The very first computers doesn’t have graphics card. The processor processes all the tasks including graphics processing. And then the graphics card were created, also called video card, it unloads the graphical portion from the processor, process it, and then output it in a display device. The graphics processing unit (GPU) is the actual chip that is put on the printed circuit board (PCB). The PCB also has other components - the memory chips, voltage regulator modules (VRM) consists of transistors, capacitors and chokes, heatsink and fan, display outputs (ports), and some other components to make the graphics card operational. Generally, graphics cards has branched out into two types based on their usage - desktop graphics card and workstation graphics card. Desktop graphics card are the most common, it is used to process desktop graphics - video files, basic 2D and 3D graphics, and gaming. Workstation graphics card are made specifically for workstation software like CAD, 3D rendering, and simulations. The two can do what the other can do, but not as efficient. You can can still play modern games using workstation graphics card but performance usually slower than when game with a real gaming graphics card. Same with desktop graphics card, you can still use modern desktop graphics card for CAD or simulations, but you better use workstation graphics card for those type of software to speed up the process. nVidia and AMD are the two major GPU manufacturers. They design the chip, make a “reference” card design and then distribute it to Add In Board (AIB) partners. These AIB partners - ASUS, Sapphire, EVGA and others - make the cards based on the reference given to them, and also make custom board designs. So what should you consider when buying a graphics card? In this article we will give you pointers in choosing a graphics, specifically gaming graphics card. Most of these pointers can also be applied when you are buying a workstation graphics card.
The GPU, like CPU, also has cores. nVidia call these cores “CUDA Cores” and “Stream Processors” for AMD, and there are several hundreds of these, or even thousands. These cores are the “workers” that processes the data fed into the GPU. More cores means the tasks will finish faster, equating to better performance.
Core Clock and Boost Clock
Core clock is the “speed” of the GPU, measured in Hertz. The core clock is usually the “base clock” while the “boost clock” is the speed of the card under heavy load. Some cards have the same base clocks but have different boost clocks. Some board manufacturers make some modifications and tuning to make this possible. Higher core clock and boost clock are good indications of a better card, if you comparing cards of the same model. You can compare the GTX 1050 card made by ASUS and the one that made by Palit, but not against GTX 1050 Ti, even though they are in the same series or generation.
Memory Clock, Bus Width, and Bandwidth
Memory clock is how fast the memory can transfer data to and from the GPU, also measured in Hertz. Bus width is how wide the “road” used to travel the data, measured in bits - 64-bit, 128-bit and so on. Wider bus width means more data can travel simultaneously to and from GPU cores. Memory Bandwidth is the combination of memory clock and bus width. Bigger bandwidth means better performance.
Memory size and Memory Type
Memory size, also called “frame buffer”, is the amount of data it can handle, measured in gigabytes (GB) nowadays. This is where processed data, often textures, are stored. Having bigger video memory means it can store more data ready for use and doesn’t need processing again. Memory type is what type of memory is used for that specific card. This is listed as GDDR or HBM in the specifications sheet. Below are the common make ups of different graphic cards. Having larger video memory doesn’t necessarily equate to better performance. Having narrow bus width will cripple the data transfer, even if you have larger video memory. Having wide bus width ensures better and faster data transfer, but having low video memory will limit the amount of data that can be transferred, resulting to some bandwidth being wasted. To maximize data processing and transfer, GPU with very high core counts are usually paired with wider memory bus width and large video memory. Basically, the GPU cores, boost clock, and memory bandwidth are the most important details of a graphics card and tells you which card is better.
Cooling and Size
The cooling components - heatsink, fan or waterblock - plays an important role as it keeps the GPU, memory and other components in their ideal operating temperature. If the GPU reaches its maximum operating temperature, the boost clock is lowered to keep the GPU in its ideal temperature at the expense of performance loss due to lowered boost clock. This process of lowering the clock to keep the GPU under maximum operating temperature is call Thermal Throttling. Better cooling components put into the card means it can maintain its temperature without reaching thermal throttling. Cards with bigger heatsink and more fans usually runs cooler, but they also take some space. Its length and thickness might block some ports of the motherboard, making them inaccessible or unusable. Some lengthy cards doesn’t fit in mATX and even in some mid tower ATX cases. But some board partners do make smaller versions of the card, having only a single fan, so you may want to take a look at different models of the card that will fit your case. There are also custom water cooled graphics card where it has its own cooling radiator in addition to heatsink and fan.
If you have a new monitor, you don’t have to worry about display output ports. The graphics card probably has the connector for your monitor. But if you are adding more monitors, your card may not have enough ports. Different cards has different types and number of ports.
You may also want to see if your card needs a PCI-E power connector. Midrange and high end cards needs a 6-pin or more PCI-E power connector, while low end cards doesn’t need this. Look at the system power requirement or recommended power supply to see if your power supply meet the required power and has the appropriate connector.
The addition metal backplate have become popular and now a standard component in midrange and high end cards. Metal backplate adds durability and prevent the bending of the card due to big heatsink. It also acts as another heatsink to help cool the backside of the card. RGB is now also a popular and standard feature in midrange and high end cards. You can ignore this feature. Unless you can use this feature as part of temperature monitoring, where you can set color for a target temperature, this feature is useless. Most board partners now also provide overclocking and monitoring software for intuitive and easy to use interface. Since the release of GeForce RTX series by nVidia, Ray Tracing started to become a trend in graphics and video game development. Ray tracing enabled cards produce photorealistic illuminations, reflections and shadows. Before, ray tracing is widely used only in workstation applications like 3D modeling and video productions. Now, not only nVidia is pushing this technology into mainstream desktop, AMD is planning to include ray tracing in their future graphics cards, including consoles. Crytek has also released a ray tracing demo using their in house game engine CRYENGINE. You may ignore this feature right now, but if you want your graphics card to play future games with photorealistic graphics, make sure you card supports ray tracing.
Recommendations - How to buy the right graphics card for you?
Have a list of games you are currently playing and planning to play in the near future, then look at their system requirements. Most of the popular online/esport games like DOTA 2, League of Legends, and Overwatch doesn’t need high end graphics card, you can even play these games on a modern integrated graphics, but a $120 to $150 graphics card will give you better and very playable results. We recommend AMD RX 560 or GeForce GTX 1050 in this price range. If you want to play AAA titles that are graphics hungry like Metro Exodus, you need at least $250. We recommend GeForce GTX 1660 Ti. Depending on your set budget, buy the best graphics card available in the price of the same model. Using the guides we mentioned above, you should be able to pick the best graphics that perfectly fits in your budget and games. Let us say you want to play RAGE 2 and your budget is $170 to $200. Is it possible that a $190 graphics card can play RAGE 2? Lets take a look at RAGE 2 system requirements. RAGE 2 requires at least GeForce GTX 780 or AMD Radeon R9 280 with 3 GB of video memory. The recommended graphics card are GeForce GTX 1070 or AMD Vega 64, which are both beyond the $200 price range. Let’s take a look at the cards within $170 to $200 price range. The cheapest $153 EVGA GeForce GTX 1650 XC is well within the minimum required graphics card of RAGE 2, faster than GTX 780 and has 4GB of video memory. But the AMD Radeon RX 580 is much faster than the GTX 1650, and since RX 580 prices are under $200, we suggest choosing the RX 580 over the GTX 1650. Now let us check the specifications of the three RX 580 cards – the PowerColor AMD Radeon RED Dragon, the XFX Radeon RX 580 GTS XXX Edition, and the Sapphire Radeon Pulse RX 580. The PowerColor Red Dragon RX 580 has the slowest boost clock at 1350 MHz, while the Sapphire and XFX both has the faster boost clock at 1366 MHz. But upon looking at the XFX’s website, the XFX Radeon RX 580 GTS XXX edition can be safely overclocked up to 1386 MHz. They do provide a guide on how to achieve this. If you are into overclocking, the XFX Radeon RX 580 GTS XXX Edition is the best graphics card under your budget. You will still be fine if you pick either the Sapphire Pulse or the PowerColor Red Dragon over the XFX, just don’t get the MSI Armor which is over your price range and gain no performance increase. If you are still doubtful on your selected graphics card, read user and expert reviews from online computer retail stores and computer related websites. From there you will read about actual user experience and broader tests conducted by the technical reviewers.