It may come across as unusual to many, but in the evolution of computing, it would not be an exaggeration to say that the graphics card was borne out of the CPU and then gradually took up its own place as a separate PC component in its own right.
In the early days, when computers were not very powerful, the CPU did both generating instructions and rendering them parts. With time, however, as graphical user interfaces, or GUI, became more mainstream, the need was felt to create a division of labor. Thus, graphics accelerators, an early form of GPUs, were born.
Now, the CPU was responsible only for generating instructions and the job of the graphics card became to translate them into image information that was to be rendered by a display device such as a monitor. The rest is just continuous process improvements in the form of packing more and more transistors in smaller spaces that resulted in both the CPU and GPU becoming more and more powerful.
In this guide, we will cover all the important information that you need to know in order to make an informed purchase decision with your hard-earned money.
Why should you buy an external GPU?
You might have gotten the impression from the introductory paragraphs that CPUs don’t have any graphics capabilities, whatsoever, now. This is not true. Most CPUs, from both Intel and AMD, themselves contain integrated graphics too. These are powerful enough to allow you to effortlessly do most lightweight tasks like watching videos, emails, word processors, and perhaps even some video games.
An external GPU becomes important when your usage involves graphically intense workloads such as playing heavy duty video games, rendering animations on Blender, and video editing etc. While integrated graphics can perform these tasks too, since they are not designed for them, it would result in hellishly low framerates in video games, and rendering and video editing taking hours to do what a GPU can do in minutes or seconds.
Type of graphics cards
While any significantly powerful GPU can both be used for both gaming or professional usage, manufacturers today have begun to optimize graphics cards for each application. This is so that every bit of performance can be extracted from the card per unit watt for that particular use case.
Gaming
These are the ones that are most often mentioned when one is discussing graphics cards. If you are building a gaming PC, then these are the kinds of graphics cards that you should be comparing. Examples include Nvidia’s RTX 3060 or the RX 6600XT offered by AMD.
Graphics card is the biggest determinant of gaming performance, even more important than the CPU. Therefore, if you are primarily using your PC for gaming, this is the component that should take the biggest chunk from your PC-building budget.
Professional Applications
These GPUs have been designed with research professionals and industrial designers in mind who render complex scientific animations like molecular interactions or creating mechanical components in AutoCAD, for instance, respectively. These graphics cards are often much more powerful (and expensive) than those designed towards the gamers segment.
However, this does not mean that they can double as gaming cards too. This extra power does not necessarily translate into more refined gaming experience. Examples of such cards are the Nvidia Titan RTX and AMD’s Radeon Pro W6000.
This means that a higher price does not automatically mean a better graphics card. You have to be clear about what you want to do with your computer in the first place and then compare among the options that are geared towards your needs.
Most Important Specs
While every spec matters to some degree, the fact that there are so many of them means that it becomes overwhelming for any novice buyer to find the “best” option for themselves. For this reason, we have distilled down the list to three of the most important graphics card aspects that most concern anyone interested in building a gaming PC. These are architecture, core count and clock speeds, and memory. We will briefly discuss each of them, one by one.
Architecture
Architecture, as its name implies, refers to how a graphics card implements the instructions that it receives from the CPU. A more involved discussion on what exactly is GPU architecture is unnecessary and beyond the scope of this discussion. Therefore, we will limit ourselves to the more practical aspects.
First, the newer the architecture, the more powerful and efficient it is. Ada Lovelace and RDNA2 are the two latest architectures released in recent times by Nvidia and AMD, respectively. It used to be that AMD cards were much more power inefficient than Nvidia cards. However, in recent times, AMD has almost completely closed these gaps due to iterative refinements in their architecture.
Second, while empirically testing a graphics card for a particular game or application is the most reliable way to determine its effectiveness, this is not possible for everyone. Most buyers have to rely on technical specifications for their buying decisions. So, an apples to apples comparison of processor count, for instance, is only possible if the two GPUs are based on the same architecture. For example, it is meaningless to compare the number of cores in RX 6600XT and RTX 3060 to determine which one is faster.
Core Count and clock speeds
Both are important as they directly affect gaming performance among GPUs based on the same architecture. For instance, the higher the number of processors, the smoother the gameplay. Same with clock speed. However, each measure provides only half the picture. The floating point operations per second, or FLOPs, based on the product of core count and clock speed, provide a fuller picture and are a better measure to directly compare two GPUs from the same architecture.
Memory (size and bandwidth)
The size of VRAM (Video RAM) is very important when building a gaming PC. Memory is responsible for displaying texture details in the game environment. This means that when you are playing games with large virtual (more textures) at high resolution (1080p or more), the graphics card has to display more information simultaneously, which means greater memory size. For 1080p gaming, a graphics card with 6GB or more of VRAM is going to work just fine. For 4K gaming, get a card with at least 8GB of VRAM, although higher is more preferable.
Bandwidth depends on memory clock speed (separate from core clock speed) and determines how quickly the texture data stored in the GPU memory can be displayed. Both bandwidth and core clocks can bottleneck each other. Too small a bandwidth means the GPU will stay idle until all the data stored in VRAM is fully loaded. Too slow core speed means a slow GPU and memory clock speed becomes irrelevant.
The takeaway is that frame rates are largely determined by the GPU cores and clock speeds. Memory bandwidth of most GPUs available today is plenty and so your purchase decision should be mainly influenced by the GPU core count and clock speeds.
Other important considerations
Besides the fundamental specifications, there are other important factors that can influence the real-world performance of a graphics card. We will discuss them one by one.
Suitable CPU
It is very important that CPU and GPU should not only be compatible but shouldn’t bottleneck each other’s performance either. We will explain how. Some customers spend too much money on a very powerful CPU and then couple it with a not-so-powerful graphics card. The result is that the GPU is working at its maximum limits but the CPU still has some room, which means it is not fully being utilized. Also, since a graphics card does most of the heavy lifting in handling computer graphics, no matter how powerful your CPU, it can’t act as a substitute for the GPU for playing AAA video games or video editing.
Something similar happens when the positions are reversed. A weak CPU sends fewer instructions to the GPU to be rendered which means the GPU remains underutilized. Feeling lost? Fret not, for we have compiled multiple lists of the best CPUs and GPUs for a particular GPU and CPU, respectively.
Monitors
While fundamentally most modern day monitors and graphics cards have no compatibility issues as HDMI and DisplayPorts are featured on both of them. However, just because a monitor is compatible doesn’t automatically mean that it is suitable too. For example, if you couple your RTX 3080 with a 1080p display, you will never be able to play 4K games on that display, despite the fact that the 3080 can easily play them.
Also, if you are considering building a multi-monitor setup, make sure that your GPU has plenty of VRAM as that will ultimately limit how many displays you can connect with your card in parallel.
PC Case
A suitable PC case is important for two main reasons: fitting and cooling. With each new year, the graphics cards that are being released are getting bigger and bigger. The RTX 4090, at 4.8 pounds, is particularly huge. Forget about this graphics card if you are considering a small form-factor system.
A power rating above 300W is not unheard of in today’s GPUs. So, even if your PC Case can house your GPU, it may not be able to eject heat fast enough if airflow is disturbed due to congestion. To protect itself from overheating, the GPU will reduce its performance. This is known as thermal throttling. One expensive consequence of this frame rate drops. Therefore, don’t skimp too much on a PC case in a bid to get the fastest GPU.
PSU
Power supply is what powers everything in your computer except for the monitor. This is also quite important. GPU manufacturers mention recommended PSU sizes alongside the TGP of the GPU itself. An underpowered PSU means frequent system crashes and overheating as the GPU demands more power than the PSU can supply. Similarly, focus on quality too. Corsair, though slightly expensive, does have a reputation for excellent quality controls.
AMD vs Nvidia
The GPU market is highly competitive and with time, only two major players are left in this market, AMD and Nvidia. Intel too is making some headways with its Arc series GPUs released in early 2023. However, only time will tell if they can earn a spot for themselves among these two giants. For this reason, we will keep our focus limited to AMD and Nvidia for this guide.
Resolution (1080p and 4K)
Generally speaking, AMD cards are generally cheaper than Nvidia cards no matter what resolution you game at. At 1080p, it might even happen that the AMD cards perform better. However, at higher resolutions, generally you also want to have raytracing and upscaling enabled, you would be better off going with Team Green as we discuss below.
Upscaling (DLSS vs FSR)
Upscaling is a new technology and, in simple words, is AI based trickery that generates extra frames guessed from adjacent frames. The result is an overall much smoother gaming experience.
Nvidia’s proprietary technology in this regard is known as the Deep Learning Super Sampling Technology, or DLSS for short. The technology used by AMD is FSR which stands for FidelityFX Super Resolution. While Nvidia owns DLSS, FSR is open-source, which means that even Nvidia can choose to use it in their graphics cards, should they so desire.
Upscaling is an area where Nvidia leads AMD by a noticeable but not significant margin. The main advantage of DLSS however lies in the fact that it is supported by more games than FSR.
Raytracing
Raytracing is also very recent, at least for its use in commercial graphics cards. It is basically a simulation of the actual behavior of light, except that it happens in the world of pixels. The result is ultra-real and almost life-like gaming.
Here too Nvidia has the lead and by a significant margin. Even the best AMD cards have poor ray tracing performance compared to their similarly priced rivals from Nvidia. If raytracing is important to you, go with Nvidia without wasting another second.
Pricing and performance
In one of the earlier, we already discussed that a higher price doesn’t necessarily equate to better performance. The two parameters do correlate, but only up to a certain point after which diminishing returns set in. In real terms, this happens when you approach the $1000 threshold. For example you can get the very powerful RTX 3080 for around $700. Or you can spend another $500 (or more) on the RTX 3090 Ti and get a 30% increase in effective speed. In other words, even for advanced gamers, most sub-$1000 cards are going to be more than adequate for their needs.
Conclusion
We have covered a lot in this guide. We will briefly recap the major points of discussion in our guide. First we talked about the main motivation for buying an external GPU in the first place. Then we discussed the types of graphics cards and their use cases. After that we discussed the most important specs and factors that influence GPU performance. Lastly, we compared AMD and Nvidia in areas that matter the most. To, at least somewhat, reduce the complexity of your buying decision we briefly talk about three highly rated graphics cards from different price brackets that you might want to consider.
Some highly rated GPUs
RTX 4090
This is the fastest GPU on the planet today, at least for the gamers community. It comes in at $1600 and features 24GB of VRAM. You can even play 8K games at high settings on this card, or you can attach up to 4 4K monitors if you. However, keep in mind that it is also very power hungry and is going to require a large PC case for adequate housing and cooling.
RTX 3080
At $700, this card is priced right at the point where the price-performance curve begins to flatten. Every additional dollar you spend is going to provide you with lower and lower performance returns. While the slightly cheaper AMD RX 6800XT offers similar performance, the differences begin to get more prominent when ray tracing is enabled. Highly recommended!
RX 6600/RTX 3060
To dispel any impressions of biasedness towards Nvidia, we decided to include one of AMD’s highly rated offerings to the mix too. Both these cards come in at $330 and are excellent options for 1080p gaming. Unless ray tracing is important to you, you can go with either of these cards.