Choosing Right GPU: What Truly Matters?

Given the wide range of Graphics Processing Units available out there, it is not easy for an average user to make the right choice without knowing what really matters and what specific things to look into before making the final decision.

This specific article will help you a great deal in that matter, making you more knowledgeable about the specific requirements of a good GPU and the limitations to avoid.

KEY TAKEAWAYS

• Check for the clock speed since higher the clock speed, the better will be the performance but make sure it is compatible with the internal structure of the GPU.
• Check the memory bandwidth of the GPU because the higher it is, stronger the GPU will be and be more able to draw more input resulting in faster performance.
• Depending on the type of computing job you do and the games you play, choose a GPU with higher VRAM.
• Also choose from the more modern type of GPUs such as the GDDR5, GDDR6, GDDR5X or the HBM of AMD instead of the older GDDR2 and GDDR3 variants.
• Check the cooling fans or radiators as well as the type of coolers to ensure that the GPU does not get excessively hot while operating.

Knowing More About GPU

The Graphical Processing Unit or the GPU is the specialized component that handles graphics in all the modern gaming, productive, and overall mid to high ranged performance PCs.

It is a similar silicon-based chip like the CPU, mounted on a PCB circuit board but doesn’t have all the added functionality that a CPU a does.

What it has are hundreds of multiple smaller cores, with a lower clock rate than a CPU core, that only process graphics at a much faster rate. A GPU has much better bandwidth and memory speed than a CPU does.

Usually, a GPU is mounted on a graphics adapter or simply a video card that has other parts like the circuit board, PCIe connecting headers, a cooling fan, etc.

With the development of the PCI-Express, older forms of connection like PCI and AGP aren’t used anymore because they have obvious advantages. Learn more about PCIe.

Graphics cards come in various sizes and budget ranges. This is because of the difference in performance, and the number of cooling fans attached to it.

Entry-level cards like the AMD Radeon RX 480 can make do with a single fan, but extreme ones like the NVIDIA RTX 2080 Ti might require three of these for effective cooling.

The more powerful the GPU is, the more heat is created, and hence more cooling is necessary.

Therefore, the case and motherboard size come to play here. Check out the differences between desktop and laptop GPU.

Some people have trouble installing GPUs because of fewer PCIe slots or non-adjustment with RAM modules.

A bigger motherboard gives you more PCIe slots and a more general headroom and this is why the ATX and Micro ATX cases can easily hold the larger GPUs.

The major and easiest way to differentiate a GPU and a CPU is the way how each works.

The CPU can handle multiple tasks at the same time, doing calculations and complex processing with 4, 6, 8, or whatever the number of cores it has.

The GPU with numerous smaller working cores is more like a specialist in graphics but an amateur in other tasks.

Using parallel processing, it can handle graphics much better than a CPU and all of its cores are dedicated only towards this goal.

As a result, anything related to graphics is effortlessly handled by the GPU because it needs to process simpler graphics instructions and has:

• More cores
• Parallel Processing
• Better memory speed
• Faster VRAM

So if a GPU is stronger than a CPU, why don’t we use it for other applications?

You see, a processor can be compared to the wardrobe, and as such, it can store and process various kinds of data.

A GPU, on the other hand, is a shoe rack that is developed only to provide a better graphics performance in holding a shoe, or getting rid of the analog, the graphics.

You cannot replace the CPU with a GPU as it won’t be able all the other types of processing as efficiently.

Moreover, it consumes more power as well so that is also a problem.

Having said so, we may not use it on a large scale, but General Purpose computing on GPU or GPGPU is a concept that is being implemented in further complex forms of processing that can utilize the more working cores of a GPU.

So let’s leave each of the CPU and GPU to let them do what they do best, and find out what should be taken care of while buying a GPU for your purpose, whatever that may be.

GPU naming:

For NVIDIA

In this case, the company uses the sub-brand GeForce to name the affordable consumer-grade GPUs, and there may or may not be a letter after this. For example, low-end cards like the GeForce 520.

The terms GT and GTX are used for performance levels, and the later ones are the most common highly powerful cards that are mostly available in mid and low prices, but there is an exception with the 1080 series.

The RTX cards are even better, and support Ray Tracing technology which provides even better graphics and faster processing with GDDR6 VRAM.

Then come generation and numbering related to a GPU name.

There have been many iterations of GeForce GPUs, but those are a legend now. Currently, there are 7^th, 9^th, 10^th, 16^th, and 20^th series GPUs present. Let us take the GeForce 1650 for an example:

• The 16 refers to the series of the GPU, and higher is better.
• The next two numbers refer to the performance level, and higher is better here as well.
• Up to the 9^th series, the first number refers to the generation, and the next two digits indicate performance level.

There can also be suffixes, each with its meaning:

• No suffix means that the GPU was the one launched first in the series.
• Ti refers to better performance and steeper price
• S or super means that the card is recently launched, and its performance and pricing sit somewhere between Ti and suffix-less cards.

For AMD

AMD naming is similar as well, but the sub-branding is named Radeon here. This is followed by R, and the letter after that indicates a series, like RX, Vega, and Fury.

But matters can get confusing since it has changed the naming scheme multiple times in the recent past, although naming like the Radeon RX 5700 XT is said to be followed in the future. The brand uses the 500, Vega, and 5000 series of GPUs in its main line-up.

So following the “RX” or “Vega” is a number sequence, decoded in the way below:

RX 590: The first digit indicates the generation of the processor, while the last two indicate performance level. Hence, it is better than an RX 560.

There are relevant suffixes like:

XT– High performance

M– Mobile GPU

XTX/ XL–  GPU based on Polaris Architecture

But although this seems simple, other names are highly awkward. For a layman, there is no way to know the Radeon VII compares to which other GPU of either the same brand or its competition.

AMD or NVIDIA

With ATI out of the scene and Intel only good regarding integrated graphics, the brand preference narrows down to two choices, NVIDIA and AMD.

Each of these are established brands, while AMD is giving tough competition to NVIDIA after its major comeback in the market a few years ago. Before you dive deep into any other details, deciding the brand is necessary.

The fight between AMD and NVIDIA has been going on for quite some time, and this is very natural as these two are the only GPU manufacturing giants currently.

There can’t be a clear winner as each has a place of strength, while weaknesses as well which the other takes advantage of.

For example, the AMD 5500 XT GPUs may be cheaper but are not as efficient as the NVIDIA competition, 1660 which are slightly slower but are power efficient.

Same is the case for the Radeon VII which performs as good as the NVIDIA GeForce RTX 2080, and better than RTX 2070, but requires more power. Also, the AMD GPU doesn’t support DirectX 12, which places it at a disadvantage.

NVIDIA and AMD can be compared on many grounds which is a topic of its own. However, we can consider some right now:

• Budget

AMD has been catering to the needs of budget buyers since the very beginning, and although NVIDIA has brought some alternatives to battle in the budget and mid-range segments, AMD is still better for someone looking for affordable GPUs to build a cheap system.

The RX 590, 580 are decent cards that may be old but are still a better choice over NVIDIA in this range.

The mid-range availability of both the brands is the same, and although NVIDIA has more options, the RX 5500 and 5600 cards from AMD are still as good and can go against the RTX 2080 or GTX 1080 type cards.

Overall, AMD seems to deliver you more performance at a given price range, but don’t make your decision just yet.

• High-end Performance

Now top tier performance is somewhat dominated by NVIDIA, as only the Radeon RX 5700XT is almost as good as the RTX 2080 super with its overall performance lagging behind about 30%.

The 2080 Ti is the absolute best in this tier, yet. AMD hasn’t been able to launch cards with Ray Tracing yet, but reports come in that those will be launched very soon, while the NVIDIA 3000 series of powerful GPUs aren’t far either.

But Ray Tracing is only good for effects, right? The extras shadows and detailed textures are great to look at, but performance differences aren’t constituted in this only.

But some RTX cards outdo any AMD card in this price range.

So Based on this, NVIDIA is still better for professionals in the higher segments, and their higher price tag is somewhat justified due to the raw performance they offer.

• Connectivity

GPUs need a connection to the monitor, and in the matter, AMD offers relatively cheaper FreeSync while NVIDIA’s Gsync has been a veteran. FreeSync supported monitors are more in number, cost-effective, but lack some of the extras that GSync provides.

But for most Gsync features, a comparable AMD answer can be found, which causes a deadlock in many matters.

But, NVIDIA is still better at smoothening scenes even at lower frame rates, while AMD is not that good enough when frame rates drop below a monitor’s refresh rate.

Both of these adaptive sync technologies are good enough, but again it is a matter of perspective.

The monitor’s refresh rate is matched with that of the GPUs frame rate, to handle screen tearing in both the brands.

But NVIDIA seems to be better and more reliable as it has better quality control than AMD.

For a casual user, all of these shouldn’t matter much since you are only interested in a hobby or entertainment of some sort. But for those serious about their PC’s performance, the question lingers.

The answer is complicated and yet simple. With a larger budget, nothing can be better than NVIDIA, but with restricted financial ease, it is better to choose AMD as it is equally reliable but lacks some extra features that the opponent offers.

Internal Factors that You Should be Aware of

• Clockspeed

For a particular GPU, the more clock rate (calculated in MHz) it can achieve, the better would be the performance. But this is not the only deciding factor and depends on various others before a conclusion can be drawn.

For a specific task, it has been observed that a GPU with lower clock speed with a different internal architecture performs better than one with a higher speed.

In many cases, the GPU can be boosted or overclocked to increase the clock rate.

But overclocking is subject to certain risks, and not all are interested in it.

Also, a card with a higher clock rate should have an effective cooling system, or else there would be problems.

• Memory Bandwidth (Gbps)

Memory bandwidth is one of the main factors, which decides the strength of a GPU.

It creates much difference in the data transfer speeds, and the more bandwidth the GPU has, the faster it can draw raw input, thus leading to faster processing.

It depends on the data bus size, and the clock speed of the GPU, and while calculating it is not much of a problem, the process is different for GDDR5 and GDDR5X types.

Modern GPUs can have a bus width anywhere from 32 bits to 384 bits, and this depends on the model and make.

The common GDDR5 type has the lowest, while GDDR6 models have the highest bus width and the most bandwidth.

Now the GPU with a bus width of 128 bits will have twice the bandwidth of the one with a 64 wide bus, given the clock speed are both in the same, and so you know which to go with.

• RAM size

You must know that some GPUs have 8 GB of VRAM present. But can you, as an average user utilize all of it? Would it be wise to spend all that extra money for the extra RAM?

This depends on how you intend to use your PC or laptop.

See gaming can be done both on 2 GB of VRAM, and 8 GB of it but the difference is that a GPU with so much RAM isn’t a lower-tier one, and in fact, it can achieve a lot more than just simple gaming, for instance, 4K or VR gaming.

In a particular game, the VRAM used for running it at a lower resolution is less than what would be used if you play at the highest settings available.

But having more RAM is beneficial in general since the GPU gets more data to work with as fast as possible. If you use a high resolution or more than a single monitor, you can use most of it.

Keep in mind that adding two or more similar GPUs (if possible) in your PC wouldn’t increase the VRAM, for example, if you use two 4 GB graphic cards, the usable VRAM is still 4 GB and not 8 GB.

What does happen is that the rendering capacity of the system gets a major upgrade, and this is possible by Crossfire Technology as named by AMD, and something called SLI coined by NVIDIA.

Both are fundamentally similar, but AMD crossfire allows you to connect two or more video cards with different GPUs, RAM, memory until they are based on the same architecture, but SLI only works on two exactly same GPUs with the same amount of RAM.

The choice of vendors is flexible in both cases. They do have other differences, but those aren’t of so much relevance here.

• GDDR5, GDDR6, GDDR5X or HBM?

When buying a GPU, you might be faced with confusion about the memory type present in it.

The older GDDR2 and GDDR3 are all but obsolete now, so the main types of memory you will find are GDDR5, GDDR6 and GDDR5X, and AMD’s HBM memory types.

You do not have to make a direct choice as every GPU comes with its memory type that cannot be changed, but for a comprehensive detail, here’s what they mean:

GDDR5

The GDDR5 is the most used memory type that has been used by manufacturers for over a decade now, and it is found in many video cards today.

Double Data Rate 5 Synchronous DRAM to be exact, they come in different sizes, starting from 512 MB to 16 GB.

The highest data rate you get is about 7-8 Gbps, but the many varieties make it the only choice in budget GPUs.

DDR5X

The DDR5X was introduced in 2016 by NVIDIA for their flagship GTX 1080 GPUs at that time, and are still used by these along with some Titan GPUs as well, but these are very few.

It is better than the GDDR5 memory type, with data rates of about 12-14 Gbps.

But this is an expensive memory type, which might explain why the GTX 1080 and 1080 Ti are so expensive even though the newer generation of video cards has been launched.

GDDR6

Next, Double Data Rate 6 Synchronous DRAM or GDDR6 is available in only 8 and 16 GB sizes, which is why it is used by high-end video cards.

The data rates are surely higher, at 16 Gbps as dual memory channels are employed.

However, unless you have a major performance requirement, these won’t be worth it.

HBM

In the HBM or High Bandwidth Memory manufacturing, the actual memory is stacked vertically, on top of logic dies, and very close to the GPU itself which increases performance.

These GPUs are very small, thus take smaller space than the other memory types and consume less power.

Although the current HBM2 or HBM2E may be faster than the GDDR5 in terms of clock speed and wider bus width, there is only one series in which AMD uses the technology, the Fury X.

Also, with GDDR6 at the scene, it is losing its importance and will soon be replaced entirely.

• Graphics Card Size

The size of the graphics card is also something that has to be considered before you buy one.

You see, some are smaller in size and hence fit on the micro ATX motherboards as well, but those with a triple fan setup, the RTX 2070 Ti, for example, are massive, and you must have enough space on your motherboard so that it fits, along with enough PCIe slots.

You wouldn’t like to spend a lot of money on a GPU and then realize that it’s too large for your motherboard, would you?

To be on the safe side, it is better to go with the micro ATX motherboard which has its own set of advantages when it comes to matters like these.

If indeed you need a smaller system and would be using a mini ITX board, make sure you know whether the GPU will fit in the concerned motherboard.

Some of which fit such smaller motherboards are the Zotac GeForce GTX 1660 and MSI GeForce GTX 1050 Ti among others.

• Cooling

All external GPUs need cooling, and this is why cooling fans or radiators are provided with them already.

Without efficient cooling, it will not be able to perform at its best for a longer time, and the excess heat might as well damage it when left unattended.

The various vendors solve this problem by providing the right number of fans with every GPU, but it is upon you to check whether it is getting too hot.

Many users also install liquid cooling which does an overall better cooling job and is suitable for overclocked systems but is expensive.

There are mainly two types of coolers, one with the open fans where the heat sink is visible, while others where the heat sink is covered by a plastic sheath.

Now the first kind is better for roomy cases that can provide good airflow, as the vents on the graphics card pass hot air through the heat sink and then release it inside the case, which can adversely affect other working components if there isn’t enough air to carry it outside.

You may even add an extra case fan so that there is ample intake and outflow.

The second kind or blower type of coolers release heat directly outside the case and is thus ideal for compact cases.

The removal path of heated air is towards the back of the card, and hence the system is kept cool even though a lot of airflow is not available. 

What Kind of PC Are You Building?

The next section deals with your requirements and demands.

The first of them is the budget, both of the GPU itself and the one stipulated for the entire build (if, of course, you are building a PC and not buying only the GPU).

Will you be gaming? Or do you need the GPU for media editing and processing? There might be some of you who will need to employ it on a workstation rig.

Each has its standard of performance, and while a high-end GPU can solve all of the above, a low-end one cannot handle even one of these.

Budget

For simplicity sake, let us consider that the viewers here have 4 main budget preferences, and this can be categorized as follows:

• Budget buyers

The buyers who have a strict budget, or are looking for an entry-level GPU for their gaming rig would be right to choose AMD options like the Radeon RX 580, RX 580, etc. NVIDIA alternatives would be GTX 1050, GTX 1660, and such within the $200 mark.

All these cards are built for budget concerned buyers, and if you aren’t into 1440p gaming or such, these should be just fine.

The max memory capacity is of 8 GB, but as we said a lot of things matter more than just this. The GigaByte 5500 XT with PCIe 4.0 is a must that you should check on this list.

• Mid-range buyers

Now some better GPUs, in about $300 that can run smoothly at 1080p and 1440p at decent frame rates, and even some at 4K.

See the previous ones might come cheap, but the ones under this category provide you the most value for your money and should be best for a majority of you.

Many options are present, like the AMD Radeon 5600 XT and NVIDIA GeForce RTX 2060 Super, GTX 1070 Ti, and GTX 1080. NVIDIA provides you with more alternatives in this range.

• High-end buyers

The best GPUs that excel over those present in the previous two segments are available here, and you can easily run stuff at 4K, be it gaming or editing.

AMD is yet outnumbered, but the 5700 XT gives a tough competition to its NVIDIA adversaries at a much cheaper price.

NVIDIA has the GTX 1080 Ti here, along with other RTX cards like the RTX 2070 Ti, 2070 super, 2080 Ti, and such. You can choose any of these, but all are supreme performers and have even 11 GB of graphics memory, and yes, DDR6 type as well.

• Premium buyers

GPUs at the extreme level would be of no use to the average content creator or gamer, and these are only suitable for those interested in complex 3D architecture, rendering, and maybe if someone is planning to build a workstation-grade PC at home.

AMD fan base would be disappointed here since only NVIDIA has such powerful GPUs at the moment. This includes the Titan series, and some worthy mentions are the Titan V and Titan RTX.

Other System Specs

A good graphics card is futile if there aren’t other good enough parts to support it. This includes the processor, RAM, motherboard and SSD storage.

For the processor and motherboard,  they must support the GPU and there should be enough PCIe slots.

The PCIe generation is not of much consequence if you aren’t a performance freak, though only the X570 motherboards support gen 4.0.

Moreover, the PCIe interface being backward compatible, the only thing you need to make sure is that the motherboard has enough space and PCIe ports to house the GPU.

What needs to be taken care of is a bottleneck, and that can arise when instances like this arise:

• Using the latest RTX GPU with an i3 or lower CPU – This is the worst kind where the CPU bottleneck takes place, and the performance is hindered. The GPU is simply too fast for the CPU.
• Using a GT 1030 GPU or lower GPU with a latest i7 CPU– This is desirable as the GPU can be fully utilized, but in reality, the GPU simply isn’t strong enough to return data at the same rate as the CPU does.

Next is the system RAM. Although every GPU has its RAM to work with, there must be a minimum amount of RAM present in the system for it to work. Too little RAM and the PC’s performance will degrade.

Do You Have the Right Monitor?

The graphics performance would only be visible if you have a monitor that has good enough refresh rate and high-resolution support.

Like, even though you render a 4K video on your PC, you can never see how good it is on a 1080p monitor.

4K monitors may come costly, but yes, there are some affordable ones present that you can buy even after spending a lot on the main build.

The GPU itself must have HDMI or other relevant video ports so that it can be connected to a TV or monitor.

The benefit of having a decent GPU is that you can use multiple monitors at the same time without trouble.

VR

Gaming on a VR headset makes the experience as real as possible, but it doesn’t come cheap.

Surely playing games like Walking Dead, Assetto Corsa and others are very exciting when you strap in a VR headset, but these require a minimum system specs list and a GPU most importantly.

RTX GPUs are the ones that first come to mind, but some others support VR gaming as well.

For this, 8 GB of system RAM is the bare minimum, along with an i5 processor or AMD equivalent.

The basic GPU requirement is the same for both laptops and desktops, and that is a GTX 1060, as can be seen in the official NVIDIA website.

But different games have different needs, and not all will run in this configuration.

Moreover, it is recommended to have more power than the minimum requirements, no matter what sort of software.

Is Overclocking Your GPU Worth It?

Overclocking is a matter of personal preference, and while overclocking your GPU may give some performance boost, it will reduce its longevity, to a great extent.

Overclocking will also increase power consumption and heat generation, so it is necessary to have superior cooling and a good PSU ready to counter that.

For the performance boost, you may see a 15% increase at most, and some better frame rates in games.

So it is worth it when done safely and provided that it doesn’t end up doing more harm than good.

Some Tips Regarding 3^rd Party Vendors

The main GPU designs are made by the manufacturing brands, AMD, or NVIDIA, but are sold by various vendors who customize the reference designs by adding more colors, fans, and other things that influence the buyers to choose a particular brand.

All of these vendors, like EVGA, ASUS, MSI, GIGABYTE, etc to name some are proven to be reliable, but while some have poor customer support, another might produce dull products that don’t exactly give off the aesthetic vibe.

Comparing these directly would be foolish, but here are some tips that you can apply so that false marketing strategies don’t get the best of you.

• Look out for more VRAM

It might be to influence your decision, but some vendors provide some more VRAM than other brands for the same model.

If the price difference is reasonable, you may choose such since more VRAM would mean that the given GPU is going to perform better.

There is no way to increase the video RAM of a particular video card once you have bought it, and this has to be considered.

• Always compare the prices

One of the basic things is to compare several brands before buying a GPU. For example, for a particular model,

ASUS might be charging more than MSI and if you aren’t careful enough you might spend more for no performance improvement at all.

Make sure the extra lighting or color design is worth spending more.

• More fans aren’t always good

GPUs indeed require cooling, but it doesn’t need to have as many fans as can be fit.

Every reference design has a stipulated number of coolers, but the vendors may add more, and one might end up thinking that just because one offers 3 fans instead of 2, it is better.

While the actual case is that the 3^rd fan might not be required at all, and only increases the length of the GPU.

When more cooling fans are running, be sure that there would be more noise too.

• Check out for faster versions

Some GPUs are available with their overclocked versions (OC).

If possible, choose those since they offer slightly more than the stock clock speeds and might save some personal effort from your side, since you don’t have to do the boosting part yourself.

These are done by the vendors themselves, and shouldn’t harm the warranty either.

However, in most cases the overclocked versions are not as good as manual overclocking, and the GPU is only boosted to its default level.

• Warranty

The ultimate thing to know before finalizing a brand is the warranty period it offers on the video card.

Some companies provide a year or two of warranty, while others offer a lifetime warranty.

Some go even further and promise warranty even if you overclock the GPU. The better offer you get, the more you benefit.

Conclusion

The GPUs are a very interesting kind of hardware. These are always the most attractive part both in looks and performance, and it is always nice to customize one’s PC.

The variables are many, but it isn’t difficult to buy a new GPU. We hope that this article would serve as a handy reference and ease your decision making.