Build the Best Gaming PC Your Money Can Buy: A Detailed Guide (Updated Sep 2014)

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

Useful Resources

Throughout the guide, I provide URLs to additional reading material to supplement what I’ve said or to provide context. While I encourage you to read them, most are optional unless specifically noted.

In addition, the following resources not only helped me organize and write this guide, but they’ll be incredibly useful to you when the time comes to select your parts for a new build, and later down the road when you decide to upgrade:

====///====Tech Reviews====\\\====
Tom’s Hardware is the probably the most popular benchmark/review site. You'll find it useful for their monthly Best-for-the-Money articles, such as “Best CPU for the Money” and so on. It simplifies part selection based on how much money you can spend. Sometimes, the choices are questionable in my opinion, but not exactly flat-out wrong, either. Just differences of opinion, which you’ll learn are not as important as people make them out to be. Tom’s Hardware has a global community and is an extremely active resource for PC hardware reviews, and with that popularity comes more advertisers, and with advertisers comes more disdain.

The rest of the sites I visit are generally the same, providing reliable benchmarks and an assortment of reviews, with no real caveats to speak of. These sites include:
> Anandtech.com
> Guru3D.com
> HardOCP.com
> PCPer.com
> Techreport.com
> Xbitlabs.com

====///====Finding Parts====\\\====
First and foremost, you should use PCPartPicker when compiling any part list. It allows you to find parts based on features/specs and price ranges. It has a built-in compatibility checker so it only shows you parts that are compatible with the components you’ve already selected. By default, the site will show you the cheapest prices of each component with or without mail-in rebates, which stores sell the part, and how much each store charges including shipping—but if you have preferential vendors, you can opt to only see parts from certain vendors along with their prices. If you create an account, you can save multiple part lists for various budgets or build types.


An easy interface to search different components at PCPartPicker


Price list for the Intel i5-4670K


An example configuration on PCPartPicker


You can monitor the price history of individual parts or an entire configuration.

A few other places to check include Newegg for its advanced search functions and volume of reviews, and reddit.com/r/buildapcsales for various promos and rebate deals.

====///====Assembly====\\\====
This guide doesn’t cover assembly, so if you’ve never built a PC, your best bet is to refer to video guides to walk you through the process. A popular tutorial is from Newegg, hosted on YouTube in various segments:
> Part 1 briefly covers part selection and compatibility, which you can skip since you’ll learn far more from this guide: Part 1

> Part 2 is a 44-minute video that walks you through the whole assembly process: Part 2

> Part 3 is a 51-minute video that walks you through software setup: Part 3

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------
Monitors and Frames Per Second

Yes, we’re starting with a part that isn’t even included in our budgets. But without this information, we cannot create an informed build. That’s because your monitor’s resolution will dictate your video card needs, and the video card commands a large portion of the budget, as it should.

This section is not about actually selecting a monitor, however. Here, we will focus on only two factors that will directly influence your overall gaming experience: Resolution and Refresh Rate. You probably already know what resolution is, so let’s save resolution options for later and take a moment to discuss refresh rate. In that respect, monitors generally come in two flavors: 60 Hz and 120 Hz. Some people mistakenly view this as a direct indication of frames per second (FPS). While this does have a significant and somewhat absolute correlation with frame rate, it’s important to understand that FPS (number of frames per second being processed) is independent of the monitor refresh rate, which is, you guessed it, the rate at which the monitor’s hardware is refreshing the display. Why does that matter? Answer: Just because you have a 60 Hz monitor doesn’t mean that anything beyond 60 FPS is useless and unnoticeable. You can notice a benefit with a frame rate higher than 60 FPS, even on a 60 Hz monitor.

Over the years, people have parroted some variation of the phrase: “The human eye can only perceive 30 frames per second.”—or 40 or 60 or 15 or 100. There are too many variations in the way digital media is presented to make such a blanket statement. To learn more, read this article: LINK

Forget anything you’ve heard about how many frames per second the human eye can see. We aren’t watching a movie. We’re playing a game that requires your physical interaction as well as dynamic input from other hardware (i.e., mouse and keyboard). It’s a different experience, and here’s why.

When you play a game, you have the option to enable or disable vertical sync (VSync). When you enable it, your video card’s goal will be to keep the frame rate of the rendering engine (the game) equal to your monitor’s refresh rate. In this case, 60. Assuming you have a video card powerful enough to achieve this, VSync turns your monitor into the frame rate bottleneck. This isn’t a bad thing, visually. 60 full frames per second is the maximum your 60 Hz monitor is capable of allowing your eyes to see. It’s the frame rate you want to achieve, minimum, in most cases.

When you disable VSync, your video card will work independently to output the highest frame rate it’s capable of producing under that game engine at the graphics settings you’ve chosen. So, without VSync, you might run a FPS monitoring program and see that you’re getting 90 FPS on your 60 Hz monitor. What you’re actually seeing is the FPS that the GPU can produce before it reaches your monitor. The monitor is processing all of those frames, but it is not displaying every whole frame. But this surplus of frames, so to speak, lends itself to another absolute necessity in fast-reflex games such as First-Person Shooters, especially online: low-latency mouse response. Mouse lag.

In the simplest of terms, when in VSync mode, your video card waits for the monitor's next refresh operation before rendering the next frame. In such a case, when you move your mouse, it must wait for the game engine to respond, which is waiting for the video card, which is waiting for the monitor to refresh, and then you see your cursor/crosshairs move. Because the video card is depending on the monitor, it has to wait for the next refresh cycle before rendering the next frame. If the frame rate is even lower than the refresh rate, the monitor is now performing more refresh cycles before the next frame is rendered, causing an even longer waiting time for the mouse to respond.

The latency between the video card and the monitor is the largest, which is why watching a cut-scene at 24–30 FPS doesn't look bad to your eyes, but trying to manipulate your character in real-time at that frame rate can feel loose, slow, and unresponsive. Ideally, you want to be able to render more than 60 FPS consistently, without VSync, so that your video card is always ready for your monitor to show a frame before the next refresh cycle.

Here’s a visual example, involving two theoretical players with the same hardware and precisely the same skill level:



The above chart only depicts a stop/aim/shoot scenario, which doesn’t even take into account the lateral movement of the players or even the crippled accuracy that results from Player B’s lower frame rate, but as you can see, the additional time spent waiting for frames and refresh cycles nearly doubles the total latency between the player’s mouse movements and the action on the screen. If you pit these two players against each other, you’d certainly want to be Player A. You would see your own movements before Player B does. You’d even see your opponent’s movements before he does because the game engine’s response is not dependent on the monitor (therefore the enemy is in the player’s virtual line of sight at the same point in time in both scenarios).

I added human reaction time to the graph to show that the input-to-display time all occurs in a fraction of the time it would take even a skilled player to react and compensate. This lag between your mouse and the monitor's display of your movements is only a matter of milliseconds, but it is noticeable when you have to make frequent, split-second movement decisions in a game.


Test your own reaction time at humanbenchmark.com

But like most benefits, there’s a cost. With extra frames comes the potential for screen tearing, which is when the screen appears torn between multiple frames. Some people hate it, and some people don’t notice it. Here’s an extreme example of screen tearing:


Image borrowed from tweakguides.com

Notice the different horizontal sections of the screen that are out of vertical alignment. As I said, this is an extreme example to show its effects clearly. Tearing most commonly occurs at frame rates higher than your monitor’s refresh rate because the monitor is trying to display more frames per refresh cycle. This will cause the monitor to display the next frame before it's finished displaying the previous frame. It’s more noticeable in some games, and completely unnoticeable in others.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------
The Video Card

You probably know the two big players in gaming graphics: AMD and NVIDIA. They’re constantly at war, and it’s definitely not one-sided. Both companies continue to edge each other out with each new release.

The never-ending war and the uncertainty of future game requirements make it difficult to choose a side. It also doesn’t help that video card specs are among the most confusing of all PC components, even for some enthusiasts. In addition to GPU and memory frequency, there’s texture rate, pixel rate, shaders, texture mapping, memory bus, floating point, memory bandwidth, and so on. Making a decision based on those specs is complicated and unnecessary. On top of that, there are various product lines meant for different levels and types of use, and those product lines have their own models and submodels. The goal in this section will be to simplify the decision-making process as much as possible. The only question that really needs to be answered is this: How well do these video cards perform in actual games?

And on that note, let’s talk about the settings that drive performance.

====///====In-Game Graphic Settings====\\\====
When you first launch your shiny new PC game, the next thing you’ll want to do is optimize your visual experience by changing the in-game settings. Some games will try to detect your hardware and automatically adjust the settings based on that, but those auto settings are too conservative at best, and way off base in many cases. Here are some of the common settings you’ll want to check:

====Resolution====
If you have a 1920x1080 monitor, you need to be playing at 1920x1080. Some people recommend lowering resolution so that you can turn up other settings. If you’re building a gaming PC and you aren’t able to run games at your native resolution, you should not have built a gaming PC.

====Antialiasing====
If you’ve ever looked at a curved or straight object in a game and noticed that the edges appeared to be made up of small staggered squares making the edge look jagged, you’ve witnessed what’s called “aliasing” (commonly referred to as “jaggies”). This occurs because what you are seeing on your screen is only a sample of the 3D data being processed by your GPU. The resolution of your monitor limits the number of pixels your GPU is able to show you. Pixels are square, not round, so the more pixels you have, the less aliasing you will be able to perceive. The image below shows an example of aliasing. Notice the jagged edges on every object.


Particularly, the jagged power lines are a quite noticeable side effect of aliasing. Image borrowed from tweakguides.com

This is where the antialiasing feature comes into play. This takes those jagged lines and resamples them (once, twice, as much as eight times depending on your settings) to blend them with the rest of the displayed textures.

A simple visual example of what antialiasing does to edges:



The results can be quite noticeable, as shown in the example below:


Image borrowed from tweakguides.com

The size of your monitor can affect your perception of aliasing. The pixel count on a 1080p monitor is the same at 27” as it is at 22”. The larger the monitor, the larger the pixels. Depending on viewing distance, the smaller monitor will provide a perceptively clearer, less jagged image. Antialiasing in general requires more RAM and GPU power compared to other settings. How much is a difficult question to answer because it depends on the game, your GPU, your drivers, and the type of antialiasing being used—and there are quite a few, ranging from GPU- and memory-intensive brute force multi/super sampling antialiasing (MSAA/SSAA) to software-based types such as fast approximate antialiasing (FXAA), the latter being less effective but causing little or no performance loss.

Because of the immense variables here, it’s difficult to plan a PC with the expectation that it will handle every game with maxed out antialiasing. In fact, MSAA/SSAA is the first thing you should reduce or eliminate if your in-game frame-rate is suffering.

As I said above, the size of your monitor can affect how noticeable aliasing is, which gives an added benefit to 1440p monitors. A 27” 1440p monitor will have more pixels than a 27” 1080p monitor, of course. More pixels per inch means smaller pixels, and smaller pixels means less noticeable aliasing. It makes 1440p a more viable option if you’re interested in a larger screen for more immersive gameplay—while you’ll ideally need a powerful GPU for that resolution, you can greatly reduce or disable antialiasing in modern games with no significant reduction in visual quality, and that gives you plenty of headroom to max out other settings.

The topic of antialiasing options can be a guide in itself. For an in-depth look at the types of antialiasing options and how they affect performance, read this page from tweakguides.com: LINK, as well as this comparison guide from benchmark3d.com: LINK


====Anisotropic Filtering====
While antialiasing improves edges, anisotropic filtering (AF) improves surfaces. 3D objects in a game contain 2D textures, which can vary in appearance depending on view angle and distance. In short, AF alters the filtering pattern of textures to make them less blurry at a distance or at an angle. I won’t spend much time on this because it’s generally an easy setting to keep enabled or at its max sampling rate. It usually does not cause nearly as much performance loss as hardware-based antialiasing.

====Texture Quality====
This is by far the most important setting in the game. I think this image pretty much sums it up:


Cropped screenshot from Skyrim: High and Low Texture Quality.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

Before we move on, here are a few things to consider about the current state of video cards:

> New cards are just around the corner – If you’re willing to wait for the latest and greatest (and you have the extra cash to spend), NVIDIA will be releasing a new generation in early–mid 2014. NVIDIA is also set to release an upgraded version of the 780, dubbed the 780 Ti, in November. The 780 Ti is rumored to be faster than the Titan and priced around $700.

> Video Memory (VRAM) is becoming a stronger point of consideration – I can’t argue with the fact that most of today’s games, even at 1440p, don’t necessarily require more than 2GB of VRAM, but I do know that with the release of the PS4 and XBox one and HD gaming becoming the norm, games are gearing up to be memory hogs by design. Remember, many popular PC titles are console ports, which were originally designed to use 512MB VRAM max. The next-gen consoles are going to have access to several GB of memory for their GPUs (even after you factor in the memory for the console’s OS, networking, and the other background functions). Those PS4 and Xbox One ports will be part of our list of things to play.

====///====Choosing the Right AIB Partner====\\\====
Oh, you thought we were done? Once you’ve chosen between NVIDIA and AMD, it’s time to choose the right brand. AMD and NVIDIA sell their GPUs, as well as a reference design of the full card, to their Add-in-Board (AIB) partners. Major partners include ASUS, EVGA, Galaxy, Gigabyte, HIS, MSI, PNY, PowerColor, Sapphire, VisionTek, XFX, and ZOTAC. Not every company is known to sell both AMD and NVIDIA.

AMD and NVIDIA leave it up to those companies to decide how they want to market their video cards. In general, you can expect the following variables:

====Cooling====
Video cards come with a wide range of pre-mounted cooling options, depending on the brand. In many cases, the cooler is the only variable worth considering when deciding between multiple cards with the same GPU.

###Reference Coolers###



As shown above, reference coolers are usually a GPU heat sink and a single fan housed in a plastic shroud. It’s called a reference cooler because it’s fabricated based on a design by the GPU manufacturer (AMD or NVIDIA), meant to provide adequate cooling to keep your GPU at the specified “safe” temperature. For the most part, it can be an effective cooler. The main drawback is that the fan can get loud at higher temperatures, and turning it down means potentially bringing your GPU to a temperature threshold where throttling (automatic GPU speed reduction) kicks in to protect the GPU from overheating. Thus, performance suffers. The fan noise is less of a problem if you game with headphones, so don’t let the noise deter you if you can find a good deal on it.

###Aftermarket/Custom Coolers###
For a better/similar cooler and less noise, you may want to opt for a model that ships with an aftermarket cooler, such MSI’s Twin Frozr, shown below:



It should be noted that no matter what cooler you choose, you will often have to set the fan profiles yourself. GPUs have target load temperatures set by default (usually around 80C), and your GPU cooler fan(s) will speed up and slow down accordingly. You might still see your load temperatures at 80 with an aftermarket cooler, but the cooler maintains those temperatures much more quietly.

You can use various software solutions to adjust your fan profile to set a lower load temperature goal, such as 65-70—the cooler will work harder and louder, but will still be quieter than a reference cooler trying to maintain those same temperatures.

There is a tradeoff to using custom coolers. Reference coolers often use a loud, fast blower fan that exhausts the GPU heat out the back of your case. This method is very effective, but only if you crank up the fan. The custom coolers, while quieter and overall more effective at cooling your GPU, usually dissipate heat inside the case rather than outside. This, of course, can add heat to other components such as your CPU, but an effective CPU cooler and proper case fan arrangement can offset this. See the diagram below:


This diagram represents a typical ATX case containing an aftermarket CPU cooler and three 120mm case fans for front intake, rear exhaust, and top exhaust.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

The CPU

In the graphics card arena, the AMD vs. NVIDIA war can get ugly. But it still doesn’t hold a candle to the vitriol exchanged in CPU debates—AMD vs. Intel. Those are the two CPU companies under consideration here, and I won’t attempt to convince you that one is better than the other in all situations. My goal will be simply to educate you on why it’s not that big of a deal in a gaming PC at the budgets we’ve set here. For that reason, I won’t spend much time discussing all of the different CPU options out there.

First, let’s clear the air regarding any preconceived notions about clock speed. It’s a common argument used among uninformed individuals who claim that AMD’s $200 4GHz CPU outperforms Intel’s $320+ 3.5GHz CPU. False.

When you consider everything that makes up a “CPU” (cores, cache, pathways, registers, etc. etc.), there's no point in comparing clock speeds. Even fairly educated enthusiasts, me included, find it difficult to make a purchase decision based solely on a CPU’s technical specifications. It’s far more practical to compare CPUs by looking at quantifiable real-world performance, benchmarking, and pricing.

====///====Working Harder vs. Working Smarter====\\\====
When a processor does its job (the workload), it breaks the load into stages/phases, then further divides those stages into units that can be processed more than one at a time. Software might provide commands/instructions in a linear manner, but a CPU can prioritize those instructions differently. Let's say a program sends out instructions X, Y, and Z. The CPU can divide those instructions into subunits (say, X1, X2, X3, Y1, Y2, and so on) and execute them in the most efficient order (which might be X2, Z1, X3, X1, Y2, etc.). That’s an oversimplified example, of course. It’s much more complex in reality.

The design of the CPU makes the clock speeds unrealistically comparable between different CPUs. It depends on how much can be done in what order, the divisions and subdivisions of labor it can provide, and the speed at which each of those units/subunits can be executed (some of it has a lot to do with the register’s capability, which stores commands, and the cache, which stores data associated with those commands).

Perhaps more difficult to explain is the difference in designs that give one CPU the ability to break down commands into more and therefore smaller subunits than another CPU because it extends beyond the number of cores. However, the end result is that the clock speed dictates how quickly those subunits will be executed, and the architecture/design dictates how many subunits can be executed per cycle—or, instructions per cycle (which is a blurrier distinction because I'm not an engineer). More subunits per instruction means smaller workload per unit, which means faster processing of that subunit at the rated clock speed. On top of that, less workload per subunit also means more subunits per clock cycle.

At this point, Intel takes the lead in doing more per clock cycle, and the differences can boil down to material choices, die size, transistor count, and overall design. But it can sometimes require a trade-off. More work at a slower pace, or less work at a faster pace? It largely depends on how efficiently the software can use the architecture, which is why CPU specs must be taken with a grain of salt and it is necessary to examine relevant benchmarks. And what benchmarks are relevant to us? The gaming ones, of course. More on that later.

====How Many Cores?====
So, what about core count? That 4GHz AMD has 8 physical cores, and the Intel 3770K/4770K only has 4 physical cores.

More cores are certainly better, but there is sometimes a misconception about AMD’s core count. The current 8-core AMD FX series are not exactly “true” 8-core processors. Their architecture uses what’s called a modular design, with 4 modules containing 2 cores each. Each of those 4 modules contains only a single floating point unit (FPU), which is responsible for carrying out computations. The cores on each module must share the FPU, as well as other components with the module, such as its cache. Granted, this does not mean it’s really a 4-core CPU. It still has 8 physical cores; they are just somewhat crippled by the need to share certain components.

A 4-core Intel CPU with hyperthreading has 4 physical cores and 4 virtual cores. As I said before, Intel’s current CPUs are able to do more work per clock cycle per individual physical core. When you add hyperthreading to the mix, you completely optimize the work output from those physical cores. Hyperthreading actively searches for any core that is currently in an idle state waiting its turn, and uses that idle core to immediately process the next set of instructions. It ensures that no cores are needlessly idle in applications that are actually coded to acknowledge multiple cores, or in general multitasking.

If we were building a 3D modeling, video editing/encoding machine, a quad-core hyperthreaded Intel i7 would perform better than AMD’s 8-core FX CPU and use less power—but it would also cost over $100 more than the AMD. However, since we’re building a gaming PC, all we really need is a CPU that won’t be the performance bottleneck in the games you enjoy.

====///====What Makes a CPU Bottleneck Game Performance?====\\\====
A CPU bottleneck means the processor is holding back the graphics card from rendering more frames per second. Meaning, the graphics card is not living up to its potential, and therefore neither is your gaming experience.

This bottleneck can occur because not only is your graphics card working hard to render what you see on the screen, your CPU is handling what’s actually happening in the game. CPUs are particularly good at handling physics, so that would include character movements, explosions, crashes, shooting, etc. in a dynamic environment.
Some games are CPU-intensive simply because they were coded that way. Skyrim, for example, relies heavily on your CPU because certain rendering components (such as shadows) were coded for the CPU.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

====What Kind of CPU Will You Need to Avoid a Bottleneck?====
For most games, not much, as the GPU is the primary source of processing power. That’s the short answer. A better answer is “It depends.” Look at these results published by The Tech Report (techreport.com) in 2012. They tested 18 different CPUs on 4 popular games using a Radeon 7950 (a fairly high-end card, as you already know).

First, let’s look at Batman: Arkham City, a GPU-dominant game with a lot of physics involvement. It’s also a game known for being poorly coded for PC.



Most of these CPUs are a bit dated, and I’ll add up front that the CPUs I will suggest for the sample budgets will have performance levels around the 2400–3570K in gaming.

To put things into perspective, the top CPU in that chart (the i7-3960X) is a $1,000 CPU, and the one right below it, i7-3770K, is a $300 CPU. Go down two rungs to the i5-3570K, and you have a $200 CPU. While those expensive CPUs are incredible pieces of hardware for other applications, they just don’t have much of an added benefit in PC games.

The Tech Report’s test also measured frame times (frame latency), which showed improvements in direct correlation with frame rate performance (i.e., the better CPUs had higher FPS and lower frame latency). I mentioned frame latency briefly in an earlier section. To read more about it, check out these links: Extremetech article and Tech Report article

Now, let’s look at the results from a notoriously CPU-dependent game, Skyrim, including a similar test conducted by Tom’s Hardware:



Tom’s Hardware used settings that are more aggressive and tested weaker CPUs, which seemed to cause a disparity between the top-end CPUs and some of the lower end chips, including the i3-2100.

Next up, we have the results from Crysis 2. The Crysis series are known for their GPU-intensive engines. At amped up settings, they’ll bring even the strongest GPU to its knees.



The Crysis 2 results are interesting, given that the CPUs ranging from $200 to $1,000 have virtually zero performance difference in such a GPU-intensive game. Now, look at Battlefield 3 (single player). The performance is nearly identical across every CPU.



The multiplayer mode of Battlefield 3 is far more CPU-intensive, however, due to the dynamic environment, and the fact that you can have up to 64 players at once in a server (requiring lots of real-time physics calculations).

Have you noticed something about all of these results? The Intel CPUs come out on top. Every single time. As I said, Intel’s architecture in those CPUs allow for far better performance per core. This still holds true in today’s CPUs despite AMD’s improved architecture. Tests are showing Intel’s i5-4670K ($220) outperforms AMD’s FX-8350 by nearly 50% in single-core performance. However, the 8350 (an 8-core CPU) is slightly ahead of the 4670K (a quad-core CPU with no hyperthreading) in multi-threaded applications. Basically, the more threads an application uses, the less of a performance advantage the 4670K will have over the 8350.

So, what does that have to do with gaming? Right now, not much. Many games are single-threaded (meaning, they only send one set of instructions to the CPU at a time), and even the multi-threaded games are only using 3 to 4 threads at a time to process in parallel. Therefore, multi-threaded games are not even using all 8 of the AMD FX-8350’s cores for there to be any advantage.

So what this ultimately translates to is negligible performance differences between these two CPUs for gaming. This is obvious for GPU-dependent games, but even in CPU-intensive games, the GPU is still doing enough work to make up for a lot of the difference in per-core performance between the two.

There are some who may argue that the next generation of consoles will bring more games coded for 8 cores—the reason being, the new consoles have 8-core CPUs. To be honest, I don’t see that happening in the near future. It’s been stated in numerous articles that the 8-core Jaguar CPU shipping with the consoles is a weak one, amounting to only about half of the performance of an FX-8350. It’s likely at least half of the cores will be devoted to the operating system and processes that must constantly run in the background. That essentially leaves a disproportionately weak quad-core CPU for gaming, meaning that coding games to take advantage of GPU power will still be important, perhaps even more important than ever.

However, I’m not saying that games won’t be coded to use more threads. On the contrary, since the console CPU is so weak, taking advantage of multi-threading will also be important, and this will certainly extend to benefit multi-core utilization on PCs as well. But this is good news for anyone with a modern PC, and especially great news for anyone planning to build one. This can mean more GPU-intensive games coupled with optimized code to take better advantage of any multi-threaded CPU.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

The Cooler

As you probably already know, processors need direct, active cooling solutions. We already discussed the types of coolers you’ll find for graphics cards, so now it’s time to consider the CPU cooling solutions. With only a few exceptions, retail box CPUs come with a cooler that’s designed to just barely keep your CPU at a safe temperature under a realistic load.


AMD stock cooler for AM3+


Typical Intel stock cooler.

The fans on these coolers are often quite loud at higher temperatures and sometimes not even adequate for a full 100% load on the CPU, even at stock speeds. This is especially true with Intel’s newest line of CPUs, which produces more heat because of the smaller die size and more integrated components. For any CPU, the stock cooler is mainly used in productivity environments—home/office with light workloads that don’t stress the CPU for an extended period. Most off-the-shelf computers at Best Buy, Dell, etc. are going to use this cooler.

Components in a gaming PC need to withstand high usage for hours on end without throttling due to excessive heat. Since your graphics card will be contributing to a higher ambient temperature inside your case, the stock CPU cooler becomes even less of a viable option. Granted, many people still use the stock cooler in their gaming PCs with any problems, but most experienced builders will recommend an aftermarket cooler of some sort. It will extend the lifespan of your CPU and allow for some mild to moderate overclocking when you are ready for it.

There are too many coolers on the market to discuss all of them here. There’s a wide range of prices on air coolers, as well as a wide performance gap with a point of diminishing returns. If you’re curious about the options, I would suggest you Google “Best CPU Cooler 2013” or something similar. Then, get comfortable and prepare for a few hours of reading test results, reviews, more test results, more reviews, followed by even more test results that conflict with the results you’ve already read. PC cooling is a hobby in itself, going hand in hand with overclocking. The options out there are incredibly vast, too extensive to cover in this guide.
Instead, let’s keep it simple. For all three sample budgets used in this guide, I recommend the Cooler Master Hyper 212 EVO or the Hyper 212+.

Both coolers have repeatedly proven to offer the best cooling performance for the price. They’re priced around $30, but the Hyper 212+ can sometimes be had for $20 with a rebate or promo. While the EVO is the newer and marginally better performer of the two, you’d be safe choosing whichever model is cheaper at the time.

Here’s a Hyper 212 EVO compared to an older stock AMD CPU Cooler:



Yes, it’s a large cooler. It’s not even the largest one out there, but you will still need to consider the size when choosing a case. The case will need to be wide enough to support the cooler’s height—generally an 8” wide case will be fine. We’ll discuss this further in the Case section.

Some people have had difficulty understanding the instructions that come with the cooler. If that’s the case, installation instructions can be found all over the web. Here’s one on overclock.net that covers both Intel and AMD mounts: LINK

====///====Thermal Paste====\\\====
When installing any cooler, you must use thermal paste, also called thermal interface material (TIM). If you want to know why, the thermal paste guide at techpowerup.com offers a good explanation: LINK

Every CPU cooler ships with thermal paste, including the stock CPU coolers (the paste is pre-applied to those types). There are dozens of thermal pastes available on the market, and there can be a significant performance gap between the best thermal paste and the worst—but as you’ll find in your research, the difference between the best and the 15th best thermal paste is only a degree or two. Performance is determined by load temperatures using each type of paste in the same cooling environment.

Luckily, the Hyper 212 EVO comes with decent performing thermal paste, but you will have to apply it yourself. There are several schools of thought when it comes to ensuring application of a thin, even layer of thermal paste. The Newegg build tutorial I linked earlier in this guide will show you one method, but I don’t recommend it. For other methods, I recommend watching this video: LINK

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

====Northbridge and Southbridge (Chipset)====
The Northbridge is the chip that generally regulated crucial functions/components such as the CPU, memory, PCIe, and front-side bus (the pathway that connects the CPU to your other components). The Southbridge is a slower chip that interfaces with the Northbridge and handles basic input/output functions such as USB, audio, regular PCI bus, Ethernet, and others. Nowadays, many of the Northbridge controls have migrated to the CPU, and Southbridge chips have been removed, integrated, or renamed/reconfigured. For example, Intel CPUs now have an integrated memory controller and PCIe controller that comprise an integrated Northbridge chip. AMD CPUs also have integrated memory controllers, but still rely on a Northbridge for PCIe functions.

The functions traditionally provided by the Northbridge and Southbridge chips are now more commonly referred to as the chipset. Given the role that the chipset plays in regulating your hardware, it’s important to make the right choice (if the CPU is the brain of your PC, the chipset is the spine). There are multiple chipsets available for each socket, but the choice is easy for a gaming PC, as I’ll explain below.

###Intel Socket 1150 and Socket 1155 Chipsets###
The desirable chipsets for Socket 1150 and 1155 are Z87 and Z77 respectively. Here’s why:
> Full overclocking support for unlocked CPUs – you’re going to want this when you decide to learn about overclocking.

> Support for SLI/Crossfire – later down the road as games evolve and become more demanding, adding a second graphics card is an easy, cost-effective performance boost. You’ll need a chipset that supports it.

> More SATA 6 GB/s ports and USB 3.0 ports – not quite as important, but nice to have the extra ports for possible expansion later on.

> 2 DIMMs per memory channel – most, but not all, of the chipsets support this, but it’s handy to be able to run up to four sticks of RAM in dual channel, so you can later add more RAM by simply adding two more sticks of what you already have, rather than having to replace them.

There are several more things unique to the Z-series chipsets, but the above features are the ones you’re most likely to care about. The other available chipsets (H87/H77, B85/B75, and more) lack one or more of the above features and are generally more suitable for HTPCs and budget home/office systems.

And if you’re wondering, the main difference between Z77 and the Z87 is that Z87 supports more native SATA 6GB/s ports (more on that later) and more USB 3.0 ports. This shouldn’t influence your decision, because the two chipsets support entirely different CPUs. If you want an i5-3570K, you’ll be using a Z77 socket 1155 motherboard, and 4670K CPUs will use a Z87 socket 1150 board.

###Intel Socket 2011 Chipset###
Intel’s enthusiast/extreme platform currently uses the X79 chipset, which has similar features to Z77—the key differences being PCIe lanes and quad-channel memory (more on those later).

###AMD Socket AM3+ Chipsets###
One great thing about AMD is backwards compatibility. They stick with the same socket for a long time, and bios updates allow today’s AMD CPUs to work with chipsets that were released years ago. In fact, Some of the older AM3 (non-AM3+) chips are able to run on AM2+ motherboards with a proper bios update, and likewise for AM3+ chips and AM3 boards, provided the boards can handle the current. With careful attention to detail, AMD makes it possible to build a competent gaming PC at very low budgets. But for an inexperienced builder, it could be more headache than it’s worth, and you run the risk of selecting a motherboard that’s “mechanically” compatible but cripples the performance of your CPU. We’ll keep it simple and somewhat painless by narrowing the choices to two chipsets:

990FX and 970
The 990FX is AMD’s “top” chipset. It has the highest level of support for SLI/Crossfire (PCIe 2.0 x16/x16) and overclocking. The 970, on the other hand, is a slightly more budget-oriented chipset. Most of them don’t support crossfire/SLI, but there are a few that support a PCIe 2.0 x8/x8 configuration, which would give slightly reduced performance over x16/x16 but still viable. Overclocking performance on the 970 chipset, as far as I’ve read, is not as good as a 990FX, likely due to the quality of the voltage regulator circuit design and other components being scaled down to cut costs. In fact, there are a few 970 boards that only support 95W TDP CPUs (i.e., FX-6300, 4300, 6100, and 4100). The differences in these two chipsets aren’t too big, but it’s generally a good idea to pair a 125W TDP CPU (the 8350 or 8320) with the higher-end chipset, especially if overclocking and/or planning to SLI/Crossfire in the future.

====SATA Ports====
Serial Advance Technology Attachment (SATA) is the modern interface used to connect your internal drives (hard drives, solid state drives, and optical drives). There are several types (revisions) of SATA ports, the main difference being transfer speed. Today’s motherboards include a mixture of SATA 3Gbps and SATA 6Gbps ports, explained below:

====SATA 3Gbps (SATA Revision 2.0)====
This SATA revision has an advertised max throughput of 3.0 gigabits per second (Gbps). Note, this is a bit, not a byte—similar to the way your Internet download speed is advertised. There are 8 bits in a byte, so an Internet download speed of 12Mbps has a max download rate of 1.5 megabytes per second (MB/s). But to further complicate things, serial protocols like SATA use a type of wire encoding called 10b8b that essentially means only 80% of the bitrate is devoted to actual data throughput—in other words, the actual max throughput of SATA 3Gbps is 300 MB/s. If you want to read the geeky details about this, have at it: https://en.wikipedia.org/wiki/8b/10b_encoding

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

====PCI Express (PCIe)====
The full name is Peripheral Component Interconnect Express. We’re going to call it PCIe. Wikipedia provides a geeky explanation of the interface, but for now, we’ll just focus on what it’s used for. PCIe interfaces with a variety of components that connect to your peripherals. It communicates with your chipset via “lanes.” A single lane consists of two pairs of wires/tracers. One pair of wires sends packets of data, and the other pair receives.

There are 4 bandwidth levels of PCIe, and they’re commonly known as x1, x4, x8, and x16. This indicates how many “lanes” have access to the PCIe slot. Technically, there is a different sized slot made for each of the 4, shown below:



Most motherboards are only going to have a combination of x1 and x16 slots. The x8 and x4 slots are rare to nonexistent these days. The reason for this is the x16 slots are physically compatible with cards that require 1, 4, 8, or 16 lanes.


A typical motherboard has a combination of PCIe x16, PCIe x1, and regular PCI slots. PCI slots are becoming more obsolete these days but are still used for various legacy needs, such as FireWire cards, and even some sound and Wi-Fi cards.

The x16 slot is of particular importance, because it holds your graphics card.

But it’s not enough to simply choose a motherboard based on the number and type of PCIe slots it has. The more important thing to consider is the number of lanes those slots share. There might be 16 lanes available for each PCIe x16 slot on a motherboard, but there are other factors at play that dictate exactly how many of those lanes can be used at once.

###PCIe Lanes and AMD###
In AMD motherboards, the number of usable lanes is determined by the chipset. AMD’s 990FX chipset provides access to a total of 38 PCIe 2.0 lanes. How they’re used depends on how many PCIe slots are being occupied. If you are running one graphics card, all 16 lanes of the slot will be used. If you add a second graphics card to an available x16 slot, that slot will also use all 16 lanes, for a total of 32 lanes—commonly referred to as an “x16/x16” configuration. Beyond that largely depends on the way the motherboard was made. If there are 3 PCIe x16 slots available, using three graphics cards will usually mean an x16/x8/x8 configuration.

The 970 chipset has 22 PCIe 2.0 lanes available. What this translates to is 16 lanes for a single graphics card, and the number of lanes for a second graphics card depends on the motherboard. The 970 standard only allows for an x16/x4 configuration, but motherboard manufacturers can include chips or switches that control the use and allocation of those lanes to allow for an x8/x8 configuration.

###PCIe Lanes and Intel###
In Intel motherboards, the CPU determines how many PCIe 3.0 lanes there are, and the chipset determines the number of PCIe 2.0 lanes. Intel’s Haswell and Ivy Bridge CPUs (e.g., 4670K, 3570K) allow 16 PCIe 3.0 lanes (and 4 PCIe 2.0 lanes in the Z series chipsets). A single card will get 16 lanes, and two cards would be in an x8/x8 configuration. Higher end Intel motherboards also contain chips that use an active switching function to effectively double the available PCIe lanes (synthetically, similar to the way hyperthreading works).

###Another Variable: PCIe 2.0 vs. 3.0###
While Intel has far fewer PCIe lanes at its disposal than the 990FX platform, there’s another variable at work here: PCIe revisions. Intel’s Ivy Bridge and Haswell CPUs support PCIe 3.0, and their respective Z chipsets support PCIe 2.0, so the Z87/Z77 motherboards use PCIe 3.0 slots for graphics cards and PCIe 2.0 lanes for everything else. AMD boards use PCIe 2.0 slots/lanes. PCIe lane speeds are measured in gigatransfers (GT/s), and PCIe 2.0 lanes have a transfer rate of 5 GT/s, while PCIe 3.0 lanes have a max rate of 8 GT/s.

###What Does this Mean?###
Right now, practically nothing. The cards on the market right now will fit in either slot, but they don’t saturate the full bandwidth of PCIe 2.0 x16, let alone PCIe 3.0. As it stands now, your single graphics card will perform the same on any of the platforms.

I would go into the technical details, but I’m assuming by now you’re willing to just take my word for it: If you were to add a second graphics card at some point, there would be no performance difference between PCIe 2.0 x16/x16 and PCIe x8/x8 except in the highest end graphics cards (such as Titans and dual-GPU cards), and only at resolutions exceeding 1440p. You would get a marginal performance gain (< 4%) at best moving from PCIe 2.0 x8/x8 to PCIe 3.0 x8/x8, and even less of a gain going from PCIe 3.0 x8/x8 to x16/x16 (available on Intel’s high-end socket 2011 platform). Ultimately, the 990FX chipset would have the slight advantage over Intel’s Haswell/Ivy Bridge due to extra lanes if you were going to use more than two graphics cards, but I have always subscribed to the notion that if you think you need more than two graphics cards at any point, it’s time to just upgrade to a more powerful single graphics card. The only situation that warrants three or four graphics cards are for extreme systems using the fastest GPUs available—systems that are likely to be using Intel’s enthusiast platform anyway.

Now, how long the difference will be negligible is up for debate. In single-card setups, which includes all three of our sample builds, it’s unlikely that it will matter anytime soon—not before you’re long overdue for a full platform upgrade, anyway.

Ultimately, I’d recommend a motherboard that supports a second graphics card*. For gaming performance, the most impactful upgrade will always be your GPU. People go through several graphics card upgrades without changing any other component in their system. Being able to simply buy another of the card you already have is convenient and cost-effective, and it’s only achievable with a motherboard that supports the second card (and a compatible power supply, which we’ll discuss in its appropriate section). A motherboard that fits this criterion should have 2 PCIe x16 slots that run in an x8/x8 or x16/x16 configuration. You can also use Crossfire in an x16/x4 configuration (but not SLI), with very slightly reduced performance but still effective.

*Note: Some motherboards only support Crossfire (AMD cards), and not SLI (NVIDIA). It’s important to read carefully if looking for a motherboard that supports both.

------------------------
++++ALERT: You are reading an out-of-date version of the guide and wasting your time. Read the PDF for the most accurate up-to-date info.It's best to download the PDF and use a proper PDF reader. Google's formatting of PDFs breaks all of the links. Link to directly download the PDF. I have stopped updating the text in the thread because the forum's limited code makes it far too time-consuming to change images and add text.++++
------------------------

====///====My Motherboard Picks====\\\====
$1,000 budget—ASRock Z87 Extreme3
Why the Z87 Extreme3: Since the CPU pick was a 4770K, we must, of course, go with socket 1150 and Z87 chipset. At the time of this writing, this motherboard was around $130, making it the cheapest Z87 motherboard that supports both Crossfire and SLI. It also has 6 SATA 6Gbps ports, decent audio chip, Intel Gigabit LAN chip, 4 USB 3.0 and 2 USB 2.0 ports, onboard headers for extra USB 3.0 and 2.0, a PS/2 port for good measure, and 8 power phases. All in all, it’s a well-rounded board with no superficial features that bloat the price.

$800 budget—ASRock 990FX Extreme3
Why the 990FX Extreme3: At this budget, we have room for a 990FX chipset with PCIe 2.0 x16/x16 support and better overclocking stability. And yes, it’s an ASRock again. I know it smells like bias, but at the time of this writing, this board was literally the cheapest 990FX board available. It supports SLI and Crossfire and has 2 USB 3.0 and 6 USB 2.0 ports, 5 SATA 6Gbps ports, 1 eSATA 6Gbps port, Broadcom Gigabit LAN, headers for extra USB 2.0, 2 PS/2 ports, and the same audio chip as its Z87 equivalent. A couple of shortfalls for this motherboard include lack of onboard USB 3.0 headers and only 4 power phases. The first issue can be solved with a cheap USB 3.0 card, while the other will require a different motherboard. The ASRock uses high quality voltage regulators, so mild/moderate overclocking shouldn’t be an issue. If it’s important to you, an extra $20 will get you the Extreme4, with twice as many power phases, 2 extra SATA 6Gbps ports, and onboard USB 3.0 headers.

$600 budget—MSI 970A-G46
Why the 970A-G46: I struggled between choosing a barebones cheap 970 motherboard vs. a 990FX board, as I made as many concessions as I could on other components so that the budget could accommodate the best graphics card possible. What I decided was that in lower budget systems such as this one, how it performs in games right now is more important than future-proofing. As you’ll see when all the part recommendations are compiled and presented at the end of this guide, the $600 build is really just a slightly scaled down version of the $800 build, replacing some convenient “enhanceability” in favor of functionality.

While its power phase design is at the low end, it is not as crippling of an issue with the 6300, which has a 95W TDP. This board does support 125W TDP CPUs, so the phase design should give you a little more overclocking headroom than a barebones no-frills 970 board. It’s also one of the very few 970 chipset boards that support an x8/x8 SLI/Crossfire setup. In addition, you get 2 USB 3.0 and 6 USB 2.0 ports, 6 SATA 6Gbps ports, a PS/2 port, the same audio chip as the two boards above, and onboard headers for USB 2.0. Not a bad motherboard for $80.