- Benchmarking is an important process for determining the power of a graphics card and can be done using various software programs.
- Frames per second (FPS) is a key metric for measuring the power of a graphics card and determining the smoothness and detail of animations.
- Resolution alone is not a reliable indicator of a graphics card’s power, as processing high-resolution animations requires more power than processing still images.
Graphics are one of the most crucial aspects of the gaming world. Thus, PC gamers tend to put exceptional thought into the graphics card they choose for their PC. Selecting the proper GPU ensures you can play the newest games with the best graphical fidelity! So, let’s consider what factors you’ll want to consider when purchasing a graphics chipset for your gaming PC.
Graphics Metrics You Should Know About
Several methods for measuring the overall strength of a graphics card exist. You’ll need to consider all of them as a package deal when purchasing a graphics card. Few, if any, graphics cards will hit the maximum metrics on every axis. So, you’ll have to choose the ones that are most important to you.
One of the easiest ways to determine the power of a data processing chipset like a graphics card is to benchmark it. This process allows you to stress test the component and see how it performs while handling different workloads.
Several benchmarking programs for GPUs exist. Heaven UNIGINE is widely regarded as the best free software for testing the strength of a graphics chipset. However, other options, such as Novabench, Passmark, 3DMark, Geekbench, and MSI Afterburner, can continue your testing.
You mustn’t confuse GPU benchmarking with CPU benchmarking. Each software uses a different combination of central and graphics chipset power to process the data. Thus, some are relying more or less on your graphics card. While software like Heaven UNIGINE and PassMark may depend almost entirely on the video card, programs like Cinebench also use your central chipset.
If you use the wrong type of benchmarking software, you’ll get an inaccurate read of how powerful your GPU is. For instance, using Cinebench will give you comprehensive data on your central and graphics chipsets and how they work together. However, Heaven UNIGINE is a much better tool for clocking your graphics card since it doesn’t rely heavily on the CPU’s power.
Frames Per Second
Frames per second (FPS) is likely the most well-known metric we use to measure the power of our graphics cards. Animations on your computer are created by the graphics chipset rapidly redrawing the image on the screen. The faster your video card can recreate the image on the screen, the smoother and more detailed the animations will be.
Smooth animations are especially important to gamers who need to see how their game’s visuals move in incredibly high detail. When a game action is shown with more frames — more images per second — optical tells for different attacks, and other critical game information are visible to the user faster. Thus, a high FPS can be a tactical advantage for users.
However, it’s important to remember that FPS is limited by several pieces of hardware, not just the GPU. While a strong graphics card is needed to produce smooth, fast animations, you will also need a monitor capable of redrawing the screen that fast. If the monitor can’t create images fast enough, you’ll be artificially locked to a lower framerate since your monitor can’t physically produce a higher FPS.
FPS is an excellent way to check the power of your GPU. High-end graphics cards can process the game’s visuals fast enough to produce a high framerate. In contrast, lower-end graphics cards must adjust settings to achieve smooth animations at high resolutions.
When we assess the FPS a graphics card can produce, we typically cross-reference it with the resolution of the visuals since the resolution of the graphics influences the framerate.
Maximum FPS vs. FPS Stability
Consider also the difference between your video card’s maximum and stable FPS rates. While your graphics card might be able to achieve some beefy numbers at times, this probably doesn’t represent a stable framerate.
You don’t just want a graphics card that sometimes spikes into high numbers; you want one that stably and consistently produces high framerates. You may especially wish for FPS stability if you play multiplayer games. Since framerates are an essential advantage in multiplayer games, you want to have a high and consistent FPS.
If your framerate is jumping all over the place, you’ll see varying amounts of game information based on nothing besides how stressed your graphics card happens to be at that moment. Thus, multiplayer gamers will want to invest in a strong graphics card that will produce a consistently high FPS to give them the maximum advantage over their opponents.
Screens are made up of little dots called pixels. The more pixels on the screen, the more detail we can see in the graphics. Simply put, more dots in a space means more unique colors. So, if I make the same image out of ten unique color dots and then again out of thirty, you’ll see more definition in the one with more dots because there are more unique colors.
Processing pixel information is a pretty heady task. The computer must do several complex mathematical equations to create a binary representation of the pixel’s color, which we then translate to a color depth and map to a pixel on the screen. We must repeat this process for every single pixel on the screen whenever the graphics card redraws the image. So, it can happen potentially hundreds of times per second.
A graphics card must be appropriately powerful to handle this amount of information. Lower-end and even older, higher-end video cards simply lack the power to process that much data fast enough to produce a good framerate.
Resolution alone is typically a poor indicator of a graphics card’s power since processing a single image of a high resolution is much different from an animation of a high resolution. Thus, we typically cross-reference resolution with framerate to see how comfortably a video card can handle the data for high-resolution video images rather than stills.
Graphics Cards Metrics to Consider When Purchasing a New GPU
To achieve the practical graphics metrics above, you’ll want to consider the practical and theoretical metrics we use to benchmark video card performance. These statistics allow us to gather information about and compare different GPUs on both axes. Let’s assess the various standards we can use to measure chipset performance.
Floating-point operations are one of the most talked-about metrics for measuring graphics card power. These complex mathematical equations use very large or very small numbers expressed as a “significand,” with an exponent showing how many places the decimal in the significand has been moved by.
Since we use a lot of floating-point operations in 3D graphics processing, a GPU must be able to balance a large number of them to output stellar 3D visuals. As a theoretical standard approximating the chipset’s power, we measure the amount of floating-point operations per second (FLOPS) that a processor can handle.
In modern graphics cards, we typically look for measurements of floating-point operations in the teraFLOPS or billions of floating-point operations per second. However, this number will be influenced by the precision of our calculations.
Precision in Floating-Point Operations
Floating-point operations come in four precisions. These measure how many bits we can use to store the floating-point numbers. When you use more bits to store a floating point number, you can put more numbers in the significand, increasing the calculations’ accuracy.
However, you need a more powerful processor to store these bigger numbers. We look at a processor’s floating-point operations per second in three precisions: half-precision, single-precision, and double-precision. Quadruple-precision calculations are typically limited to supercomputers and scientific calculations requiring an extremely high degree of accuracy.
Modern graphics cards typically rank in the teraFLOPS when we measure half- and single-precision floating-point operations. However, you’ll typically see numbers in the gigaFLOPS for double-precision ones. This disparity is typical and simply a product of how expensive the semiconductors needed to do these calculations are.
Shading units, more commonly known simply “cores,” are the primary resource graphics cards use when rendering 3D graphics. Shaders produce all the scene’s geometry and pixel-based post-production effects like bloom and motion blur.
Most modern graphics engines run on shaders. Thus, the associated graphics cards have thousands of shading units to accommodate the heavy computational need. The more shading units a graphics card has installed on the PCB, the faster it can process the graphics on your screen.
Texture Mapping Units
Texture mapping units (TMUs) are essential to modern graphics cards. They assign textures to 3D polygons. TMUs can rotate, resize, and distort bitmap images over a 3D plane. Without your TMUs, all the polygons in your game would be lifeless and gray, literally.
Render Output Unit
Raster operations pipelines (ROPs) — sometimes called a render output unit — take the pixel and texel information and process them into a final pixel depth using specific vector and matrix operations.
You’ll also want to consider whether your projected GPU has enough multiprocessors for your tastes. People who want to stream games — even just on Discord to friends — will want to invest in a graphics chipset with many multiprocessors that can handle parallel computation for both playing and streaming the game.
NVIDIA GPUs call their multiprocessors “Stream Multiprocessors,” or SM counts, while AMD chipsets call them “Compute Units.” The higher the number, the easier it is to perform parallel processing while gaming.
NVIDIA users will want to make sure they also check their prospective GPUs Tensor Core count. These processing cores allow for better mixed-precision processing when a program uses multiple precision types in its floating-point operations to allow for greater or lesser accuracy, depending on the calculation.
Ray tracing is a graphical simulation of light that is very popular amongst AAA game developers, and for a good reason: ray-traced graphics look phenomenal. However, this process is very resource-intensive, which has led the manufacturers of graphics chipsets to produce unique, standalone hardware for processing ray tracing. If you’re a fan of ray-traced graphics, you’ll need to get a GPU that can handle this task, which is no small ask!
Of course, you must also consider the physicality of the graphics card you intend to purchase. Graphics chipsets aren’t just one of the most expensive components in a PC but also one of the largest. Most modern video cards are dual-slot cards, meaning they take up the space of two PCIe cards.
However, some models of modern graphics cards can take up three or even four PCIe slots. So, if you want to install additional PCIe cards like SATA controllers, Wi-Fi adapters, etc., you’ll need to account for the space your graphics card takes up!
Thermal Design Power
Thermal design power (TDP) is also essential when choosing a new graphics card. Graphics chipsets need power, and modern graphics cards require a lot of energy. Modern high-end video cards have a TDP that would have supported an entire low-end PC a decade ago. So, you’ll have to budget appropriately.
We measure TDP and power supplies in watts. All your components have a TDP, including your motherboard, CPU, and RAM. So, your GPU isn’t the only consideration you have to make regarding your PC’s energy consumption.
In addition to your components’ raw TDP, you’ll also want some operating space for fluctuations in power draw. This is especially true if you want to overclock any of your components. Overclocking is going to jack up the energy consumption.
Graphics cards also come preloaded with their own random access memory (RAM). Video RAM is where the graphics card stores data, such as textures, models, and other assets. By doing so, the chipset doesn’t need to reload all this information every time it enters the screen; it can just load the version saved in the VRAM.
The more VRAM a graphics card has, the more data it can store in between loads. Buying a video card with a lot of memory can improve load times between scenes.
Questions to Ask Yourself When Choosing a GPU for Your Gaming PC
You can find the best graphics card for your gaming PC by asking yourself a few questions. These questions will reveal what kind of graphics chipset will best suit your personal gaming experience. Let’s examine what you need to consider in detail.
What Games Will You Play?
Invest in a newer, more powerful chipset to play the latest and greatest games. Games like Cyberpunk 2077 are currently stressing even the highest-end video cards. So, if these games are on your radar, you’ll need an appropriately powerful card.
However, if you play games primarily geared towards any and everybody, like League of Legends, a beefy graphics card might not be something you need immediately. People who like older PC games can also reduce the power of their video cards as necessary since these games work for hardware far less powerful than the current generation of GPUs.
The exception is if you’re planning on emulating games. Emulating non-PC software on a PC is a very resource-intensive task. People planning on emulating games should purchase an appropriately strong graphics card.
How Big Is My Case?
This is a huge question that more people need to remember to ask themselves. Remember, it doesn’t matter how robust your graphics card is if it doesn’t fit your case. If you plan on purchasing one of the newer high-end graphics cards, please remember that those are massive. Do yourself a favor and check the physical dimensions to ensure it fits between the case wall and the hard drive chassis.
What Is My Target TDP?
You’ll also want to quickly list the components you’re looking at and check their TDPs. Add those up and compare it to the prices of the power supplies that support a PC with that power draw to see if the price is acceptable to you.
You can’t run a PC without an appropriately powerful power supply for long or sometimes at all. So, ensure you can afford a power supply from a reputable brand that will support the TDP of your PC and give you some overhead room for fluctuations.
The image featured at the top of this post is ©Maryia_K/Shutterstock.com.