Processor

The Central Processing Unit, commonly known as the CPU or processor, is arguably the most important part of any computer. It is the brain of the computer, and executes the instructions that make up any given application, including the operating system. There are currently two main manufacturers of desktop processors — Intel and AMD.

Is the socket type important?

Modern processors are connected to the rest of the computer by insertion into a socket on the motherboard. Therefore, the socket type of your chosen processor will limit your choice of motherboard, as they must match.

What does multi-core mean?

Not so long ago processors only had a single core — or rather the idea of having multiple cores did not exist. Today, however, nearly all available processors are multi-core, and have anywhere from two (a dual-core) to 64 cores or more, most often in a power of two (2, 4, 8, 16, etc).

Software must be specifically written to take advantage of multi-core technology, and when it is, it is called a multi-threaded application. These applications may take advantage of only two cores, or four, or more, regardless of how many cores you actually have. For example, if you run an application that can use up to three cores, and your processor has six, then three cores would potentially be sitting unused. It is not that simple though, we rarely run a single application at a time — usually there are many other applications running in the background, minimized, or on other monitors, as well as the operating system processes, all of which can utilize those “extra” cores and relieving demand on the ones used by that application.

So how many cores you need? If you plan on running demanding applications such as video games or video editing software, it is best to check the system requirements of those applications to see how many cores they recommend — and then you may want to go one or two cores more than that. If a particular video game indicates it can use up to four cores, then maybe you want a six-core processor. If you do not plan to run any demanding applications, a quad-core processor will likely be enough.

Hyper-threading

Normally a single core can only process a single thread, however, some processors have a feature called hyper-threading which allows two threads per core at the same time. This does not allow a processor to operate twice as fast, as one might think, but it does allow each core to run more efficiently, switching to the second thread when the first would otherwise be idle. Be wary that some applications such as image editors may actually run slower with hyper-threading enabled, while video games typically see a small improvement. Hyper-threading can be disabled in the bios.

Is GHz a measure of speed?

Processor clock speed, also called processor frequency, is a measure of how many cycles the processor executes per second, measured in GHz (gigahertz). Generally, a higher clock speed means a faster processor, but only when comparing processors of the same brand, generation/architecture, and number of cores. A processor of an older generation/architecture may be outperformed by a newer one with a lower clock speed, because it’s architecture is more efficient. While clock speed should be taken into consideration when comparing processors, it is not the only measure of speed.

With some processors, it is possible to increase their clock speed beyond what they are rated for — this is called overclocking, and has the potential to make a processor run faster at the expense of increased power consumption and heat output. Intel processors with a K or X suffix and all AMD Ryzen processors are capable of being overclocked, as long as you have a motherboard that supports overclocking. Because of the potential of overclocking to destabilize or damage your hardware, usually only advanced builders consider it.

Cache — memory inside the CPU

A small amount of extremely fast memory built directly into the chip, the CPU cache stores the data most frequently used by the processor to speed up computation. A processor has three types of cache, L1, L2, and L3. You don’t need to know much about these other than more is better, and all else being equal, a larger cache will see better performance. Modern processors generally have adequate cache, so there is no need to focus on this unless you are seeking extreme performance.

Integrated Graphics Processing Units (iGPUs)

Some processors have a graphics processing chip built right into them, eliminating the need for a dedicated GPU. An integrated GPU will only provide basic graphics output, and should only be considered for a PC that will be limited to basic tasks like word processing and web browsing, or entry-level gaming. Read the “Selecting Features” section about graphics cards to learn more.

Take Away: Make sure you have at least the minimum recommended number of cores for the applications you will be running. Take into consideration the clock speed, but remember that it can only be properly compared between processors of the same brand, architecture, and number of cores. Hyper-threading may improve performance depending on what type of applications you are running. More cache is generally better, but is a lesser detail only concerning enthusiasts. Integrated graphics are useful for low-end builds, or as a hold-over while saving money for a dedicated GPU.

Motherboard

The motherboard, sometimes shortened as “MOBO”, is the main component that all other parts connect to. It allows communication between the processor, memory, graphics card, and storage devices, as well as external connectors such as usb, ethernet, video and audio ports. There are around 15 different manufacturers of consumer motherboards, the most popular being ASRock, ASUS, Biostar, EVGA, Gigabyte, MSI and Intel.

Compatibility

Because the primary job of the motherboard is to connect all of your other components together, your most important consideration will be making sure that it is compatible with each component you plan to use.

As mentioned in the Processor section, your processor and motherboard must have the same socket type. So if you choose a processor with an AM4 socket type for example, then make sure the motherboard supports it. Furthermore, even when the socket type matches, a given motherboard and processor may not be compatible for a variety of reasons, so make sure to check the compatibility specifications of the motherboard.

Next, the motherboard’s form factor needs to be compatible with your case. This is generally only a concern with smaller cases and usually larger cases will support any form factor that will physically fit inside. Read the “Selecting Features” section about form factors to learn more. Check the compatibility specifications of the case to be sure, and if you’re specifically interested in building a small PC, it’s best to select a case first, as that will restrict your choices.

One last note on compatibility — if you plan to overclock your processor or memory, beware that only certain motherboards allow this. With AMD AM4 boards, look for B#50 or (generally higher priced but more generously equipped) X#70 series chipsets; for Intel LGA 1200 boards, look for H#70 and Z#90 series chipsets.

Memory

Like with the processor, which type of memory (RAM) you can use will be determined by what your motherboard supports, so reference the motherboard specifications. Memory modules install into memory slots, which come in two types: DIMM, generally used for desktop PCs and SO-DIMM, which is smaller and generally only found in laptops or very small form factor motherboards. The slot type doesn’t matter for performance, just make sure you buy memory modules that match.

Then there is the number of memory slots — motherboards usually have anywhere from two to eight. All else considered, it’s best to opt for more memory slots rather than fewer — spreading the memory across multiple slots can improve performance, and having extra, unused slots available allows for additional memory to be added later. Small form factor motherboards will inevitably have fewer memory slots than larger ones.

Each motherboard also has a total limit of how much memory it can support, typically something like 32, 64 or 128 GB, but occasionally lower, so again, make sure it will support the total amount of memory you plan to install.

Expansion Slots

Expansion slots are where you plug in components such as a graphics card, audio card, network card, or any number of other optional components. There are two types, PCI and PCI-Express. PCI is an older standard that is often supported on modern motherboards because you can still buy older model components that use PCI, or you might have an old PCI network or audio card that you want to use.

All newer components now use the PCI-E standard, which comes in four formats: x16, x8, x4, and x1, where the number denotes the quantity of “lanes”. PCI-E cards can be plugged into the same size or larger slot, meaning that a x4 card will work in an x4, x8 or x16 slot, so there is no need to hunt for a motherboard with a PCI-E x4 slot just because you have a PCI-E x4 card. Similarly, if a motherboard has an open-ended PCI-E slot (a slot which will accept cards with more lanes than it has), you can plug in a larger card and it will work, but it will be limited to the speed of the slot.

Make sure that you have enough of the correct expansion slots to add whichever components you are planning on buying. If you are buying a dedicated graphics card, you will likely need at least one PCI-E x16 slot. Sound cards typically use PCI-E x1. Some solid-state drives connect using PCI-E slots.

Integrated Features

A wide array of integrated features are available. Like with processors, motherboards can have “onboard graphics”, meaning they have an integrated GPU built in. If you are not planning on buying a dedicated GPU and you didn’t choose a processor with onboard graphics, you’ll want this. If you are buying a dedicated GPU, and you are interested in doing SLI/crossfire, you will need to make sure the motherboard supports it. Read the “Selecting Features” section about graphics cards to learn more.

Nearly all motherboards come with onboard audio, meaning they have an integrated sound card. The key thing to look for here is how many audio channels are supported — if you have a 7.1 surround speaker system, then you likely want 8 channel output, if you only have two speakers, 2 channel audio is fine. Most users will be perfectly happy with onboard audio quality, generally only musicians or audiophiles will opt for a dedicated sound card.

Onboard Ethernet (LAN) support is standard, but WiFi is not. If you need WiFi, opting for a board with it integrated is a good idea, it generally doesn’t cost much more and its less hassle, but you can also add WiFi support with an inexpensive PCI, PCI-E, or USB adapter. Similarly, most boards do not come with Bluetooth support, so if that is something you want make sure to check the specs or add a Bluetooth adapter to your build.

Internal Ports

Internal ports are the connectors on the motherboard that all the other internal components connect to like motherboards, fans, and disk drives. PATA, SATA, SATA Express, mSATA, M.2, U.2, and SAS are all connectors for storage devices (also remember that some SSDs connect via PCI-e slots as well). Make sure you have the right connections (and right quantity) based on which drives you are buying, or already own.

Fan headers, unsurprisingly, are for connecting fans — don’t feel limited if a motherboard has few of these, as you can use splitters, connect fans directly to the power supply, or buy a fan controller (some cases come with a fan controller). The advantage of connecting fans to the motherboard is that your system can control the fan speeds, increasing or decreasing according to the current load. RGB headers are similar but are meant for connecting lighting for purely aesthetic reasons.

That brings us to the elephant in the room: USB headers. USB headers are where you connect USB ports to your motherboard, usually from your case but sometimes added via expansion slots. USB 2 headers are straight forward, these are for your standard “Type A” rectangular USB plugs with a black or white plastic strip in the middle. USB 3 headers are for either rectangular Type A plugs with a blue plastic strip or oval Type C plugs, but are also backwards compatible with USB 1 and 2 devices.

USB 3 comes in a few different flavors which each have a few different names, and it can get confusing. USB 3.0, USB 3.1 Gen 1 and USB 3.2 Gen 1 are all the same thing, with the latter being the most recent rebranding of the standard, have a maximum speed of 5 Gbps and support both Type A and Type C USB ports. USB 3.1 Gen 2 and USB 3.2 Gen 2 are the same thing, have a maximum speed of 10 Gbps and support both Type A and Type C USB ports. USB 3.2 (without generation) and USB 3.2 Gen 2×2 are the same thing, have a maximum speed of 20 Gbps and only support Type C USB ports. (Yeah, thank the USB Implementers Forum for this confusing naming!) If your case does not have any USB 3 ports to connect, you don’t need a USB 3 header, as you will have nothing to connect to it!

External Ports

If your motherboard has any integrated features like onboard graphics or audio, make sure the ports that are exposed match what you need. For onboard graphics you’ll typically have DVI, HDMI or DisplayPort, or a combination of those (note that some onboard graphics may only support a single monitor even if multiple ports are exposed). For onboard audio you will find a varying number of 3.5mm ports depending on how many channels the card supports, and sometimes you might find a digital optical output, which you will likely never use unless you have a high end audio setup.

Apart from that, the only built-in ports you need to worry about are USB ports. Simply make sure the board provides enough USB 2 and USB 3 ports for your needs (you’ll want to take into account how many extra ports the front panel of your case has). A good rule of thumb is at least two USB 3 ports, and two more USB 2 or 3 ports, but it’s not a huge deal as you can always add more later with an expansion card. If a case has no USB 2 ports, that’s fine as USB 2 devices can be used in USB 3 ports without issue.

Take Away: The motherboard is what connects all the other parts of your PC together — therefore you need to carefully choose one that is compatible with each part. It should have the same socket type as your CPU, a form factor that is compatible with your case, and the correct type and number of memory slots. It should have enough expansion slots of the correct type for whatever cards you will add. Decide on the integrated features according to whether you need onboard audio, video, Ethernet, WiFi, or Bluetooth. Make sure enough internal and external ports are present for whatever drives, fans, lights or peripherals you plan to connect.

Memory

Random-Access Memory (often referred to as “RAM” or just “Memory”) is the systems’ short-term memory, storing data for the processor to work with whilst it conducts computations. Nothing can be stored in RAM once the system is powered off.

Like most components, RAM comes in different specifications, and the currently used specification is ‘DDR4’. RAM comes in the format of one or more small circuit boards with memory chips mounted on them called a “DIMM”, and often referred to as a RAM ‘stick’. They are sometimes covered with a metal or plastic covering called a ‘heat spreader’. The ‘stick’ of RAM connects to the rest of the PC via insertion into a DIMM slot on the motherboard

All RAM is produced by a handful of manufacturers: Samsung, SK Hynix, Micron and Nanya cover the majority of market share. Other companies, sometimes subsidiaries of these manufacturers, buy the memory modules and build them into the sticks which are sold commercially. So whilst you will choose between G.Skill, Adata, Crucial (subsidiary of Micron) and Corsair, you may in fact be buying the same underlying chips.

Quantity

How much memory you put in your PC is important because if you don’t have sufficient RAM, the operating system will be forced to use the much slower storage drives to store working data, causing the running programs to swap data back and forth as they require it. This can induce slow-downs. Ideally, all active program data and documents can be held in RAM for very fast access. Fortunately, the most common RAM quantities are more than sufficient for general and all-round use.

8GB of RAM is sufficient for most games and general use such as document creation, web browsing and media consumption. You will not be able to undertake more complex tasks like video editing or multi-tasking without significant slowdowns at times. Some games require 16GB RAM or more to run smoothly. If you’re on a tight budget you can make do with 8GB.

16GB is sufficient for all normal use including almost all games, complex office work and basic productivity like simple video editing, photo editing or multitasking. 16GB is the sensible choice for most PCs.

32GB becomes necessary for some memory-intensive games like simulators or complex strategy games. It permits most forms of video editing including effects processing and complex content creation such as 3D art or Computer Aided Design. It’s a sensible amount of RAM for a PC used for productivity.

64GB RAM is only necessary for particularly memory-demanding tasks like editing high resolution video footage, scientific computation, or data analysis and manipulation. Check the memory requirements of the software you are planning to run.

Multi-Channel Memory

RAM can come not only as a single stick but also in kits of 2 or 4 matched sticks. When properly configured, a pair or a set of four sticks, enables ‘dual channel’ mode, giving the CPU twice the number of channels to access it. In effect, this doubles the bandwidth to the RAM and allows faster communication with the CPU, boosting system performance.

In almost all instances a pair of RAM sticks in a kit is the most appropriate choice for a new PC. This enables dual channel mode, maximizing bandwidth. On most motherboards with 4 RAM slots, it leaves two slots free for future upgrades, if required, without removing the existing RAM kit . Four sticks can, in some circumstances, enhance performance a little compared to two sticks but they are much more costly. A single stick may be appropriate if you’re on a tight budget, if performance isn’t an issue, or you if have a system with just 2 memory slots and intend to upgrade later with a second stick.

Speed and Timings

RAM is advertised in ‘Speeds’: 2666Mhz, 3000MHz, 3600MHz and so on. This refers to the number of transfers per second the RAM is capable of. It actually runs on a clock cycle at half the advertised speed, with the ‘Double Data Rate’ name indicating that it can be addressed on both the up and down swing of the clock –doubling the transfer rate.

The RAM speed combines with another property, the timings, to define the RAM’s latency. This is the total time the RAM takes to respond to a request to store or retrieve data. The timings actually define how ‘fast’ the RAM is. A broad view of timings can be obtained by reference to the ‘CL’ rating of the RAM. For DDR4 RAM, CL 16 is acceptably fast, with lower timings becoming increasingly more expensive. Higher Speed in MHz combined with low CL rating indicates faster RAM. When you see a list of timings such as ‘16-18-18-36’ this is a list of the primary timing for the most important functions RAM performs. As with CL ratings, lower numbers are faster.

Most RAM kits have a verified set of timings stored on the stick itself. These are known as ‘XMP profiles’, and you can enable them in the BIOS to set the RAM to run at the correct speed. If you do not do this, or you choose RAM that is faster than that supported by your motherboard or CPU, you may find that the RAM defaults to the base speed of 2133MHz or 2400MHz. It therefore pays to ensure that once you have chosen your CPU and Motherboard, your RAM is reasonably fast but is fully supported by those components.

Performance and Compatibility

Faster RAM can lead to higher performance but the best RAM for your situation depends on the CPU and motherboard you choose, as well as what you will use the PC for. We advise you check the CPU and motherboard specifications to be sure of compatibility, but also research what higher speed RAM options are available.

Motherboard manufacturers publish Qualified Vendor Lists or ‘QVLs’ for RAM which list kits that have been compatibility tested. If you want absolute assurance that your RAM is compatible then check the product number against this list. However, most RAM with sensible specifications will run happily at rated speeds as manufacturers cannot test every kit and configuration on the market.

Aesthetics – RGB & Heat Spreaders

There are a number of features which may help you choose which specific Kit is best for you. Heat spreaders on RAM are largely aesthetic, but matching them to other components can look great. Some RAM kits come with LED’s and light diffusers to allow RGB customization – check that it’s compatible with the RGB software of the motherboard or other RGB components to minimize additional control software. Some RAM has tall heat spreaders which can conflict with large CPU coolers or radiator placement in some computer cases, so read compatibility and dimension notes carefully. If you’re planning a compact or small form factor build seek out ‘low profile’ RAM with a height less than 40mm.

Take Away: For most systems, a pair of RAM sticks totalling 16 or 32GB is the most sensible and cost-effective option. For Ryzen Zen 2 and Zen 3 CPUs, you should look for RAM at 3200MHz, CL16 or faster, with 3600MHz CL16 being the price to performance sweet spot but don’t overspend as performance gains are limited. For an intel CPU, for K series, 3000MHz or faster with CL16 or better is a sensible option. Much faster RAM is the preserve of overclockers and will not yield significant performance benefits in normal use. If you are not buying a K series CPU, check XMP timings against the CPU and Motherboard to ensure compatibility.

Storage

Storage devices are where your operating system and applications are installed, and where all of your files are stored. They come in two basic types, Hard Disk Drives (HDDs) which have a higher capacity and lower cost; and Solid State Drives (SSDs) which are faster but more expensive. Read the “Selecting Features” section about storage drives to learn more.

Hard Disk Drive

When selecting a hard drive, the most important things to consider are capacity and price. Capacity is measured in Gigabytes (GB) or Terabytes (TB), where 1TB is equal to 1,000GB. Because large-capacity hard drives have become cheap in recent years, there is little benefit in buying small-capacity drives, as the cost savings will be negligible. If your only drive will be a hard drive, it is not recommended to go less than 500GB. If you plan to install a lot of video games, store a lot of movies, or do video editing, you will need at least 1TB or more.

Hard drive speeds are measured in RPM and vary between 5,200RPM up to 15,000RPM. For most users 7,200 RPM is totally adequate, and there is little reason to search for anything higher unless it is required for a specific purpose. Cache size also affects access speed and varies widely, but 64MB is considered to be enough for general use.

Connector

Modern hard drives connect with either a SATA, SAS or a U.2 connector. SATA is what you will likely want to use as SAS and U.2 are specialized interfaces generally only used by enterprises or enthusiasts. Most modern motherboards will have SATA connectors, but be sure to check compatibility.

Beware that SATA has varying standards: 1.5GB/s, 3GB/s and 6GB/s. From the names you can tell that these will affect the speed at which the drive will be able to transfer data to the computer. The standards are interchangeable, meaning a SATA 6GB/s hard drive will be able to connect to a motherboard connector supporting SATA 3GB/s (or vice versa) and it will work – but it will be limited to the lower speed. To ensure the higher transfer rate, make sure both the drive and the motherboard support the same standard. Some motherboards have multiple ports that support different standards, for example one is 3GB/s and another 6GB/s.

Also note there is an older standard of hard drive connector called PATA or IDE that was extremely widely used in the past. While these are no longer for sale, if you acquire an older hard drive there is a good chance it might be a PATA drive. It is rare to find a motherboard these days that has PATA connectors, so these should be avoided.

Solid State Drive

Choosing an SSD is confusing at first glance due to the wide variety of form factors, connector types and underlying storage protocols. It is not as complicated as it looks, so let’s break it down:

Form Factor and Connector

Here you have a few choices. A 2.5″ SSD has the same physical dimensions as a 2.5″ HDD, which is commonly thought of as a laptop drive. 2.5″ SSDs will use either a SATA connector or a U.2 connector, with SATA being the most common. As with hard drives, SATA comes in different standards of 1.5GB/s, 3GB/s and 6GB/s, so make sure both the drive and the motherboard support the same standard or the performance will be limited (read more about SATA standards in the Hard Disk Drive section above). U.2 is a less common choice, but has some advantages such as using the NVMe protocol — more on this below.

Another choice is an M.2 SSD — M.2 is the name of both the form factor and the connector by which it attaches to the motherboard. An M.2 drive can be thought of as similar to a stick of RAM. Not only does it look similar, but it slots directly into the motherboard with no need for a cable in between like a 2.5″ SSD does. When selecting an M.2 drive, pay careful attention to the type of connector, referred sometimes as the “keying”. The M.2 connector can be “B” or “M” keyed and will only fit in an identically keyed socket on the motherboard. Some drives have both, dubbed “B+M” keyed, and will fit in either a “B” or “M” keyed M.2 socket. Similarly, M.2 sockets can be “B+M” keyed, and allow all variations of M.2 connector to be inserted. Make sure the motherboard you selected has an M.2 slot before considering this type of SSD.

Lastly, there are PCIe “Add-In-Card” SSDs, which slot into a regular PCIe slot just like a Graphics Card, Sound Card or Network Card does. These can be PCIe x1 (single lane) all the way up to x16, with more lanes generally allowing for higher bandwidth and therefore faster transfer speeds. Keep in mind that PCIe cards with fewer lanes can be plugged into PCIe slots with more lanes (A x2 card could use a x4 slot).

Data Bus and Connector Confusion

Be careful not to get mixed up between the SSD’s “Data Bus” and “Connector Type” as they use the same terms and therefore can be confusing! The Data Bus is a chip on the motherboard that routes data from the drive to the CPU, and there are two standards: SATA and PCIe. As you would guess, an SSD using a SATA connector routes data through a SATA data bus, and an SSD using a PCIe connector uses a PCIe data bus. U.2 and M.2 connectors are less obvious: U.2 connectors use the PCIe data bus but M.2 connectors can use either.

The confusion lies in searching for parts: you might use a filter on an eCommerce website to show only “PCIe” SSDs thinking this will show you only drives that have a PCIe connector, but often this will also bring up drives that use the PCIe data bus, but actually have an M.2 or U.2 connector. The same for SATA, you could end up looking at drives that use the SATA data bus, but have an M.2 connector.

Storage Protocol

SSDs use one of two storage protocols, “AHCI” or “NVMe”. Where the data bus is the hardware that connects the SSD and the CPU, the storage protocol is the software that connects them. AHCI is the same protocol that HDDs use and while it is adequate for hard drives, it is limiting for SSDs. That’s why NVMe was created specifically for SSDs, and allows for much faster data transfer speeds. All drives using the SATA data bus use AHCI. Nearly all drives using the PCIe data bus use NVMe, excepting some older PCIe drives that actually use AHCI, but this is rare.

Compatibility Note: Not all motherboards support NVMe as the boot drive. This means that, if not supported by the motherboard, you could not use a single NVMe drive in your computer. You could, however, have a second drive, either an AHCI SSD or any hard drive as the boot drive, but install your operating system to the NVMe SSD.

Take Away: Storage drives are where you install the operating system and store all of your data. Large amounts of infrequently used data are best stored on the less expensive Hard Drive, while the operating system and frequently used applications would benefit from the increased speed of a Solid State Drive. Price and capacity are the main concerns when selecting a drive or combination of drives, but when selecting an SSD beware of the varying form factors, connector types, data buses and storage protocols to make sure the drive will be compatible with your motherboard and will meet your performance expectations.

Graphics Card

A graphics card will define the gaming performance of a PC. Whist they started out the preserve of gaming enthusiasts many applications now utilize the huge parallel computing abilities of GPUs to accelerate everything from scientific computation to graphics processing in digital art and 3D design. In this guide we’ll explain the various options in order that you can make the right choice for your circumstances.

The fundamental choices that inform your decision making are your budget, what monitor(s) your system will run, and the tasks you expect it to perform. These parameters dictate which graphics cards will be suitable.

Resolution and Refresh Rate

For gaming, the primary consideration is the resolution and refresh rate of your primary monitor(s). The main load on a graphics card is the computation required to render the pixels to the screen, and it must do this for every pixel once per screen refresh. As soon as it has finished one image, it moves on to the next.  At 1080p or ‘full HD’ a GPU is rendering just over 2 million pixels per image, a 1440p ‘QHD’ screen has 3.7 million pixels, whilst a 4K screen doubles that again to nearly 8.3 million pixels. With all else being equal, you need a graphics card with four times the rendering capability to maintain the same frame rates at 4K as you do at 1080p.

Performance is closely tied to cost and whilst a good entry level gaming graphics card for 1080p resolution can be bought for around $200, prices soar all the way to $700+ for 4k capable graphics cards with the highest performance. To avoid under-performance or overspending we recommend deciding on a monitor first, and then you can narrow the options of viable graphics cards.

AMD vs Nvidia, and Graphics Card Manufacturers

Graphics cards consist of a ‘GPU’ chip that handles computation, along with memory and power delivery circuitry on a circuit board. Consumer GPU chips are made by AMD or Nvidia, two competing companies that dominate the industry. The chips that they produce are sold to other companies such as MSI, ASUS, Gigabyte, EVGA, Sapphire and others who build them into graphics cards. Nvidia also make their own graphics cards known as ‘Founders editions’. The actual model of the GPU chip, such as ‘RTX 3070’ or ‘RX 6800XT’ is the feature that is most indicative of the card’s overall performance. If you purchase an MSI RX 6800XT it will perform almost identically to a card with the same GPU made by Sapphire. Generally all modern graphics cards are compatible with all current motherboards, and they all connect via a PCI-E slot. PCI-E 4.0 graphics cards will work in PCI-E 3.0 systems with no detriment to performance.

The important decision is to look at the GPU chips that most closely meet your expectations for performance and features and then consider options with those specific models to find a graphics card with that chip that has the features and pricing that you are content with.

Tier	Resolutions	Examples	Price range
Entry level	1080p @ 60 FPS	GTX 1650 Super, Radeon RX 5500XT, Radeon RX 580	$140-$180
1080p focussed	1080p @ 60-144 FPS	GTX 1660 Super, Radeon RX 5600XT, RTX 2060 Super, RTX 3060	$200-$260
1440p Focussed	1440p @ 60-144FPS	RX 5700XT, RX 6700XT, RX 6800, RTX 3060ti, RTX 3070	$350-$500
4K Capable	4K @ 60+FPS	RTX 2080Ti, RTX 3080, RX 6800XT	$750+
4K+ High end	4K, Production	RTX 3090, RX 6900XT	$1000+

VRAM Quantity and Speed

VRAM is very fast memory that the graphics card uses to store textures and output images during the rendering process. Graphics cards are designed with enough VRAM to perform well at their target resolutions and settings, and 4GB is now the entry level whilst 1080p focused cards tend to have 6GB and higher performance GPUs a minimum of 8GB VRAM. A card with insufficient VRAM capacity may force you to lower texture quality to allow a game to run well. A card with more VRAM is more likely to continue to provide satisfactory performance as the demands of games increase over time. The speed and bandwidth of the memory can have a marked impact on performance. There have been instances where there are versions of a card with slower VRAM that perform considerably worse. Recent examples of these are the GTX 1650 with GDDR4 VRAM instead of GDDR5, and some early versions of the AMD RX 5600XT shipped with VRAM that had 12GB/s bandwidth not the required 14GB/s for peak performance. A quick read around reviews should reveal any pitfalls of a specific GPU that you’re considering. For the most part, manufacturers specify the amount and type of VRAM that compliments the GPU so beyond ensuring that you will have sufficient VRAM for the usage you envisage, you don’t need to concern yourself with the minute details of the VRAM specifications.

Base and ‘Boost’ Clock

One frequently quoted but somewhat misleading attribute is ‘clock speeds’. This relates to how many cycles per second the GPU core runs at, but it’s not a particularly useful metric to judge cards by. Firstly, because of internal architectural differences, it doesn’t tell you much about the actual performance of the graphics card in comparison to a different model. Secondly it is frequently under quoted. All current GPUs use boosting algorithms to determine if they have thermal and power headroom to run faster and lift their clocks speed higher if they do. The quoted figures are therefore a bare minimum and almost all cards will run to higher speeds. Finally, manufacturers like to differentiate their products within the same GPU architecture by quoting different base or boost clock speeds, but these often differ by as little as 35-60Mhz and are usually accompanied by the product being titled ‘OC’ or similar. In the context of a GPU core running at nearly 2GHz these differences are insignificant and do not indicate a noticeable performance difference between products. However, if you are looking for a high performance example of the card this information can be useful as it may indicate the manufacturers confidence in the quality of the chip that is placed into these products. You’ll need to research any specific models you might be considering to find out if a more expensive card is worth the money.

Additional Features

Alongside the raw graphical power of a GPU, there are certain features specific to particular models that can enhance the cards performance or capabilities.

Adaptive sync is a suite of technologies that allow the monitor to vary its refresh rate so that it is in sync with the graphics cards output rate. This reduces or eliminates screen tearing and lag which are both artifacts of desynchronized refresh rates.  An in depth explanation goes beyond the scope of this article as there are competing standards and certifications with subtle differences. However, the landscape has become much clearer since Nvidia stopped exclusively supporting their own ‘G-Sync’ specification and now support Freesync as well. This means that Nvidia GPUs can drive adaptive sync Freesync monitors or those marketed as ‘G-Sync compatible’ if you want a guaranteed level of compatibility and performance. AMD GPUs will only be able to use adaptive refresh rates on Freesync or VESA Adaptive sync monitors, not ‘G-Sync’ monitors – but Freesync monitors make up the bulk of the gaming monitor market and are substantially cheaper than G-Sync monitors. The summary is that an Nvidia GPU will happily use adaptive sync on virtually any equipped monitor, whilst an AMD GPU needs a Freesync monitor for best gaming performance.

There are other additional features you may wish to consider. Nvidia’s ‘RTX’ line of graphics cards have hardware ray tracing and feature ‘Tensor’ cores which are optimized for Artificial intelligence. It also enables ‘DLSS’ which is an AI powered upscaling technology to improve visual fidelity and performance. Whilst Nvidia tout the benefits to gamers few games have yet to fully implement these features. Nvidia also have an inbuilt ‘NVENC’ hardware stream encoder in all their graphics cards from the GTX 1660 Super and up, allowing you to stream your own gameplay to twitch or other services with minimal system overhead or performance impact.

AMD have kept a more value oriented approach focusing on raw gaming performance rather than additional features. They also have GPU based stream encoding, and the 6000 series AMD cards offer hardware ray tracing but it’s a generation behind that of Nvidia.

Finally if you do other work with your computer, like photo or video editing or Computer aided design, then the choice of GPU may be heavily influenced by that. Specific software is often coded to take advantage of GPU acceleration of computational tasks. Nvidia GPUs use ‘CUDA’ cores and these can be utilized to render 3D images outside of a gaming environment. Of note is the popular CAD software ‘AutoCAD’ and it’s reliance on Nvidia ‘Quadro’ cards for certified drivers in workstations. Adobe software also has a strong advantage when using Nvidia GPUs and CUDA acceleration for video edit and some photo editing tasks. Meanwhile AMD’s ‘Open CL’ implementation can be of benefit to computational and rendering software as well, and they are preferred by many Linux OS users thanks to better driver support. The specifics of what kind of card you need will be dictated by the workloads you anticipate, so if you have specialist needs be sure to check the hardware requirements of the software you are looking to run.

Outputs

Most graphics cards have several outputs and these are most commonly HDMI and DisplayPort. VGA is now becoming obsolete, and DVI is also becoming less common. You should check to ensure that the graphics card you are considering has the correct number and type of video outputs for the monitors you will connect.

Power Inputs

Graphics cards can be power hungry with the highest performance variants exceeding 350W total power draw. They obtain this power via two pathways: The PCI-E slot itself can supply up to 75W and  for some cards this alone is enough. The majority of cards have additional ‘PCI-E’ power supply connectors on the side of the card. They come in 6 or 8 pin variants and graphics cards may have one or more of either. The plugs that fit them run directly from the power supply and have a split form factor enabling them to fit in either 6 or 8 pin sockets. What’s important is that every socket must be connected to the power supply, so ensure that your power supply has the right number of PCI-E connectors and total power output sufficient for the entire system.

Cooling

Graphics cards are a self contained device with a GPU chip, memory, and voltage delivery circuitry. These components generate heat which must be dissipated into the case of the PC. They achieve this with heat sinks and fans to blow air through the card and carry heat away. [D1] [JS2] 

Cooling solutions tend to vary mostly by number of fans and the thickness and dimensions of the heat sink. Manufacturers generally apply their own heat sink designs which target their chosen balance of cost, noise and cooling capacity. A card that’s kept cooler can in general sustain higher clock speeds and so perform marginally better. However the biggest differences tend to be in how loud the card is under load. ‘Blower’ style cards use a single radial fan that blows air through the shroud over the GPU and expels it at the back of the case. The high speed noise of this single fan can be quite loud and so this design has fallen out of favour. However it does have benefits in small form factor cases where the hot exhaust is better expelled directly from the case than being allowed to circulate and heat other components, and in workstations or computational situations where multiple graphics cards are stacked close together they may be preferable.

For most normal usage cases a twin or triple fan card is ideal. It’s always worthwhile to check reviews of any specific card before purchase in case they identify any problems specific to that model, such as high temperatures or a loud fan. You will also need to check the dimensions of the Graphics card against the case you have chosen. The most recent and powerful graphics cards tend to have large coolers so you will need to make sure there is room to physically fit them.

Take Away: You should begin with a target resolution in mind, and an idea of budget. From that you will likely have a relatively short list of potential GPUs to choose between. Check the performance by looking at reviews or benchmarks. Nvidia and AMD tend to have options that compete at a similar price point so check to see which fits your needs better. From there you can sort by size, price and features to build a shortlist of potential cards from different manufacturers. Finally, before purchase, check a few reviews to ensure that there are no design flaws or issues with the model you’ve selected.

Power Supply

The power supply unit, or “PSU”, is the component responsible for taking mains voltage from the wall and converting it into voltages usable by the components of the PC. Power supplies come in a variety of form factors and capacities, though typically most consumer PCs will use “ATX”, “SFX” (Small form factor) or “SFX-L” (small form factor extended) form factor power supplies.

What are all these different cables?

Power supplies use many cables with a variety of different connectors to provide power to all of the PC components. Most modern power supplies have a mixture of ATX, EPS, PCIe, SATA and molex connectors.

The ATX connector is the biggest, typically 24 pins on the motherboard side and either 24 or 28 pins on the power supply side (if using a modular PSU). This powers the motherboard and all onboard ports such as USB ports, RAM slots, PCIe slots, onboard audio and RGB. Almost all power supplies will come with a single ATX cable, though a very small number of units come with a second for multi-system setups.

EPS connectors used to power the CPU and typically have either 4 or 8 pin connectors, sometimes multiples of each. Higher wattage power supplies often come with 2 EPS cables. Some older power supplies may only come with a 4 pin “ATX” cable, which was the precursor to the current EPS cable, these usually will still work with newer 8 pin motherboards.

PCIe connectors power graphics cards, along with other high-power PCIe cards, as well as sometimes being used for supplemental motherboard power when using multiple graphics cards. They have 8 pins split into a block of 6 and a block of 2. Often there will be 2 PCIe connectors on a single cable, for high powered graphics cards like the AMD Vega 64/Radeon 7 or Nvidia RTX 3090 it is advised against using both connectors from a single cable, instead, use 2 distinct cables.

Some brands use the same pin out on the power supply side for both the EPS and PCIe cables, to allow for a more flexible approach to cable management and cable choice, make sure that your PSU brand supports this before plugging things in though!

SATA connectors supply power to storage drives, disk drives, some fan expansion cards and some RGB controllers. They have a distinct L shaped design with many pins.

Molex connectors are legacy cables used for old PATA drives. In modern PCs they are only used for some fan and RGB controllers or for supplemental power to the motherboard. They have 4 pins in a straight row.

Similar to the EPS and PCIe cables, many brands will use the same connector on the power supply side, these are often labelled as periph or peripheral ports.

What about voltages?

Modern power supplies have 3 output voltages, split across 4 rails, these are the 12, 5, 3.3 volt and 5 volt standby (5VSB) rails. The 12 volt rail powers most components, such as the CPU, graphics card(s), storage drives and motherboard. The 5 volt rail powers the USB ports along with some electronic components like the VRM. The 3.3V rail provides very little power in modern computers, only being used for the RAM slots and some PCIe cards. The 5VSB rail powers the USB ports and onboard RGB when the PC is turned off but still plugged in.

Some old power supplies may have -5 and/or -12 volt rails, these are no longer used by any modern components. Additionally, some older high wattage power supplies may use multiple 12 volt rails. Unlike modern power supplies, these are physically different rails, with their own circuitry, whereas newer power supplies that advertise multiple 12 volt rails only have 1, and instead use higher quality protection circuitry that allows the power supply to monitor multiple groups of wires separately, rather than just measuring all the 12 volt wires in one larger group.

How many watts do I need?

Power supplies come in a wide range of capacities or wattages ranging from 150 to 2,000 watts or more! Typically, most PCs will require between 400 and 650 watts, however, if you have an abundance of storage drives, multiple graphics cards or a HEDT (High End DeskTop) CPU such as the AMD Threadripper series, then you may require a higher wattage. A good way to calculate your required wattage is to look at the specifications for your desired CPU and graphics card, add the power usage values together and add around 100 watts to cover the rest of your system.

What is a modular Power Supply?

Modular power supplies use detachable cables allowing only necessary cables to be attached, making cable  management significantly easier. For example, a 650 watt power supply might have 2 PCIe cables, 1 EPS cable, 1 ATX cable, 2 SATA cables and 2 molex cables, but many users would only need the ATX, EPS, 1 PCIe and 1 SATA cable. A modular power supply allows you to omit the extra PCIe, SATA and molex cables so they don’t take up space inside your case. However, on some lower capacity power supplies the benefit may only end up being a single cable removed, as most gaming PCs require the ATX, EPS, 1 PCIe, and 1 SATA cable. Given the higher cost of modular power supplies, lower budget PCs may opt to go for a semi-modular power supply.

Semi-modular power supplies are a mid-ground between modular and non-modular power supplies. Typically they have the ATX and EPS cables hardwired, with the PCIe, SATA, and molex cables separate. These power supplies usually cost less than modular power supplies, making them an enticing option for budget builds.

What is 80+?

80 Plus is an efficiency level certification defining how much energy is lost by the PSU when converting from 120/240V to the lower voltages used by the PC itself. Higher Efficiency power supplies typically run quieter, produce less waste heat and cost less to run.

PSU Load	White	Bronze	Silver	Gold	Platinum	Titanium
10% load	–	–	–	–	–	90%
20% load	80%	82%	85%	87%	90%	92%
50% load	80%	85%	88%	90%	92%	94%
80% load	80%	82%	85%	87%	89%	90%

What is group regulation?

Group regulation is an older style of power supply design commonly seen in very cheap units, and is NOT recommended with modern PCs as they can cause unstable voltages which may affect overclocking results and component lifespan. These power supplies can be identified either by reviews or by the 12 volt capacity – group regulated units will have significantly lower max 12 volt capacity than the entire capacity of the power supply, for example, 500 watts on a 600 watt power supply. On the other hand, newer DC – DC power supplies will be able to provide the full power supply capacity on just the 12 volt wires.

Take Away: Select a CPU and graphics card first, and use the power usage values of those parts, plus around 100 watts, to determine the minimum wattage you need. Modular or semi-modular PSUs will make cable management easier, and 80+ rated PSUs will be quieter and produce less heat. Power supplies come with many different cables with varying connectors, double check that you choose one that supports all of your components.

Case

TBC

Other Parts

• Sound Card
• Fan Controllers + fans
• Expansion slot usb/card readers/etc
• Disc Drives