Choosing the right CPU for you can be complicated, but doesn't have to be.

First of all, clockspeed still does matter more than the number of cores [for games]. That said, in the past couple of years multi-core support has increased a thousand fold, and having 4 cores is actually good nowadays.

Now, when it comes between choosing for AMD or Intel, there really isn't much reason to go AMD. With AMD you can get more cores for less money, but even with that taken into consideration, Intel has something to offer for anyone at any price. There just is no reason to go AMD at all. The latest architecture (Ivy Bridge) and its predecessor (Sandy Bridge) are just that much better; at identical clockspeeds Intel will always win and they can reach the same speeds as AMD while overclocking.

The CPU with the most value right now is without a doubt the i5-2500K/i5-2550K or its Ivy Bridge counterpart, the 3570K. These are what the i5-750/760 were before that. Be careful when picking these, because there are versions without the "K" suffix. Those can't be overclocked (well, barely, not worth mentioning) - and before you say you won't overclock, you're an idiot if you don't. It's piss easy to do nowadays, with many motherboards being able to do it FOR you; a 1 GHz easy overclock is easy to achieve on air.

I advise against going for any dual-core CPU at this point. It's simply not relevant anymore.

Let me first get rid of some misconceptions in this area. First of all, bigger isn't better. The 850 watt units are no better than 500 watt units, provided you compare similar quality units. Second, you do not need more than 600-650 watts of power in any system unless you got multiple videocards, and even then it depends on which videocard you use. The recommendation on videocard boxes are pure bullshit, those are aimed at terrible PSUs that are inefficient. In fact, 500 watts is plenty for most systems, although you'll want to leave a little headroom (I'll get to that in a bit).

First things first. There's a certification standard called 80PLUS. 80PLUS is an initiative that rates PSUs based on how much of their rated power they can deliver under 20%, 50% and 100% load. There are 5 different labels:
- 80 PLUS: means the PSU delivers 80% efficiency at all load levels
- 80 PLUS Bronze: means it delivers 85% efficiency at 50% load
- 80 PLUS Silver: means it delivers 88% efficiency at 50% load
- 80 PLUS Gold: means it delivers 90% efficiency at 50% load
- 80 PLUS Platinum: means it delivers 92% efficiency at 50% load

I have not included the 20% and 100% load figures, because it is very unlikely you'll ever reach those, you'll be sitting between 20 and 80 most of the time. As an example, my E8400 and HD5870 combined with my 2 hard drives pull about 300-350 watts under load in Crysis, which is probably the game that is most taxing on your system overall.

Now, what exactly does efficiency mean? Let's take a 650W unit (gives me nice round calculations). First some load numbers: 20% would be 130W, 50% would be 325W and 80% load would be 520W. The 80% load is what's generally considered to be the highest healthy long-term usage of a PSU - so if your target system will consume 500W under extreme loads, you'll want 500/0.8 = a 625W PSU. Next step is to explain efficiency: if you are putting a 300W load on a crappy PSU, it is very likely that it'll actually be pulling upwards of 400W out of the wall socket. Now, you may not give a shit about your electrical bill, but what you do care about is how long your PSU lasts, right? Well, if a PSU is pulling 400W out of the socket, but only delivering 300W to your system, it means that the remaining 100W have to be dissipated in other ways. Generally, that means it's coming out as heat. You can already see where I'm going here; the hotter your PSU gets, and the longer it is exposed to such heat, the quicker the parts inside will degrade.

This is why efficiency is so important. Not because it'll save you some money on your electric bill, but because it'll simply make your PSU last longer. Not only does it generate less heat, the parts inside of a high quality PSU are of a lot higher quality and can withstand heat longer to begin with. Even with a basic 80+ 600W unit, if you're putting a 300W load on it, it'll generate 75W of heat; going up one step to Bronze already knocks that down to 53W of heat, and so on.

Another misconception is that all the brands make their PSUs themselves. On the contrary, very few brands do. To take a bunch of popular and common brands as an example:
- Antec PSUs are manufactured by Seasonic, Deltra Electronics, FSP or Enhance Electronics
- Be Quiet! PSUs are manufactured exclusively by FSP
- Corsair PSUs are manufactured by CWT or Seasonic; with one exception, the AX1200 is manufactured by Flextronic (but that's a useless PSU anyway )
- PC Power & Cooling have theirs made by Win-Tact (Turbo-Cool series), Seasonic (Silencer series) or Sirtec (Silencer Mk II series)

It gets worse. Taking Corsair's most popular HX series as an example, their HX650 is made by Seasonic, but the 750/850/1000 units are made by CWT! I deliberately took Corsair as an example, because I've seen a lot of "Corsair = Seasonic" on this forum, and that's complete and utter bullshit. They have 5 units in production that come from Seasonic: VX450, TX650, HX650 and the AX750/AX850. All the other ones are CWT, with the exception of the AX1200 as I said earlier. How do you find out who makes a particular unit? Check reviews, simple as that.

Next thing you'll want to look at, is modularity. There's a common misconception that modular PSUs are less efficient or unstable, neither of which is true. The big advantage to a modular PSU is that you don't have unused cables hanging around in your case.

To put this whole story into contrast, here's an example. Let's say you're considering a regular 80+ 850W unit and an 80+ Silver 650W unit. Now, as you can see in the links below, you're unlikely to reach 500W, even when running a good combo like 2 460's. If we take Hexus' example from below, putting you at roughly 400W under a heavy load from a game (Furmark is not a realistic load), that means you're sitting at respectively 47% and 61% capacity. See what I'm getting at? Simplifying the numbers a little bit, and rounding them off, the 850W unit will be generating pulling 400/80% = 500W, thus generating 100W of heat. The 650W unit on the other hand will be pulling 400/88% = 455W, thus generating only 55W of heat. The numbers I took here won't be this black and white in practice, but they'll be pretty damn close.

There are too many PSUs one can buy to make any specific recommendations here, and new units come out every month making older ones obsolete. Just decide if you want a modular unit, look at the 80+ label and its price, and if you think this is the one: Google it. A small note to be made here: there are a few sites that test PSU's under more realistic conditions than the 80+ certification. An extremely good one is www.jonnyguru.com - that's the type of PSU review that tells you all there is about a PSU. Why is this important? Well, the parts inside a PSU perform slightly worse in the real-world; for the 80+ certificate, they're tested at room temperature. Inside your case however, the temperature is slightly higher.

Some reading material:
Bit-Tech: Does anyone really need a 1kw PSU?
I took this one because it's very recent, but are tons of articles like this one out there. Basically, they use an Intel i7-980X with a GTX 480. Underload, it reaches 507W - on a 550W PSU that means it's sitting at 92% usage. Now, that kind of usage isn't healthy, but with any of the hardware I've been recommending your power figures will be lower. Taking my most power-hungry recommendation for the CPU, the i7-930, total usage will already be 50W lower. 507 - 50 = 457 => 457 / 550 = 83% usage. If you follow my recommendations for the GPU as well you can easily drop another 80 Watts even when going for the 5870. Hence my recommendation not ever to go over 650 Watts, unless you plan on running SLI/CFX, and then most specifically with GF100-based videocards (GTX 465/470/480).

Hexus: ZOTAC GeForce GTX 460 1GB in SLI - beating up on GTX 480
GPU review, but putting this here as a very recent example of power consumption. The only thing their figures lack is an indication of noise/fanspeed with the temperatures, but their power figures show what power consumption is like really for a modern system. Tests were run on an i7 965 as well, which consumes a good 50 watts more than an i7 930, which is my most power consuming suggestion for CPU.

Probably the most contested piece of hardware in any system, because some people swear by AMD and others swear by Nvidia. Let me state upfront that in the 15 years I have been building my own systems, I have had more Nvidia cards than AMD, because AMD in the past were very erratic in how their products performed and their drivers were quite terrible. The best product I've ever had was a Voodoo 1, because you didn't have to give a fuck about what driver you used as long as a game ran on OpenGL, allowing you to use 3DFX's GLIDE wrapper. Unfortunately, those days are gone and we are now stuck in the Direct3D era

On the driver front, Nvidia still have the better control panel. It has a few options which are missing from AMD's CCC and it's just generally easier to use. However, when it comes to performance and stability both companies are on par. That's possible because AMD have separate development teams for the control panel, the performance, Eyefinity and now HD3D. More on what each brand can or can't do below my recommendation list. Note that if you use Linux like I do, AMD's drivers are substantially better, at least when using Ubuntu. Nvidia's Linux drivers are horrifying, not even able to accelerate Flash. On both Windows and Linux, AMD's multi monitor support is also still way better.

Now, I'm gonna get straight down to business here and just give you a list of what the best buy at specific pricepoints are. Do not take this list and make your decision straight away though, read my comments below it!

(I apologise for the pricing in euros btw, this is just how it stands right now in most of Europe and US pricing is completely irrelevant to most of the world as it's all cheaper there)
- Budget, up to €100: AMD HD6790 or HD6770
- Budget 2.0 up to €120: AMD HD 6850
- Mid-end pt. 1, up to €160: AMD HD 6870
- Mid-end pt. 2, up to €220: AMD HD 7850 (or 6950 if you can find one)
- The odd end, up to €260: NVIDIA GTX 570 or AMD HD6970
- High-end, up to €350: AMD HD 7950 - until the GTX 670 drops to match.

First things first: if you're on any kind of budget, AMD's your choice. Nvidia has nothing under 120 EUR worth speaking off, the 560 loses to the 6870 on every metric and the 560 Ti gets destroyed by the 7850/6950 at their current price points. Nvidia are late, they still have nothing new to offer under €370. Below €220 they just have nothing interesting to offer, period. I'm sorry if you're a hardcore Nvidia fan, but if you go for an Nvidia card under €220, you're an idiot.

The €220-€260 slot is an interesting one. All of last generation's top cards have dropped to this price range. The 580 hasn't and won't, but the 570 and 6970 have. As before, these 2 are dead even performance-wise, so go for whatever deal you can find. In the slot up to €350, we've got another interesting situation. As it stands, the 7950 is your only viable option there (it's well ahead of a 580), but I expect the 670 to drop a bit more. Once it does, that's your choice there. If your budget can go up to €370, grab a 670.

Above €350 things are changing too much right now. The 680 and 7970 are tied in performance -assuming you overclock, which is again piss-easy-, but they keep dropping in price. At the time of writing, the 7970 is €80 cheaper at ~€400 than the 680 at ~€480.

As always, I advise against multi-GPU setups. If you must, a pair of 6870s or 7870s still deliver stellar performance.


And what about CUDA, PhysX, Surround/3D/Eyefinity?
Let's tackle the easiest one first: 3D. In short, just don't care about this right now. There are a whopping 10 games that support Nvidia's 3D Vision, and with that you need a whole set of hardware that has properitary Nvidia stuff in it. Now, recently, both sides have started supporting normal 3D hardware. 3D Vision does not work well (if at all) with regular 3D TV's and monitors, yet that is where the future lies. Give it another year or so for more 3D hardware to pop up, and 3D will get interesting.

As far as multiple displays go, on AMD can it run off a single card, and it's a hardware implemented solution. Nvidia needs two cards in SLI, and it's a software implemented solution. The implementations hardly differ, but AMD has an advantage because they can do it off a single card. Against popular belief, there are actually a ton of games that can handle this off a single card as well. The type of games where this technology has the biggest effect are racing games, and the 5870 has already proven to be sufficient to run games like Dirt 2, GRID and NFS: Shift across three displays at playable framerates even with 4xAA. That said, if you're looking at a multi-GPU gaming setup, it all boils down to SLI and CFX scaling. In that department, AMD have the lead with the HD6xxx series. Those cards achieve upwards of 80% scaling in a lot of games. Again though, either route is a good choice, because the 570 is an excellent card, and both sides do display spanning really well.

As far as CUDA goes, if that's something you really use then you just need to go for Nvidia. OpenCL is still a small child. However, I sincerely doubt you use CUDA unless you use some serious professional software - at which point you'll have a Quadro or something anyway . The only piece of software you might be using that uses CUDA is CoreAVC, but to be quite honest, that's mostly needed for low-end systems (HTPC's), where Nvidia's 9300/9400 rules anyway. If you use Photoshop with very large PSDs, you may also see some benefit from CUDA.

PhysX is another huge point of discussion on this forum. Nvidia fanboys swear it's the most amazing technology around, but if you look at it objectively there's not even a handful of games that really use it. The few games that do use it to create small shitty ass particle effects, which you most likely won't even notice are modeled better. One game in particular, Batman: Arkham Asylum is used as THE ultimate PhysX benchmark. The thing is, that PhysX in that game is still not used for anything gameplay related, because PhysX is such a platform limited tool. Moreover, the effects they used PhysX for in Batman: AA could've been achieved without PhysX as well, as static animations - and noone would've ever cared. Taking Mafia 2 as another popular example, I personally find PhysX in that annoying, as it takes away from the realism. In reality, you can't shoot at a concrete pillar with a 9mm, taking out a chunk of 2 litres and have that splinter into 9000 pieces. I find it ridiculous to watch, and I personally deliberately turned off PhysX because it just looks terrible. If you really are that adamant about PhysX though, they still recommend to have a dedicated PhysX card even when you are already running on Nvidia - and Batman: AA has proven that a dedicated card does indeed yield better framerates. So for this point, you should consider using a €70 9800GT or €90 GTS 250, regardless of whether you've got AMD or Nvidia.

Some reading material:
Anandtech's GPU Bench
Technically not reading material, but you can use this to verify most of what I've claimed so far.

Their selection of games is limited at best, but they're working on it. Just keep in mind not to just look at one aspect if you're comparing 2 cards: a load temperature of 80 for two cards can still mean that one card is 10 dB louder for example.

Also, don't use the CPU bench here - for gaming anyway. Their gaming tests deliberately use CPU-heavy games in which they completely eliminate any GPU bottleneck.

I've decided to drop this section for now. Prices are fluctuating far too much. Any videocard I've recommended above -including the low-budget ones!- can handle up to and including 1920x1200 fine in most games. For the heavier games (Crysis, BF3) you'll need to look at the mid-end recommendations.

I'll take this one quickly, because I can be quick and short here.

If you don't really care about sound quality, just stick with onboard sound, it'll do you fine.

If you do care, get ASUS' Xonar D2 or Creative's X-Fi Titanium. The latter is a bit cheaper, but both are the most reasonable choices. I personally have bought an X-Fi about a year ago after running on onboard sound for a good 8 years or so. I honestly didn't think I'd notice it, but I needed a separate card because the chip on my motherboard had died. The difference is immense, it really is. It's hard to explain this or back it up with articles from the web, so just take my word on it. I'm not one to put useless hardware in my system, but I'll definitely be putting a proper soundcard in any future system. Apart from the obvious audio quality, Creative have implemented some neat tricks in the driver which are noticeable in games.

I can't stress it enough, but for this piece of hardware you _really_ need to read reviews. There are thousands of motherboards available at any given time, many of them being slightly different revisions of a particular type.

What I can do however, is advise you on what _not_ to care about.

For instance, don't give a fuck about SATA3 unless you're going to spend hundreds on a high-end SSD (read the hard drive section for more info on them). Nothing needs this bandwidth yet, so it's a waste of money.
Also don't care about USB 3.0: Intel is cockblocking it until 2011, and AMD aren't going to have USB 3.0 support in their chipsets for a while yet either. Those 2 factors alone, along with some others which I won't bore you with, will make sure USB 3.0 won't be mainstream for another 2 years or so.

Now what do you want to care about?
PCI-Express 2.0 slots. These slots are rated at a speed of 1, 2, 4, 8 or 16: you'll want at least one PCI-E 2.0 x16 slot for your videocard. Now here's the tricky part: if you're going to run an SLI or CFX setup, you'll need 2 of these slots, but contrary to popular belief you hardly lose performance by going from 2 x16 slots to 2 x8 slots. Don't shell out tons of cash just to get two x16 slots, the performance you lose is 1%, maybe 2% at worst.
Also don't stare yourself blind on the onboard sound; some motherboards claim to have amazing sounds, but in reality all motherboards use an 88x chip from Realtek, which is still not as good as ASUS' Xonar or Creative's X-Fi. They're all virtually identical, for serious sound quality improvement you want one of the two mentioned cards. There are some high-end motherboards which claim they come with a riser card that gives you EAX 5.0 and shit like that, but none of them are interesting either, don't pay the extra money for those. If you want better sound, buy a cheaper motherboard without that added bullshit, and spend the cash on a Xonar or X-Fi.

Apart from that, it all boils down to your wishes. Read reviews to see which motherboards overclock easily as well, as some have cheap hardware which causes stability issues when overclocking your CPU.

Some reading material:
TechPowerUp: NVIDIA GTX 480 PCI-Express Scaling
One of their few good articles. The conclusion of the articles says it all: dropping from 16x to 8x loses you 2% performance at worst, which at 60 FPS means 60/100 = 0.6 * 2 = 1.2 FPS. Hardly something to worry about, because even for SLI/CFX setups that have to scale down on both slots -meaning x8/x8- it means you won't even lose 5% performance. At 60 FPS that means it's not even 0.6 * 5 = 3.0 FPS.

HardOCP: SLI & CFX PCIe Bandwidth Perf. - x16/x16 vs. x16/x8
Another one, although they don't test x8/x8 unfortunately. Still, this shows that there is absolutely no point in wanting 2 x16 slots.
HardOCP: SLI & CFX PCIe Bandwidth Perf. - x16/x16 vs. x8/x8
Followup on the one linked just above, proves even further that there is no noticeable drawback to running x8/x8.

This is actually a really tough decision to make, as there's literally thousands of sets to choose from. So rather than looking at specific sets, I'll try to explain what you need to look at.

First of all, the amount. I personally have 4 Gigabytes in my system right now, and that's actually plenty when running Windows 7. I often have a bunch of instances of Visual Studio, some MS Office applications (most notably Outlook, which is open 90% of the time I spend at my computer) and Photoshop open. Depending on what I'm working on I might have Eclipse or Netbeans open instead of Visual Studio, but suffice it to say that when I'm working I have a good amount of apps running. With my 4 gig, that is not a problem at all. If you're into serious video editing however, you may want a little bit more, so depending on the motherboard you go for you could opt for a tri-channel 3x2 kit, or a dual-channel 2x2+2x1 setup. More than 6 gigabytes definitely isn't necessary, and 4 is plenty really.

Besides that, there's not that much difference between all the different sets. The more expensive versions run on lower voltages or have heatspreaders on them, but that's all for mega overclockers and enthusiasts. You'll also see a lot of timings mentioned, in the form of w-x-y-zz. I'm not gonna go into detail what they are or how they work exactly, but in short it's the delay between your computer wanting to work with the memory, and actually being able to do so. So how does this matter? It doesn't.

I'm not kidding. You're not going to notice the effect of better timings, unless you care about synthetic benchmarks. In the real world, this has virtually no effect on system performance. Not even memory hogging applications like video encoding software or Photoshop really care about these lower, you're talking about a few seconds difference at best. The only applications that do benefit from quicker memory are the ones you won't be running on a system meant to handle gaming anyway, so stop caring about it.

One important thing to know about is ECC though: ECC stands for Error Correcting Code. It's not something you really need in a desktop system, so again, don't pay the extra cost for this.

The sweet spots are DDR2-800 for DDR2, and 1066 C5/1333 C6/1600 C6 for DDR3. Why maybe 1600 for DDR3? Because of the overclocking headroom. Once you get into OC'ing, you'll notice that DDR3-1600 has some more convenient frequencies, and the price difference between 1333 and 1600 is pretty much negligible at the time of writing. As far as timing goes, both types of memory have been around long enough that their initial 'slow' timings aren't even available anymore, so don't worry about that.

Some reading material:
Techspot: Intel Core i7 memory performance
It's a bit old by now, but still very relevant. While in theory faster memory should give a lot of extra bandwidth, but in practice almost no applications (can) make use of this. In the end, that's good for us, because it means we don't have to spend 50% extra because there's some "SUPER-OC MEGA-BANDWIDTH" memory available.

Anandtech: Memory Scaling on Core i7 - Is DDR3-1066 Really the Best Choice?
Especially the game benchmarks in this one will be interesting for most of you, just check them out. You'll see that most games don't give a shit, and that you're talking about single-digit FPS differences. In some games however, it has a very positive effect on MINIMUM framerate, which is what we care about most!

This is also a fairly easy department, as there are only 2 companies that have consistently delivered the top mechanical hard drives. There is a new kid on the block now though, called the SSD.

Where traditional hard drives aren't very different from optical media or the good old LP record, the SSD is. Traditional hard drives basically have a few spinning disks, with a 'needle' going back and forth to read/write data. The principle is the same as that of the media mentioned earlier, it's just that data is stored using electromagnetism. SSD's are different because they use Flash memory to store data.

Even the worst SSD's are just as fast as the fastest mechanical drives; the good ones however obliterate hard drives in every single aspect: they access data quicker, can send this data out quicker and they aren't prone to mechanical faults. They do however have a more limited lifecycle (although this is not really relevant); the more writing is done, the more it degrades, dropping the speeds. A little trick has been invented to circumvent this somewhat, called TRIM. Rather than actually writing an empty block when you delete a file for example, with TRIM support it only flags that block as unused, thus avoiding a real write cycle, and prolonging your SSD's life. All of this is handled by the controller inside the SSD, which is actually more important than the Flash memory used inside, or its link to the outside world (the SATA connection).

Do you want an SSD? Yes you do, because they are possibly the biggest jump forward in PC's in the last decade. Now, the problem with SSD's are that they have a pretty high cost per Gigabyte. And with pretty high I mean that where you're looking at €0.04-0.05 for mechanical drives, a decent SSD sits at about €1.0

That said, you do still want one, but not for data storage. To explain this, I have to go in a bit deeper. I'll try to keep it simple, so bear with me.

When you load your operating system, regardless of whether it's Linux, Windows or OS X, all your boot drive has to do is access very small files in rapid succession. The same happens when loading applications such as Photoshop or Visual studio, or to a lesser extent, games (but these typically also feature large files). Because there are no mechanical parts and it's all just electricity, SSD's can access data so fast that it's not even measurable, it's less than 0.1ms; a good mechanical drive still needs anywhere from 12-24ms. For data storage however, you don't really care how fast your drive can find the file, since it's most likely gonna be fairly big anyway, so you'll be stuck on that file for hundreds of thousands of times longer than the access time.
In fact, you probably don't even care about how fast the file can be transferred, because chances are that it's either a file that can be transferred within one second (< 150MB) or one that you'll be streaming in parts (like music or video).

If you're not up to speed on this stuff, by now you're probably confused and thinking "wtf do I want then ?!". In short: An SSD to boot your OS and some important applications, and a mechanical drive for all other things.

There's a lot of competition in this area, but the popular and good ones right now: Crucial M4, Vertex 3/4 and Intel 520. The Samsung 830's are also good, but a bit more expensive. Any good deal you can get on any of these will be fine. Failure rates are pretty similar between them and you can't really go wrong with any of them.

Right then, here's the recommendations for mechanical drives. Samsung's Spinpoint F1 and F3 have proven to be very fast drives, specifically the F3 1 Terabyte is very fast. On the other hand you have Western Digital's Caviar Black drives. These are WD's top drives and just as fast. However, the thing with mechanical drives is that the speed is also dependent on the size. Bigger drives require more platters (disks) to reach their capacity, which makes them slightly slower. My recommendation are the F3 1 TB if that's big enough for you, or WD's Caviar Black 2 TB. Hitachi and Seagate also produce good drives, but Samsung and WD consistently outperform these drives and generally do better all-round.

A crucial, but often overlooked part is the case you put all this hardware in.

I'll start off with a link: http://www.bit-tech.net/hardware/cases/
Bit-Tech are not the most technical website out there, but their case reviews are the best around, and they have lots of them. Moreover, their forums are filled with absolutely fantastic casemodders, should you be interested in customising your case. Their reviews are just fantastic though, and they have the most consistent testing methods and setup. Rather than measuring actual temperatures, they measure how much temperatures inside the case fluctuate with a given setup. That approach is good because it is incredibly hard to have the exact same ambient conditions every time. Same applies for CPU coolers, their methods are sound and consistent.

When it comes to cooling, there's a few basic principles you need to understand. First of all, hot air rises. The hot air will accumulate at the top of your case, which is why most cases these days have some kind of ventilation there. Next up is the principle of air pressure. When you pump air into a confined space, pressure builds - as it builds, it will try to leave the case on its own, without any help. That is called positive air pressure: the pressure inside your case should be higher than outside, so that the air will naturally want to flow out the case. With these two principles in mind, the ideal airflow inside a case is bottom+front in, out at the rear top end. Most decent cases go by this principle already, just make sure to check the fan orientation (there's usually an arrow indicating fan turning direction, and one for airflow direction). Most cases these days also mount your PSU at the bottom rather than the top, making sure that whatever heat it produces doesn't hang around in the top; because the PSU is in the bottom, manufacturers can also include a fan at the top of the case!

Also note that a fan with higher RPM doesn't necessarily do better than one with lower RPM, if the slower one has bigger blades. The key element here is CFM, or Cubic Feet per Minute: the amount of air a fan can move per minute.

First thing you gotta do, is decide whether you want a full tower, or a midi tower. Midi towers are generally easier to get the airflow done correctly, because it's easier to build up positive air pressure. This isn't a hard given fact though, and there's a bunch full towers that do just as well as midi towers. Next you'll want to decide whether you want a toolless design or not; meaning you can pop in hard drives, cards and what not without having to use screws. I personally don't like these toolless designs, because they are often fiddly and don't feel as secure.

Now you can start looking at case design. Everyone's taste differs, so this is purely your choice. All I can do here is recommend you to look at Bit-Tech's reviews.

A few of the top midi-tower cases are Antec's 902, Lian Li's PC-P50 and Cooler Master's HAF 922. For full towers there's Antec's 1200 and DF-85, Cooler Master's HAF X and 690 II. My current case is the NZXT's Phantom, still loving it every time I look at it !

As for CPU cooling: stock cooling isn't all bad these days. Stock coolers do require you to clean them from time to time, but if you're just going to do a mild overclock you'll probably be fine. There are however two outstanding CPU coolers, which are hands down the best in the world right now. Thermaltake's Frio, which came out quite recently, and the older (but still goddamn amazing) Titan Fenrir. Both can be found for around 35 to 40 euros, and are both quieter and cooler than stock cooling. I highly recommend getting one of these, regardless of whether you go Intel or AMD. You could also opt for something more expensive like Corsair's H50, which can fit into your case's natural airflow if done properly, but it's not really necessary.