpg 872w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-219268\" class=\"wp-caption-text\">12+4-pin 12V-2&#215;6 connector on the Nvidia GeForce RTX 5090 Founders Edition<\/figcaption><\/figure>\n<figure id=\"attachment_114763\" aria-describedby=\"caption-attachment-114763\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-114763 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07-1024x683.jpg\" alt=\"\" width=\"640\" height=\"427\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07-1024x683.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07-253x168.jpg 253w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07-300x200.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/msi-rtx-3050-ventus-2x-8g-oc_07-768x512.jpg 768w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-114763\" class=\"wp-caption-text\">8-pin PCI Express auxiliary power connector on the MSI GeForce RTX 3050 Ventus X2 OC<\/figcaption><\/figure>\n<p>Because of the high power consumption, cooling is an integral part of every graphics card. The cooler is factory-mounted to the PCB so that its base makes direct contact with the GPU die, memory chips, and usually the power MOSFETs. Heat is conducted from the base into the heatsink fins, which are cooled by fans. The fans are connected to the PCB, and the GPU typically has the ability to control their speed. The cooler usually has a shroud, which not only serves a decorative purpose but also provides mounting points for brackets or supports to stabilize the card in its slot.<\/p>\n<p>Some graphics cards use a liquid cooler instead of an air cooler. In that case, a cold plate transfers heat from the GPU to liquid circulating within a cooling loop, which the graphics card must then be connected to.<\/p>\n<p><em><strong>Please note:<\/strong> The article continues in the following chapters.<\/em><\/p>\n<script async src=\"\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js\"><\/script>\n<!-- responsive -->\n<ins class=\"adsbygoogle\"\n     style=\"display:block;background-color:transparent\"\n     data-ad-client=\"ca-pub-8150419924824893\"\n     data-ad-slot=\"6522017574\"\n     data-ad-format=\"auto\"><\/ins>\n<script>\n(adsbygoogle = window.adsbygoogle || []).push({});\n<\/script>\n<p><!--nextpage-->A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated in the processor (or earlier, in the chipset). Here, however, we will focus on discrete graphics cards.<!--more--><\/p>\n<h3 class=\"western\">Compute Units<\/h3>\n<p>A GPU is a processor that operates on the principle of highly parallel instruction processing. Therefore, it consists of multiple replicated compute units. In Nvidia GPUs, these compute units are designated as SMs (Streaming Multiprocessor), in AMD GPUs they are labeled as CUs (Compute Unit) or WGPs (Workgroup Processor), and in Intel GPUs as Xe Cores.<\/p>\n<p>These compute units are typically further grouped into higher-level structures, whose hierarchy differs depending on the architecture used. Some GPU compute resources are shared within a certain block level. For example, each SM or CU block may have its own registers and some caches, but texturing or rasterization units may be incorporated in a higher level block that combines several basic compute units. Examples of these bigger blocks are TPCs and GPCs in Nvidia&#8217;s GPUs and Shader Engines and Shader Arrays in AMD GPUs.<\/p>\n<h3 class=\"western\">Shaders (\u201eCUDA cores\u201c)<\/h3>\n<p>In modern gaming and compute tasks, the units for general-purpose compute operations, located in blocks in quantities of, for example, 64 or 128, play perhaps the biggest role. In simplified maner they are called shaders, stream processors, shader ALUs, or CUDA cores (in Nvidia GPUs). The code you run on the GPU, so-called shader programs, is processed on these units. These were originally simple and had limited use in games (the term shader derives from their use for shading), but over time they gained increasingly universal capabilities and now represent a compute resource usable for both general purpose non-graphics computations. In graphical use, shaders can work with image pixels, but also with the geometry of 3D models, which is also calculated using shaders today.<\/p>\n<h3 class=\"western\">Texture Mapping Units (TMU)<\/h3>\n<p>Texturing units or TMUs are an evolutionarily older feature, present almost from the first 3D accelerators. Their task is to process operations necessary for mapping textures onto models in a 3D scene. A texture is a two-dimensional raster image intended to represent the &#8220;surface&#8221; of an object. But for the object to be wrapped with this surface, points from the texture raster must be variously adjusted to be displayed in the correct perspective, rotation, and deformation. It is also necessary to scale them to the correct size. These operations are referred to as filtering (which can use various methods\u2014bilinear, trilinear, anisotropic) and require a certain amount of performance.<\/p>\n<p>The task of the texture mapping units is precisely to process filtering operations without consuming the performance of the GPU&#8217;s general-purpose units (shaders). They can also work directly with various compressed texture formats that save memory space without needing to decompress them first.<\/p>\n<p>In earlier times, a value called texture fill rate in (mega or giga) texels per second was stated in GPU specifications. This figure indicated how many outputs (texels, which means pixels with applied texture) the TMUs in a given GPU could handle overall. Today, this parameter is no longer in the forefront of interest, as the performance of texturing units is typically not a performance bottleneck.<\/p>\n<figure id=\"attachment_260746\" aria-describedby=\"caption-attachment-260746\" style=\"width: 2048px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky.png\"><img loading=\"lazy\" decoding=\"async\" class=\"noborder wp-image-260746 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky.png\" alt=\"\" width=\"2048\" height=\"687\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky.png 2048w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky-300x101.png 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky-1024x344.png 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky-768x258.png 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-skladby-GPU-Nvidia-GB202-generace-Blackwell-ukazuje-ruzne-funkcni-bloky-1536x515.png 1536w\" sizes=\"auto, (max-width: 2048px) 100vw, 2048px\" \/><\/a><figcaption id=\"caption-attachment-260746\" class=\"wp-caption-text\">Block diagram of the Nvidia GB202 Blackwell generation GPU shows various functional blocks<\/figcaption><\/figure>\n<h3 class=\"western\">Rasterization Units (ROP)<\/h3>\n<p>ROP units (Render Output Unit, or Raster Operations Pipeline) are another of the basic components of GPUs, present practically from the beginning. The task of these units is to merge all outputs generated during the rendering of the 3D image by other GPU units and create from them the final pixels forming the output two-dimensional frame of the game or 3D scene. ROP units can apply anti-aliasing methods such as MSAA (where multiple input samples are used to create one output frame pixel to suppress so-called aliasing).<\/p>\n<p>A value called pixel fill rate in (mega or giga) pixels per second can be specified for GPUs, which was prominently advertised in the past. This figure indicated how many image pixels (pixels) of the output frames the ROP units in the GPU could render overall. Similar to texture fill rate, this parameter is usually not at the center of attention today.<\/p>\n<h3 class=\"western\">Memory controller<\/h3>\n<p>The GPU has a memory subsystem with its own memory space. This memory is controlled by a special controller, whose function is very important for GPU performance. Very high bandwidth is required for graphics memory, so a wider bus is usually chosen, although cheaper GPUs have 128-bit or even only 64-bit memory widths (the width in bits is often referred to as the &#8220;memory bus&#8221; width). In reality, multiple controllers are used in parallel; for GDDR-type memories, a controller has a width of 32 bits, so for example, a GPU with a 256-bit bus has eight of them, and a 384-bit bus means twelve controllers. Besides the digital logic of the controllers, the GPU also contains PHY blocks, i.e., circuits generating the actual electrical signal for communication with the memory.<\/p>\n<h3 class=\"western\">Cache<\/h3>\n<p>Similar to a CPU, the GPU uses a more complex hierarchy of the memory subsystem with cache memories at several levels\u2014L0, L1, L2, and sometimes also L3 (Infinity Cache in AMD GPUs is basically an L3 cache). The task of cache memories is to hold the most frequently used data, which is available from them with lower latency and higher data bandwidth than if the GPU had to load it from the main GPU memory.<\/p>\n<p>Serving part of the data requests from the cache memories means that the volume of requests that have to go to the main graphics memory is reduced. Thanks to this, the whole memory subsystem can handle more read operations. Larger caches thus ultimately allow to achieve a higher overall performance level, for a given bandwidth of the main memory. Memory bandwidth has long been one of the most significant limiting factors preventing increases in GPU performance.<\/p>\n<div class=\"n4sEPd\">\n<div class=\"FFpbKc\">\n<div class=\"xMmqsf\">\n<div class=\"nidPne\">\n<div id=\"ow16\" data-show-delay-ms=\"250\" data-append-to-body=\"true\" data-propagate-tooltip-mouseover-events=\"true\" data-anchor-corner=\"bottom-left\" data-enable-skip-handler=\"true\" data-popup-corner=\"top-left\">\n<div class=\"dig2sb\">\n<div class=\"scLCMd\">\n<div id=\"itamenu\"><span class=\"jCAhz\"><span class=\"ryNqvb\">Cache memory can also have a positive impact on energy efficiency.<\/span><\/span> <span class=\"jCAhz ChMk0b\"><span class=\"ryNqvb\">Retrieving data from cache memory usually consumes less energy than reading from main memory, which is why large and efficient cache can reduce GPU power consumption.<\/span><\/span><\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<h3 class=\"western\">AI accelerators<\/h3>\n<p>Over the last decade, running artificial intelligence applications has become one of the most important, if not the most important, tasks for GPUs, utilizing their high computational power. For this reason, special units that perform matrix multiplication operations for AI needs were added to GPUs. The performance when using these units is higher than what is available on general-purpose compute units (shaders). However, these special units support a smaller set of computate operations and often also work only with lower-precision data types, which allows them to achieve higher performance. While FP32 (32-bit floating-point value) is the standard precision for GPU compute operations, 16-bit, 8-bit, and even 6-bit or 4-bit values (INT8, FP8, FP6, FP4) are used for artificial intelligence.<\/p>\n<figure id=\"attachment_219201\" aria-describedby=\"caption-attachment-219201\" style=\"width: 1833px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4.png\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-219201 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4.png\" alt=\"\" width=\"1833\" height=\"1031\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4.png 1833w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4-300x169.png 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4-768x432.png 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/01\/FP4-1024x576.png 1024w\" sizes=\"auto, (max-width: 1833px) 100vw, 1833px\" \/><\/a><figcaption id=\"caption-attachment-219201\" class=\"wp-caption-text\">5th generation Tensor Cores in Nvidia GeForce RTX 5000 graphics cards support matrix computations, now also with the FP4 data type<\/figcaption><\/figure>\n<p>Nvidia GPU architectures contain AI accelerators labeled as Tensor Cores. Intel GPUs have similar dedicated units called XMX. Compute GPUs from AMD use Matrix Cores, but Radeon gaming GPUs do not use this approach yet and have AI acceleration built into the general-purpose compute units.<\/p>\n<h3>Ray tracing accelerators<\/h3>\n<p>Game graphics cards long used various simplified models for lighting calculations. However, in 2018, ray tracing technology began to appear in games. This is a previously known technology enabling realistic simulation of lighting, as well as shadows, reflections, and light scattering from mirror-like or similar surfaces. Ray tracing simulates the path of light rays and their interaction with objects. It is thus possible to achieve a more realistic image with it. However, this technology was long unattainable due to high performance demands.<\/p>\n<figure id=\"attachment_240744\" aria-describedby=\"caption-attachment-240744\" style=\"width: 1800px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-240744 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech.jpg\" alt=\"\" width=\"1800\" height=\"720\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech.jpg 1800w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech-300x120.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech-768x307.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2020\/09\/Akcelerace-ray-tracingu-na-RT-j\u00e1drech-1024x410.jpg 1024w\" sizes=\"auto, (max-width: 1800px) 100vw, 1800px\" \/><\/a><figcaption id=\"caption-attachment-240744\" class=\"wp-caption-text\">Ray tracing acceleration on RT cores<\/figcaption><\/figure>\n<p>Therefore, since 2018 (Nvidia&#8217;s Turing architecture), GPUs contain special units aimed at accelerating certain parts of the ray analysis process. Primarily, this entails the analysis of their intersections with the triangles of objects in the scene, which uses so-called BVH auxiliary structures.<\/p>\n<p>These units again have different designations in different GPU families\u2014in Nvidia architectures they are called RT cores, Intel graphics designate them as RTUs, and AMD calls them Ray Accelerators.<\/p>\n<h3 class=\"western\">Multimedia engine<\/h3>\n<p>Over time, besides 3D gaming graphics, multimedia playback functions, especially video, were also transferred to the GPU. Their predecessors were the acceleration of color space conversion between YUV and RGB, image scaling (e.g., enlarging to full screen) or deinterlacing. Then GPUs began to be equipped with full-fledged blocks for complete video format decoding.<\/p>\n<p>These decoders (NVDEC in Nvidia GPUs) are capable of decompressing video with low power consumption and high performance thanks to usage of specialized circuits (during transcoding, it is sometimes possible to achieve hundreds of frames per second). Their use means that it is not necessary to load the computer&#8217;s CPU when playing back media.<\/p>\n<figure id=\"attachment_171285\" aria-describedby=\"caption-attachment-171285\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2023\/10\/nv-proviz-encode-decode-diagram-2461893.png\"><img loading=\"lazy\" decoding=\"async\" class=\"noborder wp-image-171285 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2023\/10\/nv-proviz-encode-decode-diagram-2461893-1024x588.png\" alt=\"\" width=\"640\" height=\"368\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2023\/10\/nv-proviz-encode-decode-diagram-2461893-1024x588.png 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2023\/10\/nv-proviz-encode-decode-diagram-2461893-300x172.png 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2023\/10\/nv-proviz-encode-decode-diagram-2461893-768x441.png 768w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-171285\" class=\"wp-caption-text\">Overview of video format support for decoding and encoding NVDEC\/NVENC on GeForce RTX 4000 (Ada Lovelace) graphics cards<\/figcaption><\/figure>\n<p>Following decoders, GPUs later began to integrate similar blocks for encoding, i.e., video compression (in Nvidia, the encoder is called NVENC). Their use is for example for video conferences, for recording or streaming video from games, or for general work in video processing and editing. Today&#8217;s multimedia engines in GPUs support H.264, HEVC, VP9 formats and, most recently, the AV1 format.<\/p>\n<h3 class=\"western\">Video output block (Display Engine)<\/h3>\n<p>The output of the ROP units consists of so-called framebuffer data, which the display output block can then send for display on the monitor. This functionality includes the possibility of vertical synchronization (VSync), which ensures that data is sent to the monitor when a new frame is ready, and it does not happen that the frame being sent is replaced by a new one mid-transmission (which causes so-called tearing, where you see an older frame in the top part of the monitor, but somewhere in the middle there is a split and the bottom part of the monitor shows a newer one).<\/p>\n<p>The display output of modern GPUs has also gained more advanced functionality\u2014today, for example, it can refresh the image on the monitor with a variable frequency (this is so-called adaptive refresh using technologies like FreeSync, Nvidia G-Sync, or HDMI Variable Refresh Rate) and also send data to the monitor in a compressed form using VESA DSC technology.<\/p>\n<p><em><strong>Please note:<\/strong> The article continues in the following chapter.<\/em><\/p>\n<script async src=\"\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js\"><\/script>\n<!-- responsive -->\n<ins class=\"adsbygoogle\"\n     style=\"display:block;background-color:transparent\"\n     data-ad-client=\"ca-pub-8150419924824893\"\n     data-ad-slot=\"6522017574\"\n     data-ad-format=\"auto\"><\/ins>\n<script>\n(adsbygoogle = window.adsbygoogle || []).push({});\n<\/script>\n<p><!--nextpage-->A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated in the processor (or earlier, in the chipset). Here, however, we will focus on discrete graphics cards.<!--more--><\/p>\n<h3 class=\"western\">Technical details for the curious: How does a GPU as a processor differ from a CPU?<\/h3>\n<p>A GPU can be understood as a processor somewhat similar to a CPU, but built for massive parallelism using up to thousands of simultaneously processed threads. A good thing to keep in mind is that in a GPU, the part corresponding to cores of a CPU would be the entire compute units (SM blocks in Nvidia, so-called CUs or WGPs in AMD), not what is commonly called CUDA cores, shaders, or stream processors.<\/p>\n<p>A GPU core uses very wide <a href=\"https:\/\/en.wikipedia.org\/wiki\/Single_instruction,_multiple_data\">SIMD units<\/a> and the individual shaders or CUDA cores are actually just one SIMD &#8220;lane&#8221; of such a unit. The essence is that in a SIMD unit, one instruction always performs the same operation simultaneously with many data points in the individual lanes, like cars going in the same direction side by side on a multi-lane highway. In a GPU, these SIMD units typically have 32 lanes, meaning 32 values with which the same instruction is calculated at once (one compute unit, i.e., a &#8220;GPU core,&#8221; can however contain two or four of these SIMD units). Individual shaders are therefore not independent and must always process the same instruction.<\/p>\n<figure id=\"attachment_260747\" aria-describedby=\"caption-attachment-260747\" style=\"width: 402px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro.png\"><img loading=\"lazy\" decoding=\"async\" class=\"noborder wp-image-260747\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro.png\" alt=\"\" width=\"402\" height=\"679\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro.png 804w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro-178x300.png 178w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro-607x1024.png 607w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Schema-jedne-vypocetni-jednotky-SM.-Obsahuje-ctyri-SIMD32-jednotky-cache-texturovaci-jednotky-a-RT-jadro-768x1296.png 768w\" sizes=\"auto, (max-width: 402px) 100vw, 402px\" \/><\/a><figcaption id=\"caption-attachment-260747\" class=\"wp-caption-text\">Diagram of one SM compute unit. Contains four SIMD32 (working with 32 values at once) units, shared cache, texturing units, and an RT core<\/figcaption><\/figure>\n<p>Unlike a CPU, a single GPU core (i.e., an SM or CU block) does not have very high single-threaded performance and cannot handle memory latency or branching well. Therefore, not just any code can run efficiently on a GPU. It is suitable for processing massive amounts of data when the processing can proceed in many parallel threads simultaneously, and for compute tasks without complex branching and algorithm behavior and interdependence between individual threads. Therefore, the GPU is suitable for game graphics calculations, numerical calculations on a massive scale, or AI acceleration (which consists of matrix calculations over large volumes of data). Tasks such as decompressing a zip file, processing JavaScript on web pages, or running the operating system would, on the contrary, perform poorly on the GPU architecture.<\/p>\n<h3 class=\"western\">The role of API<\/h3>\n<p>An important difference between a GPU and a CPU is also that we usually do not run program binary code on a GPU in the same way as you run a program on a CPU. Running code on the GPU is mediated by a software layer in the drivers, which translates the program code (e.g., game shaders) into compiled code using the instructions of the specific GPU type. Therefore, GPUs do not have a stable and compatible instruction set like, for example, x86 processors. Instruction sets in GPUs can change between generations, and these differences are hidden behind the driver (software &#8220;stack&#8221;), which itself represents the stable interface that the software developers target.<\/p>\n<p>Instead of directly executing compiled binary code, various APIs are used when programming GPUs, which create an abstraction from the actual architecture of the specific GPU. For games and 3D graphics, the DirectX API on the Windows platform and the cross-platform Vulkan and OpenGL APIs are traditionally used. For compute usage, there are the OpenCL API and, very widely used today, the CUDA software layer offered by Nvidia GPUs. The equivalent for AMD GPUs is the ROCm layer, and for Intel GPUs, it is oneAPI. These APIs are implemented and provided to applications by the drivers.<\/p>\n<h3 class=\"western\">Graphics pipeline: How is a game image created in the GPU?<\/h3>\n<p>The process of rendering 3D graphics frames takes place in the so-called graphics pipeline, which consists of a larger number of steps. In the historical beginnings, more steps of this process were entrusted to the CPU,before they began to be accelerated on the GPU. The pipeline initially largely consisted of so-called fixed functions, meaning that operations were performed by the graphics processor in a given way, usually on hardware units dedicated for that purpose (this principle still applies today for units like ROPs and TMUs).<\/p>\n<p>However, over time, with the growing importance of shader programs in games, an increasingly larger part of the pipeline became flexible and programmable. Currently, its predominant part is processed by general-purpose compute units (&#8220;shaders&#8221;) as shader programs. This has the advantage that the software author can modify and influence these processes in various ways and thus has greater freedom of options when creating games or other applications.<\/p>\n<h4>1. Application computations on the CPU<\/h4>\n<p>The beginning of image processing in the graphics pipeline lies outside the GPU itself. GPU is not a standalone computing system, and the application\u2014whether a game or some work program\u2014runs primarily on the computer&#8217;s CPU. In the case of a game, its &#8220;engine&#8221; and other components run on the processor, as well as the processing of inputs from the gamepad, keyboard, or mouse, and other inputs (simulation of enemy AI in the game, movements of opponents received over the network). The application must also handle object collisions in this phase, for example.<\/p>\n<p>The game engine tasks the graphics card (GPU) with rendering the next frame based on these inputs. This task includes all elements (so-called &#8220;primitives&#8221;) that are to be in the resulting frame, i.e., objects composed of polygons and textures, their lighting, and effects applied to them. Along with the objects, the &#8220;camera&#8221; is also defined, specifying what parts of the scene will be projected onto the final image.<\/p>\n<p>When the application hands over the frame elements to the GPU for rendering, it communicates via an API and the graphics driver that implements the API and which itself &#8220;talks&#8221; to the actual graphics chip. The shader code is also first processed on the processor; the GPU drivers must compile the shaders for the architecture of the used GPU chip. This compilation is usually performed in advance (when starting the game, before loading the level) and the result is often stored in a cache on the disk so that compilation does not have to be repeated constantly. Shader compilation is not a regular part of the graphics pipeline; it stands outside (however, shaders can sometimes be compiled &#8220;just in time&#8221; during the game&#8217;s runtime).<\/p>\n<h4 class=\"western\">2. Geometry and lighting processing<\/h4>\n<p>The first step that is performed on the GPU itself, is the processing of the scene geometry and the manipulation and transformation that need to be performed on it. Geometry refers to the &#8220;wireframe model&#8221; of the 3D scene formed by polygons, which consist of individual triangles and their vertices.<\/p>\n<p>In 3D graphics, every shape is composed of a mesh of triangles (a triangle is one of the basic graphical &#8220;primitives,&#8221; besides points and lines) connected by their edges and vertices, which overall form the shape of the object. The triangle is used as the basis because it always lies in a single plane\u2014in history, there were attempts to use quadrilaterals as the basis of 3D graphics, but they more or less did not catch on. The visible surface of the object is created by assigning textures, in which the two-dimensional image of the surface is stored, to this combination of triangles.<\/p>\n<figure id=\"attachment_260748\" aria-describedby=\"caption-attachment-260748\" style=\"width: 512px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Priklad-polygonoveho-modelu-objektu-na-kteerm-je-viditelna-skladba-z-trojuhelniku.png\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260748 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Priklad-polygonoveho-modelu-objektu-na-kteerm-je-viditelna-skladba-z-trojuhelniku.png\" alt=\"\" width=\"512\" height=\"512\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Priklad-polygonoveho-modelu-objektu-na-kteerm-je-viditelna-skladba-z-trojuhelniku.png 512w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Priklad-polygonoveho-modelu-objektu-na-kteerm-je-viditelna-skladba-z-trojuhelniku-300x300.png 300w\" sizes=\"auto, (max-width: 512px) 100vw, 512px\" \/><\/a><figcaption id=\"caption-attachment-260748\" class=\"wp-caption-text\">Example of a polygonal model of an object, showing the composition of triangles (source: Nvidia)<\/figcaption><\/figure>\n<p>Scene geometry was once processed in a fixed manner, but today it is programmable using geometry shaders, mesh shaders, and other types of shader programs (historically, the first type of shaders working with geometry were vertex shaders). Geometry processing today can also itself generate new geometry elements directly using code running on the GPU\u2014an example is so-called tessellation, which can add additional vertices and shapes to a part of the original model.<\/p>\n<p>As part of geometry processing, so-called culling also takes place. This involves detecting objects that are not visible either because they are outside the camera&#8217;s field of view or are occluded by objects closer to the camera. GPUs architectures constantly try to improve their capabilities to detect such parts of the scene because removing them from the scene at the earliest possible stage of processing helps preserve more performance for objects that are visible.<\/p>\n<p>The processing of lighting in the scene also partially belongs to geometry processing, although it is actually realized in a later phase. In this phase of the graphics pipeline, values (gain) indicating the degree of illumination can be calculated and assigned to individual triangles, or vertices, which will then be applied later in the next phase. Lighting can be global or may come from local sources. However, its simulation is significantly changed by ray tracing rendering, if it is used.<\/p>\n<figure id=\"attachment_260749\" aria-describedby=\"caption-attachment-260749\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260749 size-large\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely-1024x824.jpg\" alt=\"Textury a v\u00fdsledn\u00fd objekt po namapov\u00e1n\u00ed na polygonov\u00e9 modely (zdroj: Nvidia)\" width=\"640\" height=\"515\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely-1024x824.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely-300x242.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely-768x618.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Textury-a-vysledny-objekt-po-namapovani-na-polygonove-modely.jpg 1098w\" sizes=\"auto, (max-width: 640px) 100vw, 640px\" \/><\/a><figcaption id=\"caption-attachment-260749\" class=\"wp-caption-text\">Textures and the resulting object after mapping onto polygonal models (source: Nvidia)<\/figcaption><\/figure>\n<h4 class=\"western\">3. Rasterization<\/h4>\n<p>In the previous steps, the scene&#8217;s objects existed in a form defined by their geometry in a 3D space given in numerical values. Rasterization consists of projecting the scene as a two-dimensional image into a plane, which is now formed by a raster of individual pixels, based on the previous data. During this projection, the frame&#8217;s pixels are thus created.<\/p>\n<p>Each individual pixel (sometimes the term fragment is used) as a set of color values is calculated by determining which primitives (triangles) form it. Textures (surfaces) are mapped onto the triangles according to the task the GPU received at the beginning\u2014according to this mapping, texture samples corresponding to the calculated pixels are selected. The effect of lighting intensity (gain) calculated in the previous step is also applied. And before the pixel is written, the required pixel shaders (or &#8220;fragment shaders&#8221;) are applied to it.<\/p>\n<figure id=\"attachment_260750\" aria-describedby=\"caption-attachment-260750\" style=\"width: 640px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Geometricke-utvary-a-jeich-vystup-po-rasterizace-ilustrace.gif\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260750\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Geometricke-utvary-a-jeich-vystup-po-rasterizace-ilustrace.gif\" alt=\"Geometrick\u00e9 \u00fatvary a jejich v\u00fdstup po rasterizaci (ilustrace: Drummyfish, Wikimedia Commons)\" width=\"640\" height=\"415\" \/><\/a><figcaption id=\"caption-attachment-260750\" class=\"wp-caption-text\">Geometric shapes and their output after rasterization into pixels (illustration: Drummyfish, <a href=\"https:\/\/commons.wikimedia.org\/wiki\/File:Top-left_triangle_rasterization_rule.gif\">Wikipedia Commons<\/a>)<\/figcaption><\/figure>\n<p>During this phase, it is also calculated how pixels affected by, for example, semi-transparency or transparency of some object should look, by using alpha (transparency) values. To improve quality and prevent aliasing (the effect of visible transitions between pixels, leading to the perception of jagged edges and lines), anti-aliasing methods are used, such as MSAA, where multiple samples are used to determine the color values of a pixel, simulating a higher resolution.<\/p>\n<p>Rasterization as such used to be a fixed process, and fundamentally that still holds true. But over time, it gained programmability. Various types of shader programs can also be applied to the points, making it possible to implement more complex effects than those possible strictly through the process of rasterizing polygons and textures.<\/p>\n<h4 class=\"western\">4. Post-processing and display<\/h4>\n<p>Once the raster image is complete, it can be sent to the monitor, which is the final stage of the graphics pipeline. This happens by the GPU&#8217;s output block starting to read the data from the frame buffer, where the result is stored in memory, and gradually sending it via the monitor interface. To prevent sending an incomplete frame, double buffering and the previously mentioned vertical synchronization can be used, where drawing is done into a different space than the one from which the GPU&#8217;s output block reads data for the monitor. After completing a frame, these buffers are swapped.<\/p>\n<p>However, in modern games, many additional operations working further with the frame assembled in the previous step, called post-processing, can take place between the step of rasterization into points and sending the frame to the monitor. Shader programs can apply additional effects here, perform frame upscaling, or color adjustment (for example, for the needs of HDR image mapping). Some anti-aliasing methods operate precisely in this phase, working as a post-processing pass.<\/p>\n<figure id=\"attachment_260753\" aria-describedby=\"caption-attachment-260753\" style=\"width: 1200px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-260753 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny.jpg\" alt=\"\" width=\"1200\" height=\"630\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny.jpg 1200w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny-300x158.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny-1024x538.jpg 1024w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2025\/09\/Polygonalni-model-tvorici-zaklad-3D-sceny-768x403.jpg 768w\" sizes=\"auto, (max-width: 1200px) 100vw, 1200px\" \/><\/a><figcaption id=\"caption-attachment-260753\" class=\"wp-caption-text\">Polygonal model forming the basis of a 3D scene. Together with textures, it approximates the real world (illustration, source: Nvidia)<\/figcaption><\/figure>\n<h3 class=\"western\">Upscaling and frame generation<\/h3>\n<p>In recent years, upscaling or frame interpolation technologies have become important in games. Their goal is to increase the output frame rate of 3D rendering beyond what the pure computate performance of the GPU allows. Upscaling (also called Super Resolution technique) consists of the game or program being rendered at a lower resolution, which is then upscaled to a higher one within post-processing.<\/p>\n<p>To bring the quality closer to that of full (so-called &#8220;native&#8221;) resolution, advanced techniques such as deploying <strong>artificial intelligence<\/strong> for scaling the raster image of the frame are used. This replaces common scaling algorithms. The AI model for this use is trained on a corpus of image data composed of high-quality reference frames and versions of the same frames at lower resolution. The AI then, during training, learns to derive an image that most closely resembles the high-quality reference frame from the low-resolution version of the frame. During game upscaling, the artificial intelligence then makes these learned changes. These techniques can utilize special units for AI acceleration in the GPU.<\/p>\n<p><strong>Temporal reconstruction<\/strong> applied after upscaling is one of the most effective techniques for improving quality of upscaling. It uses data about movement in the scene for so-called motion compensation, which allows combining the image information in the current frame with information from previous frames. This is possible because, based on the knowledge of motion vectors in the scene, we know which points from previous frames can be used as a reference for a point in the current frame (i.e., which points in different frames belong to a certain object).<\/p>\n<p>By combining them, we have more visual information for the current frame than is present in its rendered version with reduced resolution, so by adding this information, the image in the frame can be improved. Thanks to temporal filtering across several consecutive frames, this technique can suppress some image artifacts that spatial techniques (which means those processing only within the pixel raster of a single frame) have difficulty eliminating or preventing, and which can manifest as shimmering or flickering of the image. On the other hand, temporal filtering risks introducing its own set of image errors (&#8220;ghosting&#8221; or blurring) will arise from incorrect application of the motion compensation and filtering, which requires careful attention when designing and tuning these techniques.<\/p>\n<p>What is important for the temporal reconstruction is that that samples of the input lower-resolution frames prior to the upscaling are rendered by the GPU with small subpixel positional offset, which is varied in every consecutive frame (the offsets are compensated for during the upscaling and the temporal filtering so that they don&#8217;t negatively affect the image seen). The offset positioning (jitter) of the samples effectively improves the resolution of the upscaled frames, once the temporal reconstruction technique is used to combine the offset information in previous frames into the current one. Resulting level of visual quality is therefore higher than what would be achieved if the samples of the input lower resolution frames used for the upscaling were from the same positions.<\/p>\n<figure id=\"attachment_116763\" aria-describedby=\"caption-attachment-116763\" style=\"width: 3588px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-116763 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation.jpg\" alt=\"\" width=\"3588\" height=\"1797\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation.jpg 3588w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation-300x150.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation-768x385.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/nvidia-dlss-3-motion-optical-flow-estimation-1024x513.jpg 1024w\" sizes=\"auto, (max-width: 3588px) 100vw, 3588px\" \/><\/a><figcaption id=\"caption-attachment-116763\" class=\"wp-caption-text\">DLSS 3, when generating frames, combines vectors communicated by the game with vectors found by the Optical Flow technique for better quality (source: Nvidia)<\/figcaption><\/figure>\n<p>Motion compensation using motion vectors also underlies the technology of so-called <strong>frame generation<\/strong>. This technique estimates the movement of objects in the scene between frames that the GPU has already rendered based on motion information. Based on this, it can then assemble a new intermediate frame that estimates what the position of the objects should be at a point in time between the actually rendered frames. It is a similar technique to that used to increase the frame rate by TV Sets. The GPU can thus add one or more intermediate frames between every two frames actually rendered by the game.<\/p>\n<figure id=\"attachment_116762\" aria-describedby=\"caption-attachment-116762\" style=\"width: 2048px\" class=\"wp-caption aligncenter\"><a href=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works.jpg\"><img loading=\"lazy\" decoding=\"async\" class=\"wp-image-116762 size-full\" src=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works.jpg\" alt=\"\" width=\"2048\" height=\"1028\" srcset=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works.jpg 2048w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works-300x151.jpg 300w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works-768x386.jpg 768w, https:\/\/www.hwcooling.net\/wp-content\/uploads\/2022\/09\/how-nvidia-dlss-3-works-1024x514.jpg 1024w\" sizes=\"auto, (max-width: 2048px) 100vw, 2048px\" \/><\/a><figcaption id=\"caption-attachment-116762\" class=\"wp-caption-text\">Functioning of the DLSS 3 pipeline with frame generation (source: Nvidia)<\/figcaption><\/figure>\n<p>The disadvantage is that generated frames can suffer from lower quality or visual errors caused by incorrectly estimated motion or resulting from the fact that only incomplete information is available as a basis for generating these frames. This technique also requires that the last two frames are already finished to start its work, and one of the frames then always waits for the display of the generated frames before it can be displayed as well. Frame generation therefore slightly increases the latency (delay between input and image) of the game. The image in the generated frame also does not react in any way to control input or events in the game, because it is assembled purely from the output visual data of adjacent frames and the game program had nothing to do with it, not being able to influence the contents.<\/p>\n<p>These techniques appeared in 2018 and the following years, first in Nvidia&#8217;s DLSS technology (requires GeForce graphics cards), then also within alternatives: FSR from AMD and XeSS from Intel. Their arrival coincides temporally with the spread of ray tracing graphics and they can serve as a tool that compensates for the high computational demands of ray tracing. At the same time, they can help achieve higher frame rates on special high refresh rate monitors or on monitors and televisions with very high resolutions, such as 8K.<\/p>\n<p style=\"text-align: right;\"><em>English translation and edit by Jozef Dud\u00e1\u0161<\/em><\/p>\n<script async src=\"\/\/pagead2.googlesyndication.com\/pagead\/js\/adsbygoogle.js\"><\/script>\n<!-- responsive -->\n<ins class=\"adsbygoogle\"\n     style=\"display:block;background-color:transparent\"\n     data-ad-client=\"ca-pub-8150419924824893\"\n     data-ad-slot=\"6522017574\"\n     data-ad-format=\"auto\"><\/ins>\n<script>\n(adsbygoogle = window.adsbygoogle || []).push({});\n<\/script>\n","protected":false},"excerpt":{"rendered":"<p>A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated [&hellip;]<\/p>\n","protected":false},"author":26,"featured_media":40123,"comment_status":"open","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[770,658],"tags":[],"class_list":["post-261260","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-analysis","category-graphics"],"yoast_head":"<!-- This site is optimized with the Yoast SEO plugin v24.9 - https:\/\/yoast.com\/wordpress\/plugins\/seo\/ -->\n<title>How a graphics card works. Intended for secondary schools - HWCooling.net<\/title>\n<meta name=\"robots\" content=\"index, follow, max-snippet:-1, max-image-preview:large, max-video-preview:-1\" \/>\n<link rel=\"canonical\" href=\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/\" \/>\n<meta property=\"og:locale\" content=\"en_US\" \/>\n<meta property=\"og:type\" content=\"article\" \/>\n<meta property=\"og:title\" content=\"How a graphics card works. Intended for secondary schools - HWCooling.net\" \/>\n<meta property=\"og:description\" content=\"A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated [&hellip;]\" \/>\n<meta property=\"og:url\" content=\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/\" \/>\n<meta property=\"og:site_name\" content=\"HWCooling.net\" \/>\n<meta property=\"article:published_time\" content=\"2025-10-06T10:00:29+00:00\" \/>\n<meta property=\"article:modified_time\" content=\"2026-08-01T00:16:13+00:00\" \/>\n<meta property=\"og:image\" content=\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-630px.jpg\" \/>\n\t<meta property=\"og:image:width\" content=\"1200\" \/>\n\t<meta property=\"og:image:height\" content=\"630\" \/>\n\t<meta property=\"og:image:type\" content=\"image\/jpeg\" \/>\n<meta name=\"author\" content=\"Jan Ol\u0161an\" \/>\n<meta name=\"twitter:card\" content=\"summary_large_image\" \/>\n<meta name=\"twitter:label1\" content=\"Written by\" \/>\n\t<meta name=\"twitter:data1\" content=\"Jan Ol\u0161an\" \/>\n\t<meta name=\"twitter:label2\" content=\"Est. reading time\" \/>\n\t<meta name=\"twitter:data2\" content=\"40 minutes\" \/>\n<script type=\"application\/ld+json\" class=\"yoast-schema-graph\">{\"@context\":\"https:\/\/schema.org\",\"@graph\":[{\"@type\":\"WebPage\",\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/\",\"url\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/\",\"name\":\"How a graphics card works. Intended for secondary schools - HWCooling.net\",\"isPartOf\":{\"@id\":\"https:\/\/www.hwcooling.net\/#website\"},\"primaryImageOfPage\":{\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage\"},\"image\":{\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage\"},\"thumbnailUrl\":\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg\",\"datePublished\":\"2025-10-06T10:00:29+00:00\",\"dateModified\":\"2026-08-01T00:16:13+00:00\",\"author\":{\"@id\":\"https:\/\/www.hwcooling.net\/#\/schema\/person\/1a1c9f238b83289e7c87b6fc07ad20ee\"},\"breadcrumb\":{\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#breadcrumb\"},\"inLanguage\":\"en-US\",\"potentialAction\":[{\"@type\":\"ReadAction\",\"target\":[\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/\"]}]},{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage\",\"url\":\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg\",\"contentUrl\":\"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg\",\"width\":3840,\"height\":2560},{\"@type\":\"BreadcrumbList\",\"@id\":\"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#breadcrumb\",\"itemListElement\":[{\"@type\":\"ListItem\",\"position\":1,\"name\":\"Dom\u016f\",\"item\":\"https:\/\/www.hwcooling.net\/\"},{\"@type\":\"ListItem\",\"position\":2,\"name\":\"How a graphics card works. Intended for secondary schools\"}]},{\"@type\":\"WebSite\",\"@id\":\"https:\/\/www.hwcooling.net\/#website\",\"url\":\"https:\/\/www.hwcooling.net\/\",\"name\":\"HWCooling.net\",\"description\":\"Performance can have many forms...\",\"potentialAction\":[{\"@type\":\"SearchAction\",\"target\":{\"@type\":\"EntryPoint\",\"urlTemplate\":\"https:\/\/www.hwcooling.net\/?s={search_term_string}\"},\"query-input\":{\"@type\":\"PropertyValueSpecification\",\"valueRequired\":true,\"valueName\":\"search_term_string\"}}],\"inLanguage\":\"en-US\"},{\"@type\":\"Person\",\"@id\":\"https:\/\/www.hwcooling.net\/#\/schema\/person\/1a1c9f238b83289e7c87b6fc07ad20ee\",\"name\":\"Jan Ol\u0161an\",\"image\":{\"@type\":\"ImageObject\",\"inLanguage\":\"en-US\",\"@id\":\"https:\/\/www.hwcooling.net\/#\/schema\/person\/image\/\",\"url\":\"https:\/\/secure.gravatar.com\/avatar\/2bd32e7382bccc62b0ad57f0b361009d6d4624e68c8f39fb28935312b6d9b5bc?s=96&d=mm&r=g\",\"contentUrl\":\"https:\/\/secure.gravatar.com\/avatar\/2bd32e7382bccc62b0ad57f0b361009d6d4624e68c8f39fb28935312b6d9b5bc?s=96&d=mm&r=g\",\"caption\":\"Jan Ol\u0161an\"},\"url\":\"https:\/\/www.hwcooling.net\/en\/author\/jano\/\"}]}<\/script>\n<!-- \/ Yoast SEO plugin. -->","yoast_head_json":{"title":"How a graphics card works. Intended for secondary schools - HWCooling.net","robots":{"index":"index","follow":"follow","max-snippet":"max-snippet:-1","max-image-preview":"max-image-preview:large","max-video-preview":"max-video-preview:-1"},"canonical":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/","og_locale":"en_US","og_type":"article","og_title":"How a graphics card works. Intended for secondary schools - HWCooling.net","og_description":"A graphics card is a computer component that processes graphics in games but can also serve for compute or various professional graphics tasks. Besides compute units, it has another simpler role\u2014it provides the actual video output to the monitor connected to the graphics card\u2019s ports. This role can also be handled by a GPU integrated [&hellip;]","og_url":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/","og_site_name":"HWCooling.net","article_published_time":"2025-10-06T10:00:29+00:00","article_modified_time":"2026-08-01T00:16:13+00:00","og_image":[{"width":1200,"height":630,"url":"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-630px.jpg","type":"image\/jpeg"}],"author":"Jan Ol\u0161an","twitter_card":"summary_large_image","twitter_misc":{"Written by":"Jan Ol\u0161an","Est. reading time":"40 minutes"},"schema":{"@context":"https:\/\/schema.org","@graph":[{"@type":"WebPage","@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/","url":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/","name":"How a graphics card works. Intended for secondary schools - HWCooling.net","isPartOf":{"@id":"https:\/\/www.hwcooling.net\/#website"},"primaryImageOfPage":{"@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage"},"image":{"@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage"},"thumbnailUrl":"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg","datePublished":"2025-10-06T10:00:29+00:00","dateModified":"2026-08-01T00:16:13+00:00","author":{"@id":"https:\/\/www.hwcooling.net\/#\/schema\/person\/1a1c9f238b83289e7c87b6fc07ad20ee"},"breadcrumb":{"@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#breadcrumb"},"inLanguage":"en-US","potentialAction":[{"@type":"ReadAction","target":["https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/"]}]},{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#primaryimage","url":"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg","contentUrl":"https:\/\/www.hwcooling.net\/wp-content\/uploads\/2021\/01\/rx-6800-xt-vs-rtx-3080-00-v2.jpg","width":3840,"height":2560},{"@type":"BreadcrumbList","@id":"https:\/\/www.hwcooling.net\/en\/how-a-graphics-card-works-dedicated-to-secondary-schools\/#breadcrumb","itemListElement":[{"@type":"ListItem","position":1,"name":"Dom\u016f","item":"https:\/\/www.hwcooling.net\/"},{"@type":"ListItem","position":2,"name":"How a graphics card works. Intended for secondary schools"}]},{"@type":"WebSite","@id":"https:\/\/www.hwcooling.net\/#website","url":"https:\/\/www.hwcooling.net\/","name":"HWCooling.net","description":"Performance can have many forms...","potentialAction":[{"@type":"SearchAction","target":{"@type":"EntryPoint","urlTemplate":"https:\/\/www.hwcooling.net\/?s={search_term_string}"},"query-input":{"@type":"PropertyValueSpecification","valueRequired":true,"valueName":"search_term_string"}}],"inLanguage":"en-US"},{"@type":"Person","@id":"https:\/\/www.hwcooling.net\/#\/schema\/person\/1a1c9f238b83289e7c87b6fc07ad20ee","name":"Jan Ol\u0161an","image":{"@type":"ImageObject","inLanguage":"en-US","@id":"https:\/\/www.hwcooling.net\/#\/schema\/person\/image\/","url":"https:\/\/secure.gravatar.com\/avatar\/2bd32e7382bccc62b0ad57f0b361009d6d4624e68c8f39fb28935312b6d9b5bc?s=96&d=mm&r=g","contentUrl":"https:\/\/secure.gravatar.com\/avatar\/2bd32e7382bccc62b0ad57f0b361009d6d4624e68c8f39fb28935312b6d9b5bc?s=96&d=mm&r=g","caption":"Jan Ol\u0161an"},"url":"https:\/\/www.hwcooling.net\/en\/author\/jano\/"}]}},"_links":{"self":[{"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/posts\/261260","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/users\/26"}],"replies":[{"embeddable":true,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/comments?post=261260"}],"version-history":[{"count":3,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/posts\/261260\/revisions"}],"predecessor-version":[{"id":279714,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/posts\/261260\/revisions\/279714"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/media\/40123"}],"wp:attachment":[{"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/media?parent=261260"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/categories?post=261260"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.hwcooling.net\/en\/wp-json\/wp\/v2\/tags?post=261260"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}