A great many of the folks who land on Tom's Hardware are already deeply passionate about technology and PC hardware. But we know that others are looking to learn more. We're an inclusive bunch. So if you've never built your own PC, fear not. Our editorial team does it all of the time, and we're happy to walk you through the steps, starting with picking the right parts.
There's a good chance that, even if you haven't gotten your hands dirty inside of a case, you still have a basic knowledge of the components that go inside. Experienced builders often have their ideal configuration in mind before they choose a case. But even a seasoned pro needs to be sure that everything's going to fit inside the right chassis. And of course, enclosures vary depending on what you want to do with your PC. Home theater systems, all-in-ones, flashy gaming boxes, and business-oriented workstations all have their own requirements.
Traditional cases follow the size categories below. However, more modern designs tend to stray from those well-defined standards in the name of differentiation. Mid-tower designs, for example, are now found in nearly full-tower scale. To make matters more confusing, they can even be referred to as full towers, even if they lack the drive bays inside that used to define the form factor.
| Traditional Case Sizes | |||||
|---|---|---|---|---|---|
| Type | Full Tower | Mid Tower | Mini Tower | Mini Cube | Desktop |
| Height | 21-24 inches | 17-19 inches | 12-14 inches | 7-9 inches | 3-7 inches |
| Width | 6-8 inches | 6-8 inches | 6-8 inches | 8-9 inches | 14-17 inches |
| 5.25" bays | 4-9 | 3-6 | 1-2 | 1-2 | 1-3 |
| 3.5" internal bays | 6-12 | 2-6 | 1-2 | 1-2 | 2-4 |
| Motherboard Form Factor | ATX, EATX | ATX | microATX | mini-ITX | ATX, microATX |
| Card slots | Seven | Seven | Four | Two | 2-7 |
| Power supply | PS/2 or larger | PS/2 | PS/2 or SFX | SFX or TFX | Various |
Full towers were traditionally tall enough to hold two power supplies, though many had a second hard drive rack where you might expect to find the top power supply. The interior space of a full-tower chassis is useful in some configurations; however, most mainstream users (and even most enthusiasts) simply don't have enough hardware to fill it.
A better justification for picking a full tower is that the top bays are easier to reach when the case is sitting on your floor. A modern example of the traditional full tower, Rosewill’s Blackhawk Ultra, is the right-most case in the image below.

ATX mid-towers are usually capable of holding full-sized motherboards, full-sized power supplies, several full-sized optical drives (DVD and Blu-ray burners), and multiple hard drives. Well-designed units like the Cooler Master Storm Enforcer (above-left) are well-suited for gaming and video enthusiasts, simply because they support a greater number of expansion cards and hard drives than smaller units. A comparison of our current case reviews to models from ten years ago show that good ideas stand the test of time.
A majority of cases give you room for seven expansion slots around back. Typically, that's enough for a couple of graphics cards, add-in sound, and even back-panel brackets exposing USB or eSATA connectivity. But let's say you love your games, and you're dead-set on building a system with three or even four graphics cards. Specifically seeking out an ATX case with eight or more expansion slots might be necessary, since high-performance cards have thick cooling solutions that use the case’s slot hole for support and ventilation.
MicroATX mini-towers are nearly as versatile as mid-towers in applications ranging from office workhorses to high-end liquid-cooled SLI-powered gaming monsters because of their less-imposing profile and easier trasportability. Mini-towers typically support one or two optical drives and one or two hard drives, and the microATX form factor supports a maximum of four expansion slots. All of those limitations are acceptable for most users.
Mini-ITX cubes typically support a single expansion cards and only the smallest power supplies, though the slightly-oversized Lian Li PC-Q08 above (center) supports larger parts. Relying mostly on integrated features and capabilities, these space-saving enclosures were once only good as office- and productivity-oriented platforms. Now, thanks to more efficient host and graphics processors, we also have access to ultra-compact gaming machines and home theater consoles. Though you'll commonly see these referred to as “small form factor”, the term form factor is better applied to the mini-ITX motherboard found inside. Variations of the cube aesthetic alternatively support ATX and microATX form factors.
Formerly used to raise small CRT monitors up to eye level on flat desks, today’s horizontal desktop cases are mostly restyled for home theater systems. They range from the gaming-themed mini-ITX Raven RVZ01 (pictured bottom-center, above) to the eight-inch-tall full-ATX pedestals laying on their sides. Many of the slimmer models use special half-height expansion cards, though the model pictured above uses a right-angle adapter (called a riser card) to situate a full-sized graphics card sideways. If expansion is important to you, beware of models that use a custom-sized power supply, as those may not be upgradeable.
Want something smaller? The yellow box above is the most compact unit we’ve tested to truly qualify as a performance-oriented machine. Called the Brix Pro, it holds two notebook-sized memory modules, an on-board mSATA SSD, and a 2.5” notebook drive. Shorter single-drive units are available with similarly scaled-down performance, and Intel even jumped on the tiny bandwagon with its similar-appearing NUC (Next Unit [of] Computing) form factor. Most of these machines are available either as a barebones system (no drives or memory) or a complete PC, and all of them use external, notebook-style power adapters.
Processor selection can be summed up in three words: performance, power, and price. Our Best Gaming CPUs For The Money column includes general performance and pricing data that applies to both gamers and non-gamers alike, and additional performance data is found in our CPU Performance Charts. It’s also important to know that when our gaming gurus recommend an overclockable “unlocked” processor, the non-overclockable version may offer the same standard-speed performance for less money. Overclocking is a group of techniques designed to push a part’s frequency beyond its designed operational parameters (voltage, heat, etc).
Those same CPU charts show idle and peak power draw, and specific power draw under various types of applications can be garnered through thorough reading of our CPU reviews.
Today, enough software relies on multi-core processing that AMD and Intel have all but eliminated single-core products from their product portfolios. Single-threaded workloads are still fairly common at the consumer level, and technologies like Turbo Boost (from Intel) and Turbo Core (from AMD) are designed to accelerate CPUs when they encounter those lighter tasks. As you read through our processor reviews, the Apple iTunes workload we run is a good example of a single-threaded test.
It’s certainly nice to know that modern operating systems can spread the load of multiple tasks over several cores, and that software developers can break certain tasks into jobs that multi-core processors can handle concurrently. But you're still wondering how many cores you need. If some are good, are more better?
Not necessarily. Software isn't optimized to run across an infinite number of execution cores, and the more resources you duplicate on-die, the more complex your processor becomes, drawing more power. As with all things, there's a balance to strike, depending on what you use your PC for. If you're browsing the Web, responding to email, and writing in Word, most modern dual-core CPUs will feel plenty-lively. But once you start transcoding videos for your tablet or editing pictures taken on your DSLR, it gets a lot easier to overwhelm mainstream hardware.

Game developers have been trying to take advantage of multi-core processing for several years, yet we’ve rarely experienced a significant performance increase from having more than four cores. That’s probably why the largest manufacturer of desktop CPUs, Intel, focuses its gaming-oriented message primarily on four-core processors with the latest advancements in per-core and per-clock productivity.

Unable to match its chief rival in per-clock performance, AMD first countered by releasing processors with more cores for less money. The extra resources can come in useful in heavily-threaded tasks, but a big bump in clock frequency was the only thing that could keep AMD's older technology competitive in gaming circles. That came in late 2013 with two factory-overclocked models.

Power consumption is a major concern in environments where acoustics have to be kept in check. Typically, as you increase power, cooling requirements go up too. And that often means faster-spinning fans, which make more noise. The latest generation of low-energy Intel and AMD processors makes great strides in performance per watt used. Intel also offers even more miserly S-series variations of its Core i7 and Core i5 CPUs that can reduce heat inside high-performance machines.
Once you have a general concept of your own performance and power needs, the above-referred CPU Performance Charts, reviews, and Buyers Guide should help you narrow down a list of specific models you’d like to try. Of course, if you need a little extra guidance, check out Tom's Hardware's CPU forums, where you can ask questions and get answers.
General purpose applications, gaming, high-definition (HD) content, and professional 3D modeling all pose unique requirements for the graphics subsystem. Typically, power users spring for discrete cards, which you drop into an open expansion slot on your motherboard. Both Intel and AMD are adding increasingly capable graphics engines to their host processors though, so you might not even need to buy a card if your needs are basic enough.
If that's the case, the information and resources linked on the previous page are good enough to get you armed with a capable CPU. But if you're interested in playing the latest games using high-quality detail settings, mining cryptocurrencies, accelerating video rendering workloads, or building a workstation designed for heavy lifting, add-in graphics plays a big role in your system's performance.

Extreme gaming hardware is specifically designed to drive the most realistic detail settings at very high resolutions. We’ve even seen game demonstrations using three 4K displays configured in panoramic view, a technology that AMD calls Eyefinity and Nvidia dubs Surround.
Our benchmarking experience suggests that you’ll probably want two high-end graphics processors to enjoy smooth frame rates at the highest detail settings in the latest games using just one monitor at 3840x2160. A Full HD display with a native resolution of 1920x1080 only has one-fourth as many pixels to drive, so you'd get similar performance from a more mainstream graphics card.
Frankly, Don does a stellar job keeping our Best Graphics Cards for the Money column up to date each month. If you want specific guidance on the right card to buy at any given budget point, that's the resource to bookmark. Of course, you can always check out our 2014 Graphics Card Charts for more specific performance data on the GPUs you're trying to choose between. Specific performance differences between specialized card models, along with analysis of new technologies and alternative cooling methods, can be found in our graphics reviews.
As you move away from traditional desktop use cases and toward professional workstations cranking on business-class software, AMD's FirePro and Nvidia's Quadro graphics cards become more apropos, mostly because their drivers are optimized for OpenGL performance and validated extensively with the most notable ISVs. OpenGL is a multi-platform application programming interface that software developers use to render graphics, and it's particularly prevalent in the workstation space. Expect to pay a lot more for correspondingly-tuned cards, even though the GPUs under their heat sinks are exactly the same as what you get from the consumer equivalents.
It's tempting, then, to save a few grand and tap a Radeon or GeForce card for those heavy lifting tasks. But even if you disregard potential accuracy/image quality differences, remember that the desktop boards lack those driver optimizations, and consequently aren't always as fast. If you're using your PC to make money, the smart move is to go with hardware designed for the job. We recently published Workstation Graphics: 19 Cards Tested In SPECviewperf 12, which should help put the potential of professional and gaming graphics products into perspective.
The motherboard is one of the most critical selections affecting the functionality of any build. So, why didn't we mention it first? Well, choosing your case, processor, and graphics solution first may narrow hundreds of possible models down to just a few best matches. Fortunately, our Beginner's Guide to Motherboard Selection contains most of the information needed to initiate first-time builders. In fact, the plethora of previously-published information available allows us to narrow motherboard selection down to a list of criteria:

- What form factor best matches the case you want to use? As seen above, smaller boards can be fit into larger cases, but not vice-versa.
- What interface does your CPU of choice use? Cross-compatibility is severely limited on AMD's sockets (such as AM3 processors in AM3+ motherboards), and Intel's LGA interfaces are exclusive (meaning no cross-compatibility).
- Has the board been approved to work with the processor you picked? In some cases, even if a CPU fits into a particular socket, it may not be supported by a given motherboard's most up-to-date firmware. CPU compatibility lists on each motherboard's website usually refer to specific BIOS versions, and you wouldn't want to end up with a board manufactured three months ago if the BIOS your CPU needs is only two months old.
- How many graphics cards will be installed? Most graphics cards use PCIe x16 slots, and many motherboards appear to have three of them, but the third slot is often impeded in some technical manner. It's important to read motherboard reviews to find out how this might affect your build.
- "Riser cards" allow case manufacturers to produce thinner cases by turning expansion cards sideways. If the case uses a riser card, does it match the motherboard’s slot?
- Non-graphics expansion cards usually fit into PCIe x8, x4, x1 or legacy PCI slots. How many do you plan to use, and what slot type is required for each? Shorter PCIe cards can be placed in longer PCIe slots, but the reverse isn't usually true. And some motherboards share resources between slots, making it necessary to read the board's specifications table or our motherboard reviews.
- If on-board graphics are used, which display outputs are required? Some motherboards give you VGA, HDMI, DisplayPort, and DVI connectors. Others don't give you any. Most on-board graphics processors support a maximum number of two or three displays, as discussed on the manufacturer's specifications table and in our chipset coverage.

- If on-board sound is used, what type of audio system connection is required? Audio over HDMI is nearly universal, but standalone digital audio systems typically use optical or coaxial cables. And live compression of 5.1-and-above sound streams to a digital output typically requires either DTS Connect or Dolby Digital Live (DDL), which is outlined both by the manufacturer and at the bottom of the features table in our motherboard reviews.
- How many network connections will be used?
- Will eSATA, Thunderbolt, or other specialized interfaces be useful?
- What other external connections might be required?
- How many Serial ATA, mSATA, M.2, or SATA Express drives will be installed?
- Will RAID be required? If so, what modes are needed?
- How many memory modules will be installed?
- Will the board be overclocked?
Once you know the answers to these questions, you're ready to take a closer look at our motherboard reviews!
There are a ton of options when it comes to system memory. From data rates to latencies to voltages, the number of combinations can become overwhelming. The easiest answer in the debate of what kit to buy sounds deceptively simple: just buy 1.5-volt DDR3-1600 (PC-12800) modules with CAS 9 timings. All Socket AM3+, FM2+, LGA 1150, and LGA 2011 processors are designed to support at least this memory speed. It's inexpensive as both 4 and 8 GB sticks, and it’s available in both dual- and quad-channel kits.
Yet there are noticeable performance benefits for similarly-priced DDR3-1866 (PC3-14900), particularly if you're using a CPU's on-die graphics engine for gaming. And this speed functions normally, even with processors that are not officially designated to use it (primarily older models or low-energy platforms). And the same easy benefits of DDR3-1866 are even available with most DDR3-2133 kits and modern performance-oriented processors.

The problem with recommending faster memory kits is that they often require at least some manual configuration. If you're not comfortable tooling around in your motherboard's firmware, they might actually drop you to lower performance levels.
You see, Intel’s XMP (eXtreme Memory Profiles) technology facilitates extended memory settings beyond the basic automatic-configuration technology called SPD. Though XMP originally allowed motherboards to set overclocked options like nonstandard voltages and data rates, most of today's XMP-capable modules operate at standard voltage levels and frequencies. Still, when you first boot up, they typically default to either DDR3-1333 or -1066. Going higher requires that you manually enable an XMP profile. Even some DDR3-1600 modules employ XMP (rather than SPD values) to achieve their rated performance levels, and this is particularly true of reduced-latency (CAS 7, CAS 8) modules.
Memory faster than DDR3-2133 is usually expensive and not really required. Our tests have shown that DDR3-2400 is barely beneficial, and only in situations where you're leaning on integrated graphics. We've even seen data rates above 2400 MT/s hurt performance as the motherboard attempts to increase stability.
In terms of memory quantity, Tom's Hardware recommends at least 4 GB for the cheapest Web surfing Windows-based systems. Gamers could probably get by with 4 GB, but we’re more comfortable with the 8 GB that has become the norm in high-performance machines. Few applications push memory needs past that point, though users of memory-intensive programs who also multi-task (such as Tom’s Hardware editors) can occasionally find an excuse to install even more. Users who need more than 8 GB usually know their needs in advance, based on experience with a previous machine.
Even those exceptional circumstances only push us to 12 GB, though 16 GB is easier to install in dual-channel mode (via two 8 GB modules). If you’re desperate for an excuse to add even more, installing RAM disk software (which uses some of your system memory as a virtual hard drive) could be your impetus.

Our memory reviews show a wide range of options, and buying name-brand modules with lifetime warranties from reputable venders is good insurance against unexplained system instability.
Choosing internal mass storage once meant deciding between the performance of a solid-state drive (SSD), the capacity of a mechanical hard drive (HDD), or the greater expense of both. But as with system memory, advancements in manufacturing and maturing technology put medium-capacity SSDs within reach for most enthusiasts. We're even seeing 256 GB drives under $100. That SSD might not make processing-bound workloads run faster, but they'll certainly launch quicker, access the data they need more expediently, and respond in a way you simply won't experience with a hard drive.

Of course, you'd still need a handful of high-capacity SSDs if you were planning on storing your photo, movie, and game collections on solid-state storage. Fortunately, adding the expense of a 1 TB disk for under $60 makes the combination of just-right SSD and big hard drive more palatable.
The flash-based capacity you'll want depends on what you do with your PC. A Windows installation rarely exceeds 32 GB without additional programs installed, even after many months of collecting temp files, cookies, and other "temporary" trash. Popularly-used apps like the Office suite and Adobe's Creative Cloud software can easily consume many times that much space, and games regularly eat up more than 10 GB each all on their own. Most of us could squeeze Windows and essential programs into a 128 GB SSD without much effort, but 256 GB drives are the sweet spot if you're adding a few games, too.
Mechanical storage becomes critical once you start piling on years of pictures, music, and movies. DVD and Blu-ray disc images consume up to 8.4 and 50 GB, respectively. If you love to archive video, your capacity needs will expand very fast this way. Game install packages can be even larger than the games themselves, and those of us with less-than-perfect Internet access are reluctant to delete source data, even when installation finishes.
Although SATA is the most popular desktop storage interface, other drive form factors are becoming more popular. Among them, mSATA is both widely available and mature. Designed to install onto a motherboard, these have become so common that some companies produce adapters to install mSATA drives into 2.5” bays using standard SATA data and power cables.

Beyond mSATA, we're also starting to see M.2- and SATA Express-capable platforms. They're still not very common, but because they both enable PCI Express-based transfers, the performance of future storage products will outstrip today's SATA 6Gb/s drives. As a reminder, just one PCI Express 2.0 lane gives you up to 500 MB/s of bidirectional throughput. A two-lane link should be theoretically capable of 1 GB/s. Meanwhile, SATA 6Gb/s is rated for up to 600 MB/s, though a more practical ceiling is in the 550 MB/s range.

Though most systems use either one large drive or a combination (a smaller SSD and larger hard drive, for example), other configuration options let you choose between additional performance, more capacity, increased data security, or a combination of these.
RAID stands for Redundant Array of Inexpensive Disks, a group of methods that allows data to be spread across several drives concurrently. Most enthusiast-class motherboards support at least RAID modes 0, 1, 0+1, and 5. Each array of disks appears to be a single disk to programs other than the RAID utility.
The possible use of RAID affects the number and capacity of drives selected, so a very brief description of these modes is in order:
- Level 0 divides data into chunks that are spread across two or more drives at the same time, providing up to double the transfer rate (in the case of a two-drive config) and the combined capacity. Because of the way the data is divided, this mode is also referred to as "striping" by in-the-know storage gurus. The major drawback is that if a member drive fails, the array's data is lost.
- Level 1 mirrors two or more drives so that if one fails, data can be recovered from the other. The major drawback is that because both drives (again, in a two-drive array) store the same data, available capacity doesn't increase.
- RAID 0+1 allows four (or more) drives to be set up as a "mirrored" set of "striped" drives. In other words, it's a RAID 1 array composed of two RAID 0 arrays. If one striped set (RAID 0 array) fails, data can be retrieved from the other. Total capacity is still limited to that of one striped set.
- RAID 5 creates parity bits for data recovery. Data and parity bits are distributed across all drives, increasing transfer rate, while sacrificing only the amount of space required to store the added parity bits (the capacity of one drive in the set).
Generating parity bits for RAID 5 requires processing, which means that RAID 5 enabled in software can hog resources. Conversely, RAID Levels 0 and 1 generate little CPU overhead. Gamers with little regard for long-term data storage may choose Level 0 for performance, and anyone with a significant amount of valuable data may choose Level 1.
Tom's Hardware continuously reviews drives and storage controllers, with several of these articles going into additional detail concerning RAID modes, benefits and consequences.
Although it doesn't get its fair share of recognition, the power supply is the single most critical component for system stability and longevity. We've seen cheap models literally go up in flames, taking out several key pieces of hardware in the process. Picking an underpowered model might get you crashes or even boot failures. Since low-quality parts often fall short of their specifications, we'll start off with a link to our power supply reviews and a list of reputable units that have surpassed the expectations of our forum experts. You’ll notice that power supplies don’t get updated as often as other parts, because that technology doesn’t progress as quickly. Quality units have “staying power”.
How much capacity your system needs depends on its hardware configuration. Graphics cards are the most power-hungry components in gaming systems, while CPUs take priority if you're using integrated graphics. Several power supply calculators are available on the Web, though some are more up-to-date than others. The good news is that oversized power units can easily sustain undersized systems without damage, though efficiency sometimes drops when the unit is loaded by less than 20% of its rating.
Power supplies are divided into multiple primary (12 V, 5 V, 3.3 V) and secondary (-12 V, -5 V, 5 V standby) voltage outputs. Better-quality power supplies provide separate over-current protection on each of these output levels, called "rails". Additionally, Intel specified that each rail could provide no more than 18 amps, to reduce the risk of connector meltdown/cable fire.
As the need for more than 18 A of 12 V power became obvious, most manufacturers started dividing their 12 V output into multiple 18 A rails. That created load-balancing trouble as, for example, a two-rail unit could have two highly-loaded cables on one rail and two relatively unloaded cables on the other. This would trip the amperage protection circuitry, even though the internal transformer had power to spare. So-called single-rail power supplies were then devised that violated Intel's mandate, but allowed these systems to at least function. And "smart" power protection circuits have since been employed to reduce the risk of a fire from a single connector (which was the reason for the mandate in the first place).
Simple calculators might do the job for basic configurations, but the highest-end graphics cards place higher load bias on +12 V rails (so much so, in fact, that AMD's Radeon R9 295X2 even has a very specific +12 V rail requirement). Most of today's highest-performance power supplies are correspondingly designed to serve up lots of current on the +12 V rail, though cheaper parts occasionally skimp in that specification. Be on the lookout for this as you shop. AMD and Nvidia originally guided customers to the PSUs with enough 12 V amperage through their lists of CrossFire- and SLI-certified supplies. However, 80 PLUS and its efficiency ratings are also popular sources for determining higher-quality products.
Power supplies are rated in output, and one benefit from 80 PLUS reports in that they contain efficiency data from 20% to 100% load. This enables Tom’s Hardware readers to find a similar configuration in one of our builds, read the input power that we report, and calculate the required output power using 80 PLUS efficiency ratings. For example, a complete machine that draws 647 W through our meter at 85% efficiency needs a 550 W-rated unit (647 x 0.85). Even if you add a little over-capacity for USB-powered peripherals and future drive upgrades, that same machine can run comfortably on a high-quality 600 W unit.
Power supply form factors are not named after motherboard standards, in spite of the way they’re often sold. The ATX motherboard form factor does specify how they’re wired however, and an ATX-compliant power unit could follow one of several sizing standards. These include PS/2, PS3, SFX, or TFX, plus propriety parts.
| Power Supply Form Factors | ||||
|---|---|---|---|---|
| Type | PS/2 | PS3 | SFX* | TFX |
| Height | 5.875" | 5.875" | 2.50" | 70 mm |
| Width | 3.375" | 3.375" | 5.00" | 85 mm |
| Depth | 5.625" | 4.00" | 4.00" | 175 mm |
Often called “ATX”, the PS/2 power supply form factor is a carry-over from the 1980s, long before ATX even existed. Its mounting pattern continues to be used in most mid- and full-tower ATX systems, but large-capacity units are often far longer (deeper into the case) than required by the original specifications. The odd-appearing metric dimensions are artifacts from an original design based on fractional inches. But the inch-based screw threads aren’t as friendly to metric conversion.

Using the same mounting holes as standard PS/2 units, PS3 allowed Hewlett Packard to shorten the overall depth of its 1990s full ATX mini-tower cases. Confusion over PS3’s age can be attributed to the extensive time it took for Intel to add the existing standard to its power supply guidelines. Conflation with SFX can also be blamed on Intel’s placement of its physical dimensions within SFX design guidelines.

One might say that SFX is two form factors, one that’s 5” by 4” and the other 4” by 5”. As a potential third candidate for SFX naming, Intel also specifies a 50 mm-tall version as “SFX, 40 mm Profile” in reference to its fan size. The three (sub-standards) can be differentiated by visual inspection as being wider, deeper, or thinner than the other two. The wider one is more common in consumer-level cases, and the one that’s coincidentally (and mistakenly) most often referred to as microATX. This form factor also allows up to 17 mm of fan housing to extend from one side of the lid, into the computer case.

The narrow TFX form factor allows some companies to make their slim cases even slimmer, though it also intrudes farther into the case. Because PS3, SFX, and TFX are often sold side-by-side under the microATX banner, buyers must often look at the pictures to determine what the seller is actually selling.
EPS supersedes ATX as the electrical standard for high-amperage power supplies, with a 24-pin “EPS” main connector powering most on-board devices and an 8-pin EPS 12 V connector delivering power to the CPU. Most manufacturers make these connectors divisible, with 4-pin sections breaking away to allow fitment in 20-pin ATX and 4-pin CPU power headers.

Also shown is an 8-pin PCIe supplemental power cable for high-end graphics cards, from which two pins can be split away to make it work with 6-pin headers. The plastic insulator surrounding these pins is shaped differently from the 8-pin CPU power connector, preventing accidental misuse.
There’s also some cross-compatibility between wider and narrower cables. Many systems with 8-pin CPU power connectors will operate sufficiently from a 4-pin cable, lacking the extra current needed to support a high overclock. And it’s often possible to hang the end of a non-divisible cable over the end of a narrower connector.

Drive power cables include the old-fashioned 4-pin “ATA” style, a smaller “floppy” style, and the more modern “SATA”. Increasingly, power supplies lack the floppy power cable, but, because some accessories use it to power other things, you often get an adapter for one of the ATA-style connectors. In this day of SATA-based storage, the four-pin ATA leads rarely hook up to drives, but rather power cheap fans, fan controllers, and multi-drive backplanes.
In total, builders must find a power supply that’s quality-made, fits their case, has enough capacity, and has all the required cable ends. If that last measure isn’t met, adapters are usually available.
It's possible to complete your build and get to gaming with nothing more than the previous-mentioned components (along with a thumb drive for your operating system). The number of downloadable programs has increased to the point that many of our readers never need an optical drive (CD, DVD, BD-ROM). Cases come with mounting screws and usually include cooling fans. Most retail-boxed CPUs have a heat sink and fan. And a majority of motherboards are bundled with cables.
For other enthusiasts, the ability to run old programs or play media is critical. Overclockers, especially, will immediately toss aside whatever thermal solution their multiplier-unlocked CPU came with in favor of something more effective.
Even the most tight-fisted builder should be able to afford a DVD writer, with typical online prices ranging from $20 to 40 on the latest models of many popular brands. Blu-ray writers are more expensive, though not nearly as bad as they once were. Combo drives with Blu-ray read and DVD write capabilities used to fill the pricing gap, but that market shrank as the gap decreased.

Other power users prefer additions like a premium sound cards or TV tuners, though integrated sound is quite good nowadays and Internet-based streaming services make TV cards largely superfluous. Then again, that's what makes the PC so great. You have the freedom to swap parts in and out as your needs change.
The debate over optional or mandatory upgrades heats up when we get to the world of overclocking.

CPU coolers range from tiny devices as small as a 2” cube to enormous liquid-cooled radiator systems. We review the entire range, and think that first-time builders who would eventually like to overclock will have the most success with something that’s easy to install. Depending on your case, that could still give you dual-fan sealed liquid system or big air options.
Any of these items are optional for most builds, so let’s get back to the mandatory steps.
Online merchants leverage lower operating expenses to price products far below those needed to keep the doors open at brick and mortar shops. But shipping costs can still kill your hopes for big savings, particularly if you shop across multiple storefronts. Per-item shipping often gets better as more items are added to the order, so the savings attributed to buying online are maximized by purchasing from the fewest possible sources.
A difficult cascade of questions may consume you if you consider many sellers, various components at different prices, and a range of shipping rates. The easy path is picking one vendor able to give you the best deal on your complete list. Keep in mind that single-item shipping rates quoted through shopping engines should drop significantly as order size increases, and if this doesn't happen it's time to check the next vendor on your list!
Local stores must increase prices to cover their higher operating expenses, but many receive items in large enough quantities to save you some of the money you'd otherwise pay on shipping. Consider the example of a single stack of DVD-R media: online pricing might be $6 plus $8 shipping, totaling $14. If a local store bought 100 stacks at a 10% discount, squandered that 10% savings on bulk shipping, and added a huge 50% mark-up, it'd still be able to sell them for $12...saving you $2 and several days of waiting.

"Loss leaders" are another way for buyers to save when purchasing locally. These are items that larger stores like Best Buy or Fry's Electronics sell at a loss in order to lure you in, hoping sales staff or glamorous displays will get you to pick up a few more things on the way out. Relying on one-time deals often requires substituting a lesser part to get a better value, however.
Level Of Service
It's often said that you get what you pay for, and service is one area where local stores have the ability to outperform their online rivals (though not all of them do). Because small shops are constantly trying to build their reputations, and because they deal in lower volume, they're usually willing to go the extra mile to answer questions and earn your business. Larger electronics chains focus on volume instead, and would rather sell you another part than figure out why the one you have isn't working. Online merchants expect you to have enough knowledge to figure things out on your own.
Consider the situation of dealing with a compatibility issue:
- Smaller, locally-owned shops will usually offer advice, inspect the item for free if you believe it's defective, or diagnose it in your system for a reasonable fee (again, that's not to say all of them will). On the other hand, they may not be willing to provide a refund if you try to return a new component in used condition.
- Most online merchants don't provide adequate tech support, instead going directly to the return process while charging a 15% "restocking fee" for any returned item. You'll have paid shipping both on the delivery and the return, and your 15% fee will go towards someone else's "open box" price reduction.
- Favoring irresponsible buyers, "big box" retailers might give you all your money back if you come up with a good enough reason (or plausible excuse) for the return.
Seller Integrity
Local stores live and die by word of mouth, and will normally try to settle disputes amicably. Larger chain stores will generally try to dodge the bullet, though it might take a while for you to reach a satisfactory outcome.
Online merchants need to keep the majority of customers happy, but a minority can fall through the cracks. Many price comparison engines—such as Google Shopping and Amazon—have rating systems linked to viewable buyer comments.
Auction sites are a great place to find discontinued hardware, but final selling prices on newer parts often exceed those of larger discount sites. Manufacturer warranties may not apply (especially to gray-market parts) and seller warranties are only as good as the seller's word. Be careful, though, and learn from my personal experience. I found a seller who had spent more than three years building his reputation as a power seller, and had a favorable rating of over 99%. His "retirement" plan, apparently, was to advertise items he didn't own during his final month of sales, and he was able to abscond with a six-figure salary of ill-gotten gains, a few hundred dollars of which were mine. It has become more difficult to succeed at these scams in recent times, as payment companies with buyer protection will now track down criminals who've cost them insurance money. Yet, we still hear of sellers sending a box full of rocks or paper to prove shipment. Unless the seller has a history of doing this, it's your word against his concerning what actually arrived. Auction sites become a reasonable option whenever the benefits substantially outweigh the risks. Just be sure that you take all necessary precautions, and are prepared for any hardships that might come in spite of your caution.
Purchasing Summary
Online merchants offer the lowest price, but shipping policies favor large purchases. If you can get most items from one site, your savings could be significant. Inexpensive orders are often best-sourced locally due to shipping fees.
Human interface components like keyboards, mice, and game controllers are so dependent on individual ergonomics that it's always best to try a few before making a purchase. Large retail chains may provide an adequate selection of parts to try out, but some buyers use these stores to "window shop" before placing an online order.
Final assembly is usually the quickest part of a build. Component selection may require days of consideration, and finding the best source for your gear can take up the better part of a day. But plugging connectors and inserting screws shouldn't take more than a few hours, even for the most inexperienced builder.

If you're familiar with a few simple hand tools, you could assemble a complete PC in less time than it takes to read this guide. But troubleshooting (that is, going back to figure out what went wrong if things don't work the way they're supposed to) could slow things down significantly. Phobias aside, you're unlikely to damage your hardware or yourself if you follow a few very easy precautions, and we hope this final segment will eliminate hours of post-build trial and error.
First Precautions
Nothing creates a sinking feeling faster than damaging a critical component before you're even finished putting everything together. Major concerns include electrostatic discharge, dropped parts, and breakage caused by force fitment or scratched circuits.
Accidental electrostatic discharge (ESD) can destroy PC hardware, a fact that causes many building guides to exaggerate this danger. In truth, few experienced custom PC builders take more than the most basic precautions against ESD; even when it does occur, it's likely to follow the component's ground plane rather than zap its most sensitive parts.
The most basic precaution is to occasionally touch a ground, such as a large metal office desk or the metal case of a plugged-in system, to discharge your body. Additional ESD risks come from the use of carpeted workspaces and extremely dry environments, so another level of protection may come from the use of an antistatic mat under the chair and a humidifier for extremely dry rooms. Grounded wrist straps are an over-the-top method of protection rarely used outside of production environments, yet the extra-cautious will attain peace of mind when wearing one.
Fallen components seem easy to prevent in theory, but damage from droppage is a far more likely cause of broken components than ESD. Hard disk drives are often mishandled during installation and other parts can be easily knocked from a desk. Reducing fall distance is as easy as moving work away from the edge of a desk, and reducing damage from parts getting knocked to the floor is as simple as leaving them in the box until they're ready to be installed.
But one physical issue that even the most cautious of us can't prevent 100% of the time comes from dropping processors into their interfaces at a slight angle. This problem is specific to Intel’s latest LGA interfaces, because the contact pins have gotten thinner as the company has added more of them. Intel’s pins act like springs, so that even the slightest damage can cause insufficient contact pressure. There’s no hard-and-fast rule to prevent you from damaging the motherboard as you guide your processor into place. Just be extra careful during those critical few moments.
Beyond having a CPU slip out of your fingers, assembly damage can include situations where parts are misaligned and forced into place. Most components require only a small amount of pressure to seat the connector, but a few do need more aggressive tactics. We’ll cover those specifics as we install each part.
Many technicians refer to the CPU, motherboard, DRAM and graphics as a platform. These parts can be assembled and tested outside of a case by connecting a power supply and power button. And, except for a discrete (separate) graphics card, they can usually be inserted as an assembly into an empty enclosure.
Socketed processors have followed a common theme for at least 20 years: an arrow on one corner of the CPU aligns to another arrow on the CPU socket. This is the first method manufacturers use to assure proper orientation, but AMD also uses missing pins with blocked interface holes to further prevent improper installation.

CPU pins are easy to bend, so if you're really rushing through the motions, it's certainly possible to force a processor into its socket the wrong way, smashing its pins in the process. With the tension lever released as shown, the CPU should literally drop into the socket under its own weight, with no force applied. These are known as Zero Insertion Force (ZIF) sockets.

After checking to make sure the CPU is fully inserted, press the tension lever into the horizontal position to lock it in place.
LGA processors have edge notches to prevent incorrect installation in addition to being marked with an arrow as a visual guide. A load plate holds the pinless CPU tight against socketed contacts, called lands. One or two locking levers apply the load.

After making sure that the CPU is correctly installed (as shown above), lower the steel load plate over the CPU and rotate the wire clamp into its locked position.

Thermal interface material (also known as thermal compound, paste, or grease) fills tiny spaces between the CPU and its cooler to assure optimal heat transfer. Most factory-supplied coolers have a stiff factory-applied TIM that becomes soft when heated by the CPU, but other coolers require the manual application of thermal transfer grease or paste.
Igor Wallossek’s article on thermal paste installation shows a perfectly acceptable way to add today’s thick thermal materials without creating a mess. A small blob in the center of the sink will indeed spread as shown in the above photos, and thermal softening will likely spread it even more as the system is used. But I like to maximize contact surface area all the way to the corners, so I usually put a slight smear of paste around ¼” from each corner in addition to the small blob in the middle. My old method of dabbing it on worked only with the low-viscosity pastes of the past.
Excess paste will squirt out around the edges of the CPU, so it's important not to apply so much as to create a mess. Cleaning pastes out of crevices can be particularly difficult, and becomes necessary when using certain metallic thermal solutions.
Clip-on CPU coolers are still used by AMD for its Socket AM3+ and FM2+ processors, and the clip is still compatible with most of the firm’s older socket interfaces. With the cooler in position, slip the non-levered end over the corresponding plastic hook, then repeat the process on the levered end. Finish the installation by flipping the lever to apply pressure.
Pinned-on CPU coolers use mounting holes rather than the more traditional clip bracket. Introduced with Intel’s LGA 775 package and retained through the company's modern LGA 1150 interface, installation requires pushing each pin into the corresponding motherboard hole until a click is felt or heard.

The lower pin (translucent white, above) is hollow, split on one end, and has barbs on the split end. This part goes through the mounting hole first. The upper pin (black, above) protrudes through a hole in the lower pin’s center to wedge the barbs apart. Twisting the top of the pin ninety-degrees counterclockwise unlocks the spring pressure, allowing the cooler to be removed.
Because a counterclockwise twist defeats the latching mechanism, check that all pins are properly twisted fully-clockwise before attaching the cooler.
Screw-on coolers solve the problem of fragile plastic pins and the four points of motherboard strain by using screws and a load-spreading support plate. This greater security and motherboard protection is particularly useful with large and heavy coolers that require increased contact pressure across the CPU’s heat spreader. The support plates are typically designed to fit Intel's four-pin mounting holes, or replace AMD's clip-style brackets. Intel’s LGA 2011 motherboards ship with a support plate already installed, and many coolers also ship with a second set of mounting screws to use its threaded holes.

Because the support plate must be placed behind the motherboard, these coolers should be mounted before the motherboard is installed into the chassis. Many cases have an access hole in their motherboard trays specifically for this purpose, but it’s usually easier to reach the screws with the motherboard unobstructed by case walls.
Installing RAM
System memory is keyed so that it only fits into the slot one way. Because this key is off-center, backwards modules cannot be fully inserted. Check to make sure that the notch in the module's contact area aligns with the slot's key, and press each module into the slot until a click is heard or felt from the latches. Fully seating modules may require a relatively significant amount of pressure.

Our configuration called for a pair of modules in corresponding slots to enable dual-channel mode. Check your motherboard manual to see which slots should be used for this performance-enhancing orientation.

Also note the slot numbers, which are usually written on the board, and compare them to the module installation order outlined in the motherboard manual. This was particularly critical with LGA 1156- and LGA 1366-based motherboards because they relied on a DIMM in the second slot of each channel for termination, though many LGA 1150 and 2011 motherboards aren’t as fussy.
Most enclosures support a range of motherboard sizes, each with a few different mounting points. These points connect a layer of the motherboard called the ground plane to the case's mounting tray, reducing signal crosstalk due to radio frequency interference (RFI). Thus, the mounting points are usually grounded.

Misaligned mounting points could contact a hot trace on the motherboard's back side, so case manufacturers usually make them removable via metal spacers called standoffs. It's important to observe the exact location of each mounting hole in the motherboard before placing a standoff in the corresponding tray location. A mistake made here could potentially damage the board, though the most likely result of an improperly-placed standoff is a system that simply refuses to power on. Arrows in the photo below illustrate the matching mounting points where standoffs were placed.

The ATX form factors specifies the size and location of a rectangular plate, called an I/O shield, which fills the gaps around the ports and connectors on the back of the motherboard. That is to say, an I/O shield fits a customized port selection to a standardized hole in the chassis. Cases often include an old-fashioned standard plate that must be snapped out before inserting the new, custom replacement.

Note that the upper tabs of this I/O shield hang down because it arrived in a semi-flattened state. These need to be bent approximately ninety-degrees from the surface to prevent them from blocking nearby ports during motherboard installation. The left tab in the photo below has been bent to the proper orientation. Many of today’s most popular boards instead use foil-faced foam to contact the ports.

Recheck standoff positions before inserting the motherboard at a slight angle, aligning ports with cover plate holes while guiding the board until it rests flat against the standoffs. Grounding tabs or foil-faced foam on the I/O shield will typically push the motherboard out of position, but the board should be easy to push into place. Align one hole perfectly with the standoff and affix a screw, then push the board into alignment for a second hole before tightening the second screw. The first two screws should prevent the board from twisting out of position while installing the remaining screws.

The power supply is usually secured with four coarsely-threaded screws, though it’s not always mounted to the back of the case. Some enclosures relocate the power supply and use an extension cable to place power on the back. Variations in design may demand that the power supply is installed before the motherboard, as specified in the case’s manual or installation guide.
Expansion cards are usually available as PCI Express (PCIe), since legacy PCI is nearly extinct. Available in single-, four-, eight-, and sixteen-lane versions, the PCIe standard retains compatibility between shorter cards and longer slots. The below image shows a PCIe x1, PCIe x16, and PCI slot for comparison.

PCIe allows shorter cards to be placed in longer slots, such as an x1 card in an x16 slot. Conversely, longer cards can only be placed into shorter slots when the forward end of that slot has no cap. Because the difference between open-ended and closed slots isn’t easily seen in photos or explained on motherboard specification sheets, many manufacturers use x16 slot connectors for their four- and eight-lane interfaces. If you noticed that half of the electrical contacts are missing from the above PCIe x16 slot (the little metal pieces are tough to see), it’s because that slot is wired as an x8 interface.
Though our example motherboard includes on-board graphics, we chose to use a PCI Express graphics card for enhanced performance. The PCIe x16 card is inserted until a latch on the slot engages the card's hook. These latches are present on most PCI Express x16 slots, but are not found on lower-bandwidth PCI and PCI Express x1 interfaces.

As with other cards, a case screw or quick-release latch secures the top of the card's metal bracket at the opposite end.
Internal 3.5" drives are traditionally secured with coarse threaded (UNC) case screws, while external drives, 2.5" drives and bay devices usually have fine metric threads. External drives typically slide in from the front, while internal drives often slide in from inside the case.

Several manufacturers offer tool-free installation using drive rails, sliding latches, or other pin-loaded devices that engage with screw holes. Our case reviews highlight several designs.
The motherboard cables of new systems are usually based on the expanded EPS12V standard, which encompasses previous ATX standards. Previously found on server-sized EPS power supplies, the 24-pin main power cable is both forwards and backwards compatible with the earlier 20-pin part. The below example shows how a 20-pin plug fits into a 24-pin socket; the wide latch is designed to work with either 20-pin or 24-pin plugs.

One of the reasons for a 24-pin power cable includes added amperage supplied to PCI Express slots compared to older interface standards. While most cards won't overdraw a 20-pin connector, graphics card makers occasionally have suggestions for a higher minimum level of available power.
The 4-pin or 8-pin ATX 12V connector satisfies the electrical demands of the CPU. Formerly known as the "P4" power connector, it was added by Intel to supplement its Pentium 4 processors, and later adapted by AMD motherboard designers. The newer 8-pin versions were originally meant to address phenomenally power-hungry Pentium D and Prescott-based Pentium 4s, but many modern AMD and Intel processors are efficient enough to once again work from 4 pins. Most 8-pin boards will work with both 8-pin and 4-pin power, as the connectors are cross-compatible.

Also seen in the photo above is a 4-pin CPU fan power connector and the front-panel audio connector. On-board 4-pin fan connectors are designed to provide pulse width modulation (PWM) automatic speed control, but the connectors are once again cross-compatible with 3-pin fans. Some motherboards are able to control fan speed via either voltage changes or pulse width, while others will run the “wrong” fan at full speed, continuously, without harming the system.
Front panel audio cables are often available with both AC97 and HD-Audio connectors, where HD-Audio is a slightly newer standards and AC97 is extinct. Using the "wrong" connector may temporarily reduce the number of available audio channels, but will not harm any components. The key-pin for audio headers is in a different location from other panel connectors to ease installation.
The case's power switch, power indicator light, reset switch, and hard drive activity light are usually connected at the motherboard's lower-front corner. LEDs pass current in only one direction, and positive pins (indicated by a "plus" sign below) normally connect to the colored wire on each lead. A black or white lead wire usually indicates negative or ground state. If your power and reset switches work but your power and HDD lights don't, your LED connectors are probably flipped.

USB connectors have been standardized for over fifteen years, and we suggest that first-time builders not attempt to incorporate any parts older than that. The missing pin location is blocked by most front-panel USB connectors to assure that the connector is polarized correctly. A reversed connection would damage the motherboard, so 4-pin, 8-pin, or single-row internal break-out cables require special care. The missing pin indicates the negative/ground end of the connector.
Device Cable Installation
Serial ATA (SATA) power and data cables are keyed on the sides, as seen on the drive below. Some early SATA drives were also able to accept older 4-pin ATA power connectors. The sticker warns that builders should choose either SATA or legacy power, but not both.

Many PCI Express graphics cards require more power than the slot is able to provide, and use the 6-pin input connector shown below or a newer, higher-amperage 8-pin version. The 6-pin connector must never be confused with 4-pin or 8-pin motherboard power, as its polarity is the opposite of those! Fortunately, the newer 8-pin PCI Express power cables are designed in such a way that they cannot be forced unintentionally into a motherboard’s 8-pin connector.

Are you trying to rescue data from an outdated drive on your new system? New motherboards don’t support these drives natively, old SATA-to-ATA converters often had limited compatibility, and old drive interface cards used legacy PCI. If you really are stuck trying to pull childhood photos off a deceased family PC, I’d like to suggest installing the drive in an external USB-based adapter. That is, if you can still find one.

ATAPI and Ultra ATA drives have pin 1 on the "other" side of the connector, as seen when facing it (on the right in the photo below). A key is located on the top of all 80-conductor ATA cables to prevent upside-down insertion.
Final Words
No system is complete without software, and most operating systems are available on a bootable DVD. The system's boot order can be selected in the motherboard BIOS, usually under the "Advanced BIOS Features" menu, and should be set to boot from CD first. Many modern motherboards will list the actual name of the drive in the boot order, while others will only list it by device type.
Further BIOS tips and tricks can be found in our BIOS for Beginners.
We hope that this series has made your build a complete success, but if it hasn't, members of our Community Forums eagerly anticipate your technical questions.



