As ever you are getting more and more for less and less, so this section needs a bit of updating, so here goes. Furthermore AMD have released a half decent linup of Phenom IIs consisting of Tripel and Quad cores.

Budget conscious: AMD ATHLON X2 (mid range X2 costs £30, $40)
pros: price.
Cons: Absolutely bleeping bloody OWNED by everything else namely core2duo and Pentium E5xxx. A 2.4ghz Core2Duo IS BETTER THAN a 3.2ghz Athlon X2. The way Core2Duos are built and work is so much better, it makes up for the core2duo having a slower “speed” per price. USELESS FOR OVERCLOCKING (will not go beyond 3.3ghz. Even if you have a 3.2ghx to start with…)

Value for money: Pentium E5xxx & E6xxx/Core 2 Duo E7xxx
pros: These ARE genuinly cut down core2duos, so you get (almost) all of the Core2 goodness for a fraction of the price. Due to the lower FSB (bad) and less cache (bad) than a core2duo it actually overclocks BETTER. In the hands of an experience overclocker a 2.5ghz E5200 can reach 4ghz, at which point it will beat a fully blown core2duo (say 3ghz) in basic applications and games. In more complex applications (e.g. video editting) whilst still very fast, the smaller cache really does begin to hurt it (25% slower than an equally clocked Core2duo?)

Normal Budget (£400, $500) Intel Core 2 Duo/Quad
Pros: double FPS of an equivalently clocked Athlon X2. Overclocks amazingly (2.5ghz can reach 3.3ghz even from a novice)
Cons: Not very future proof for the price you pay. Of course they update them and new versions, but the "Core 2" architecture has been around for a while now...

High performence Intel Core2Duo/Quad
pros: Core2duo E8500 - E8700 (3.5ghz) and Core2Quad Q9550 and above. The actual proccessors cost the same as a basic I7 920, but the motherboards are £150 less, and the memory is half price too.
cons: Phenom II will equal or exceed the large majority of Core2Quads and E8xxx in games and especially more demanding apps. A Triple or Quad core Phenom is more future proof than an expensive and actually quite old Core2Duo/Quad.

High Performence AMD Phenom II 940/955
pros: High end Phenom IIs match I7 920 in games. Low MOBO costs. Very easy to upgrade (it will work in a 3 year old motherboard) they work with DDR2 and DDR3 RAM. Ok overclocking unlike any other AMD proccessor from the past 3 years, not quite as good as Intel though.
cons: errm... now that i come to think about it... apart from overclocking and the fact people are still somewhat wary of putting a lot of money into an AMD (who have been dominated by Intel for the past 3 or 4 years) rig I can't think of any.

Ridiculously Extreme Performance (£2k, $1.5k) Intel Core I7 (AKA “Nehalem”) 920, 940, 965, 975, 930, 950
Pros: Fastest processor in the world. 2.8ghz I7 is 20% faster than a 2.8ghz Core2Quad. Overclocks beyond belief. Base (2.6ghz) model can hit 3.5ghz in the grasp of a moderately skilled overclocker.
Cons: price. But if you are looking at this, it’s likely price means nothing to you. Base model = $285/£200. Core I7 Extreme = $999/£1300. The only problem: Price of motherboards. Base model motherboard £300+/$400+ Soon to be replaced by I7 930 (2.8ghz) 950 (3.0ghz) and 965 (3.3ghz)

click to enlarge, click again to zoom in?

How much do I need? How much can I have? How much is enough?



WTF Does PC2-6400C5 mean?!?!

This, actually quite simple once you get your head around it, model code scheme naming malarky tells you the:

generation
bandwith
speed
latency

The "PC" number at the front denotes the generation, e.g. PC2 = DDR 2, PC3 = DDR 3 etc
the 4 to 6 digit number in the middle is the bandwidth, in MB/s. E.g. the RAM in the above example has a bandwidth of "6400" MB/s or 6.4GB/s.
Since 8 bits are in one byte, we can work out the speed from the bandwidth. bandwidth (or middle number) / 8 = speed. For example 6400/8 = 800MHz
Some RAM, e.g. Corsair, will have an extra bit at the end "Cx" "C" stands for "CAS" and "x" is the time, in CPU cycles. E.g. "C5" = "CAS 5." This is getting quite complicated, and truth be told isn't worth worrying about for 99% of users. A lower CAS value is faster, but the difference is far to small to warrant the extra £10 they charge for CAS 4 RAM, instead of CAS 5. HOWEVER What is acceptable? For DDR2 you really want "CAS 5" (or lower, but as I said, not a big difference) and for DDR3 CAS 8 is acceptable.

(DDR = double data rate. So technically speaking, your 800mhz RAM actually runs at 400mhz, but can move 2 bits per cycle, hence is the equivalent of 800MHz)


DDR2 vs DDR3

As of 2005 DDR2 is “current.” Know-one can predict when DDR3 will take over the world (DDR3 will become mainstream when companies like Dell deem it cost-effective-enough to splash out on (Which will not be soon… at all) When loading big files, such as:

Turning the computer on and off
Loading programs (e.g. MS Word, photoshop)
Rendering (e.g. CAD, DCC, video editting, proffesional photoshop)
Loading game levels (e.g. when it loads a map in COD)
WinRAR

Then a big difference in speed (e.g. DDR2 1066 to DDR3 1600) makes quite a big difference in time (e.g. 20 - 40% faster depending on app) Time meaning how long it takes to load. I.e. it will load faster. However it makes 0 difference to FPS.



Recommendation:

DDR2 is silly cheap, DDR2 motherboards are tried, tested and cheap. Any dual core CPU, apart from Core I7, is designed to use DDR2 (core2duo/quad/athlon x2 etc will work with DDR3, but was not designed specifically for)

All DDR2 is silly cheap. May as well get the fastest of all the DDR2 because the price differences are so small between slow 533mhz DDR2, average 800mhz DDR2 and the fastest 1066mhz DDR2 (you can find DDR2 in 1250mhz, but only from one manufacture and it costs very silly money)

4gb G.Skill 1066mhz DDR2 can be had for £35 ($45)

Brands: If you overclock, or want good performance look for some of these:

Geil
G.Skill
Mushkin
Patriot

Fantastic price, best brand for overclocking. Full stop. G.skill/mushkin no questions asked (it will overclock 200mhz odd further than say Samsung) Needless to say each company does different models. So, if you are REALLY REALLY serious about overclocking, take a look at:

Corsair Dominator
G.Skill Black Pi
Geil Black Dragon

If you buy OEM, you have no choice and will get some cheap stuff with no name that will not overclock (i.e. it will just crash when you try to make it go faster) You will just get the “average” speed. Currently "average" means 800mhz (as of may 09) which is really no slouch it has to be said. However as with all computers, two years ago 800MHz was the utter bomb and expensive, now it is relagated to "average" and "cheap".

If you buy a custom built PC: 90% come with a brand name (elixir, Samsung, PNY) and will overclock decently. G.Skill, Geil and Mushkin etc as said own any other brand if you plan to OC. If it comes pre OC’d it will most likely have one of those said three top OC’ing brands.

1) Pick the size you want.
2) Pick the budget
3) The budget will define what sort of drives you are lookign at: 16mb cache, 32mb cache, RAID 0, 10k, 15k, SSD?

If buying to build yourself:

1) Make sure it’s at least a “7200rpm” hard drive
2) Make sure it’s at least a “16mb cache” hard drive (32mb is a tiny bit faster, but obviously costs more)
3) make for absolute sure it is a “SATA II” hard drive (aka serial ATA II) and NOT and “IDE, PATA, SATA or SCSI” hard drive.

All OEM/Custom PC will have at least a 7200rpm and 16mb cache drive so no need to worry. Some websites still sell slower 5400rpm drives with 8mb cache, so if building at home, be sure to check this.


WTF IS RAID?!?!

Redundant array of inexpensive disks. come again? attaching two hard drives together to make one uber-drive.

RAID 1: 500GB + 500GB = 500GB. This gives you a real time, PERFECT backup. It saves everything to two disks at once. Therefore if one disk fails, you still have EVERYTHING backed up.

RAID 0: 500GB + 500GB = 1000 very fast GB. 2 hard drives in this config pretty much doubles your read/write. 3 drives triples it. At 4 drives it starts to get expensive and you will reach the limit of your motherboard. You could spend lots of money of a PCIe RAID controller if you wanted 10 drives for example. Increased read/write has the same effect as fast memory. exactly the same effect, so go and re-read the list in the memory section (faster turning on/off computer, faster loading of photoshop etc) If one disk breaks you lose everything.

RAID 5: two drives in RAID 0 for speed. But then it has two drives that are a RAID 1 mirror of the RAID 0 drives. hence you have two drives workign together to be very fast. AND you have two drives working together to back the other two drives up, should something break. RAID 5 is expensive because you need 4 disks. You also need to buy a good motherboard that supports RAID 5. e.g. P45, 790GX

“The 10K Drive” These are fast hard drives. Very expensive (£130/$150 for 160gb) but super fast for loading windows, games, photoshop etc. Consider only if you already have a damn good CPU/RAM/GFX card etc.
One hard drive, with a normal speed (7200rpm) and a normal size cache (16mb) is absolutely fine, whether you play loads of games or whether you use your pc for word and msn.

The "SSD" drive, should be used in the same way as a 10k drive (i.e. have a small fast drive for windows & games and a big drive for media) SSD drives cost about $300 for 120gb.

This is fairly simple. If you buy OEM you get no choice at all, they will often be TOO SMALL TO UPGRADE COMPONENTS LATER be built cheaply, not have great airflow and cooling characteristics.

If you buy a cheaper custom built PC, it will have a cheap crappy case. If you buy a good Custom PC it will likely come with at worst, a good case, at best lots of options.

If you build yourself, just read user reviews, make sure it has good airflow. Good airflow does NOT mean 5 spaces for fans. Look for removable motherboard trays, PSU at the bottom, dust filters on the fans, good brands (Thermaltake, Antec)

If you want to spend < £70 on a case there really is no option than the Antec 300 gaming case. PERFECT air flow, looks stylish, £40, very cheap, steel instead of some cheap plastic. > £70 then look at thermaltakes, the antec Nine, Twelve and P-series. If you are spending more than £70 on a case, instead of getting a better PSU/GFX/MOBO then you probably already know all this.

DO NOT UNDER ESTIMATE THE IMPORTANCE OF THIS. I CAN NOT STRESS THIS ENOUGH.

Lesson number 1) NEVER BUY A PSU IF IT’S LESS THAN £30

Lesson Number 2) If you want a good performence computer NEVER BUY A POWER SUPPLY THAT ISN’T MADE BY:

Corsair
PC Power and Cooling
Silverstone

Lesson number 3) If you are building a budget computer that doesn't get stressed a huge amount NEVER BUY A UNCOMMON BRAND.

Lesson number 4) NEVER BUY THAT 600w ONE FOR £15. WATTAGE IS NOT IMPORTANT relatively speaking, compared to +12v rail rating. I could explain why wattage is not very important, but then i'd have to take up another 10 lines.


“AMP RATING OF THE +12v RAIL IS IMPORTANT” NEVER BUY ONE WITH LESS THAN 30 AMPS ON THE +12 RAIL IF YOUR BUILDING A POWERFUL SYSTEM.


OEM PSUs ARE CRAP AND THEY WILL NOT BE POWERFUL ENOUGH TO HANDLE ANY UPGRADE IN PSU OR ANY OVERCLOCK. THEY DON’T HAVE THE CONNECTORS FOR GRAPHICS CARDS AND NO CONNECTORS FOR EXTRA FANS/LIGHTS/DVD DRIVES.

When buying a PSU, you need to look for:

All the bits of your computer need different size/shape connectors, so you need to buy a PSU with the right ones.


PCI-E 6 pin

These power your graphics cards. If you want a GFX card that will be good enough to play games it WILL NEED at least one “PCI-E 6pin connector.” HD3870/4870 nVidia 9800GTX+ need TWO connectors. If you have two 4870 in crossfire, you would need 4 (2*2=4) Some powerful GFX cards need a “PCI-E6+2” or “PCI-E 8pin,” so look out for this. It is called 6+2 because you can take the 2 off if your GFX card only needs 6 pins.

A note about these connectors: if you only need a VX450, but your GFX card needs say two PCI-E 6pin connectors, you can buy adapter kits. You plug one adapter into TWO 4pin molex (see below) and you get ONE extra PCI-E 6pin. These cost about £4 on average.

SATA

SATA is the name of connection your hard drive and any disk drives need. One hard drive needs one SATA connector. One DVD drive needs one SATA connector. You get the point. 3 hard drives, 2 DVD needs 5.

4pin MOLEX (aka 4pin peripheral)

These are the “old fashioned” version of SATA pins. YOU CAN’T PLUG ONE OF THESE INTO A SATA COMPONENT OR VICA VERSA. However, these 4pin “molex” things are great for any extras (lights, fans, fan-speed-controllers, LED fans) So if you want 3 fans, you need three of these. (some fans plug into motherboard via a small 3pin wire. AVOID THESE MOTHERBOARD-CONNECTING FANS)


THE BEST PSU TO BUY IS:

£45 – Corsair VX450 (about 500w average power. 31 amps on +12v rail. 8x peripheral. 8x SATA. 1x PCI-E)

£55 – Corsair VX550 (about 600w average power. 42 amps on +12v rail. 8x peripheral. 8x SATA. 1x PCI-E)

£65 – Corsair TX650 (700ish watts average power output. 52 amps on +12v rail. 10x Peripheral. 8x SATA. 2x PCI-E)

£70 + Any Corsair TX/HX, Any PC power and cooling 600w +, any Silverstone 600w +



They all have the connectors you will ever need. All have adequate +12v ratings. Corsair plain lie about their power ratings to be on the safe side of any legal action: VX450 is rated at 450w, but produces 570w flat out, and averages about 520w. Any other company would take the other approach “well it produces 430w at max, so lets just round that up to 450w, they will never know it only produces an average of 360w” A recent Tomshardware article took a cheap 500w PSU and found it caught fire when you tried to use more than 262w!!! NOT GOOD.

There is no such thing as a "cheap" GFX card that will "max out" all the settings/options/effects on a brand new game whatever the youtube comments or marketting blurb tells you.

The performance and feature-set (a fast GFX card is useless if it doesn't understand the latest programming-code used in a game, for example) of GFX cards changes very quickly and quite substantially. Buy a card for £100 and it's worth £50 a year later. Likewise buying a mid-range £100 GFX card may net you 50 FPS in "game x," but in just one year that GFX card's direct replacement (costing near enough the same, £100) will typically give anywhere from a 20% increase to DOUBLE the FPS.


How to choose a graphics card?

1) What games do you want to play (serious sam 2, consoles ports (e.g. COD games) or PC-specific titles (far cry 2, crysis etc)
2) How do you want to play them? (satisfied with high and not "highest" settings?)
3) The above two steps dictate how much money you HAVE to spend
4) What is the rest of your PC like? (It would be stupid to spend £200 on a quick GFX card, when it will be held back by a slow CPU that can only achieve 35 FPS. You've spent £200 and your game will still feel "laggy" in places.)


A lot of people come and ask me “will a *insert random number and letters here* be better than a 7800GT” I am not going to list everything, or any combinations. Work it out for yourself. GFX card = graphics card. GPU = graphics processing unit. Some cards have two graphics processors on one graphics card. This can hugely increase performance, but can only be found on the highest end cards. Said dual-GPU cards can be hard to identify, for a while the in-thing was to name call dual-GPU cards "ATI blah blah X2" or "nVidia blah blah GX2." However at the moment (mid 2010) they just prefer to add a very high number to the end of the name (see naming schemes in a second)


NVIDIA Naming System

First Number: Generation of graphics card ("Quadro" = SERIOUS CAD/DCC (not to be used for gaming) G... = 2009/2010, 9 = 2008, 8 = 2007, 7 = 2006, 6 = 2005, FX = 2003)
Second number: low to high. 9 = top of range, 3 = bottom of range
Letters: GX2 > Ultra > GTX > GTS > GT > GSO > GS > SE > LE (GX2 is top of range. LE is bottom)



Some General nVidia RULES:

All 8800 are better than all 8600. Even a lower end (say a GS) 8800 is better than a higher end (say, a GTS) 8600
All nVidia models are extremely-close-to-being-100%-equivalent to the next lowest model number of the next generation (8400GS = 9300GS, 8500GT = 9400GS etc)
Newer models will support NEWER features. Latest Medal of Honour needs something called “DirectX 10” So whilst a 7900GX2 is much faster than an 8400GS, the game simply won’t work because the 7900GX2 only has DirectX 7. The 8400GS is much slower but has DirectX 10, so it will work. For this reason, it is generally best to buy the latest generation of card.



Using these rules, lets look at some examples:

8800GT vs 8800GS (GT is better than GS, so 8800GT wins)

7900GX2 vs 8400GS (8 means 8400 is newer. 4 means 8400 is lowest end. GS means it’s pretty damn low. 7 means 7900 is older than 8400. 9 means it is top of the range. GX2 means it is better than all 7900GT/GS etc 7900GX2 WINS because it was the flagship model one year ago. 8400 may be new, but it is a budget model)

8600GTS vs 9500GS (8 is older. 9 is newer. 6 is mid range. 5 is mid-low range. GTS is higher than GS. However, remember rule number 2: “equivalent to the next lowest model number of the next generation” so a 8600GTS is equivilant to the next lowest number/letter of the next generation. The next lowest number from 6 is 5. The next generation from 8 is 9. So 9500GS = 8600GTS performance wise)

Most common: 8600GT or GTS vs 7900GS (this is an exception, 7900GS is slightly better than both. Just)

GTX480 vs GTX285 (These are the cards that came when nvidia ran out of numbers [7000 series, 8000 series, 9000 series, ugh... errr... I know, let's call it the "G..." series] "GTX" means both, in their day, were very high end cards. The number to the end, 80 and 85, shows they were also quite high up in the "GTX" line. But the 4 shows the GTX480 is much newer than the GTX285 and thus much faster)




ATI Naming System


Similar idea, first number/letter denotes the series/year, the second number shows it's approximate target usage and the final number shows it's position within that target usage.

First letters/numbers:

Radeon 7000 – 9000: 2003/04
X000 era: 2004/05
HD2000: 2006/07
HD3000: 2007/08
HD4000: 2008/09
HD5000: 2010
HD6000: ????


Second number:

HD4800, HD4300, HD5900 etc

3 - 5 is low end, only very basic games or lowest settings on "console ports"
6 - 7 is mainstream gaming. Will play all games on decent settings.
8 - 9 is high end gaming. Will play 99% games on highest/extreme settings.



Third number:

e.g. HD4850 or HD4870. It really does vary between the 300s or the 500s or the 800s but generally in LOW end cards the end number makes a small differe (e.g. HD4570 isn't much quicker than HD4560) but in HIGH end cards it makes a big difference (HD4890 isn't far off twice as fast as the HD4830)

3 - lower end
5 - mid
7 - high
9 - extreme


As for a 2900 vs a 3450. It’s pretty much same story as intel, but they are more spaced out. Looking at most modern cards, a 4870 is as good as TWO 3870. a 4850 is waaaaaay better than one 3870.





There are lots more things to know about Graphics cards, but I won’t go into any of it here. HD5970 for example, a dual-GPU 2010 card, has 3600 cores for example. That wasn't a spelling mistake. And those PS3 users think they're 6 core CPU is special...

The high end ones tend to use very futuristic memory, e.g. high end GFX cards in 2008 had DDR5 memory, but high end CPUs in 2008 only took DDR2 system memory! so the GFX cards are a long way ahead on that one...

Tech people like codenames and try to confuse us with those and big numbers, the PC world adverts for example are hilarious: "The new toshiba blah blah blah WITH A MASSIVE 3 GIG OF MEMORY!!!!" The people that will buy that laptop most-definitely have no idea if 3 is good or not, nor do they know how much a "gig" is. nVidia call their GPUs stuff like "G84/G86/G92/G92b/GT200" whilst ATI have their RV... numbers, e.g. RV670s is kinda a code name for the numbers at the front, and LE/PRO/XT as a codename for the last numbers. Thankfully there is not much reason to worry about any of this.




Anywho. The graphics market changes so quickly, I can’t give a recommendation. The single BEST thing you can do however, is look at reviews like tomshardware.com or fudzilla.com and find a review in which they BENCHMARKED (this means actually playing cod5, and recording how many FPS they got, then comparing that to lots of other cards) the GFX card in question. They will always have a summary paragraph that tells you whether it is worth the money.

Tomshardware.com do an excellent (only if your in US of A) feature every 2 or 3 months, called “best gaming graphics cards for *month*” they split the article up into < $50, 50 – 100, 100 – 200, 200 – 300, 300+ and reveal the best card in each price bracket.

Motherboards are a bit like graphics cards. Intel/AMD/Nvidia design “chipsets” and then sell the design to companies like asus/gigabyte/Foxconn/DFI/XFX etc who then make the motherboards and then sell them. Below are some things to consider when buying a motherboard, before we get to chipsets.


SLI/Crossfire

If you want to have two nVidia GFX cards you HAVE to buy a “SLI” mobo. If you want two ATI GFX cards, you HAVE to buy a “crossfire” mobo. You can put one ATI card in a SLI motherboard. You can put one nVidia GFX card in a Crossfire mobo.
SLI currently goes up to three GPUs (tri SLI)
Crossfire (CF) goes up to four GPUs

Having two graphics cards does not mean twice the performance. It will be somewhere from 1.5 to 1.8 times the performance of a single card, depending on how old the GFX card is (the older the card, the longer people have had to optimise the drivers to make use of two cards)


PCI-E Lane Width

If your mobo only has one PCI-E slot (maybe I should have explained this. PCI-E is the name of the slot your GFX card goes in on the mobo) then not to worry, as it will most probably be a full speed “x16” link. If you have SLI or CF look for a motherboard that will use Slot (aka lane) one @ “x8” and slot two @ “x8” minimum (do not accept “x4”)


Memory

First off, if you want DDR2 memory, you need a DDR2 motherboard, if you want DDR3 memory you have to have a DDR3 motherboard. They are NOT backwards compatible.

Secondly, is even if the motherboard is DDR2, it may not support all the different speeds of DDR2. Most support 533mhz, 667mhz and 800mhz. The more performance orientated chipsets support 1066mhz too, the best ones support “OC 1200mhz.” This means it officially supports memory that has been OC’d up to 1200mhz. If you want 1066mhz memory, your mob needs to support 1066mhz memory.


FSB (only for intel CPUs older than about 2008/09)

Again, something I really don’t need or want to go into detail about. All you need to know is this, if you buy an Intel processor with a 1600mhz FSB, it would be best to get a motherboard that supports 1600mhz (if the mobo only supported 1333mhz, it would still work, just not perfectly) This whole match-FSB-to-mobo only applies to Intel CPUs.

AMD call FSB the "hyper transport" (HT) and luckily it's less confusing because an AM3 CPU will have HT3.0 and likewise any AM3 motherboard will also have HT3.0



RAID

Cheap MOBOs only support RAID 1.
Better ones support RAID 0 and RAID 1
Performence ones support RAID 1, 0, 5 and possibly RAID 10.


The Chipset/Names

Motherboards as said, have various designs with names. nVidia name theirs a bit like GFX cards (730i, 730 ultra, 750i, 780i SLI, 790i, 650i, 680ultra SLI are a few examples) The only AMD ones you might want AMD call 790FX/GX (but you if you are silly/cheap enough to get an AMD AthlonX2, then don’t waste money on such an expensive motherboard) Intel also make ones, which we will look at In a mo.

AMD make motherboards capable of Crossfire, and for AMD processors (socket AM2. oh god… I haven’t explained sockets… ugh)
Nvidia make motherboards for SLI and AMD processors.
Intel make motherboards for Intel, CF and occasionally SLI!

When the nVidia 700 series was current (laste 2008/2009) if you wanted SLI, then you have no other option than nVidia chipsets. In about 09 nVidia sold Intel the technology it needed to enable SLI on SOME intel motherboards.



If you do go AMD, it doesn’t matter unless you want to overclock. In which case get the highest "numbered" motherboard you can afford. BUT BEWARE! It's better to buy a 785 from a great brand with lots of features than a 790 from someone who we've never heard of.



Intel chipsets have been "the bomb" for intel CPUs from about 2008 (current date is mid 2010) The Intel naming system:

First letter: series (also can denote socket, thus which type of CPU is compatible)
First number: year (also denotes socket, thus type of CPU compatible)
Second number: range of features (e.g. RAID, PCIe width, CF/SLI) 8 is loads, 1 isn't.

"G" - great for basic computers
"P" - stands for performance
"X" - stands for extreme

"30/40" is roughly 2007/09 and relates to all "Socket LGA 775" products (all core duo and core2 products, most pentium E.... products, some Pentium 4 etc)
"50" is roughly 2010 and relates to "Core I" products.
"8" > 5 > 3 > 1

E.g. an "X58" is for extreme users (X) is from 2010 (50) and has loads of features (8)

A P35 is for performance users (P) is from 2007 (30) and has a lot of features. It probably has similar/less features than the P43, even though the P43 is newer.

what is it? Each object (i.e. CPU, GFX card, memory) on your motherboard has an address (like your house's address) A 32bit operating system has addresses with 32 characters in it. A 64bit system has 64 character addresses. Your operating system gives each individual byte of memory an address. A 32bit system (most likely what your using now) has a LIMIT OF 4GB OF RAM, because:

2^32 =~ 4,000,000,000

So a 32bit OS can only invent 4 billion addresses. Since each byte of RAM needs one address, you are limited to 4 billion bytes. 4 billion bytes = 4GB. You can put more in, but the OS will not find it, and won't be able to use it.

Note: you will often hear something about a "3.5gb or 3.3gb" limit as opposed to a "4gb limit." This is because the CPU, motherboard, BIOS etc etc need a little bit of RAM to function properly. These components get the RAM they need first (roughly 300 - 700mb) and then the rest of the RAM is given to the OS, hence you usually get just under 4gb.

A 64bit system DOES IN FACT HAVE A LIMIT to the amount of RAM that can be used. However as you can see below, the number is so big (that is, errm, 0, killo, mega, giga, tera, peta, err almost 18 EXABYTES. EXABYTES. 18 exabytes = THE WHOLE OF THE INTERNET x 9000) that clearly the limit is irrelavent...

2^64 =~ 18,000,000,000,000,000,000



What to look out for:

2000/ME/Server/XP/Vista etc all come in various versions such as Home/Business/Pro/Basic etc etc.

However, each of these versions can be bought in either "32bit" or a "64bit" version. All proccessors since the Pentium 4 and Athlon 64 are compatible with 32 and 64bit, so no need to worry about compatability of hardware.

However, 99.99% of software and drivers are made for 32bit machines, and a small minority of 32bit programs DON'T work on 64bit. I must stress it really is a small amount, but it can make a big difference. If your are CONSIDERING 64bit, first you MUST CHECK ALL OF YOUR EXISTING SOFTWARE AND DRIVERS ARE AVAILABLE IN 64bit VERSIONS.


Which one is right for me?

As we've seen, a 32bit OS has a limit of 3.3gb USEABLE memory. So if you are into games like crysis and far cry 2, and are running Vista Business or Ultimate then 4GB is actually about perfect. For those two games and Vista basic or home, 3gb would be fine. If your into heavy video editting then 4gb is hugely recommended. If your into CAD or DCC (in which case you won't need to be reading this... but still) 8gb is a more sensible number.

SO, if you only ever need about 4gb max then 32bit is fine. If you have 4gb, but may need to upgrade or your PC has to last a while in the future (i.e. for Windows 7, or next gen DX11 games like crysis 3 maybe?) then go 64bit.

Although 64bit has a massively irrelavent limit. There are software limits applied, based on how much you pay. I.e. Vista Basic 64bit has a limit of 8gb, Vista Home has a limit of 16gb, Ultimate has a 128gb limit. Windows Server has a very silly limit (i think it's terabytes)

I think that is all to do with 64/32bit OS...

OVERCLOCKING IS EASY.
ONLY COMPLETE IDIOTS CAN ACTUALLY BURN UP A MOTHERBOARD/RAM/CPU.
IF YOU ARE AN ADULT, THERE IS NO REASON WHY YOU CAN'T OR SHOULDN'T OVERCLOCK.
ONLY DO IT IF YOU HAVE READ 58 "HOW TO" ARTICLES ON GOOD SITES.


IT CAN BE FUN.
IT IS REWARDING.
IT IS FREE.
IT SAVES YOU HUNDREDS (I7 2.66GHz is $800 less than a I7 3.2GHz but can easily be overclocked to 3.2GHz)


BEFORE YOU OVERCLOCK YOU NEED:

A SOLID UNDERSTANDING OF HOW OVERCLOCKING ACTUALLY WORKS
A FEW DETAILS ABOUT YOUR HARDWARE (see later on)
TO KNOW RULE NO. 1 OF OVERCLOCKING:


YOU HAVE SEEN THAT ON A WEBSITE *RANDOMNAME* GOT HIS E5200 TO 4GHz. UNLESS YOU ARE EXPERIENCED, YOU WILL NEVER EVER GET ANYWHERE NEAR THIS.

Nothing worse than some complete noob saying "oh i only want to overclock to about 4ghz. I will take it easy for my first go" when only the luckiest AND most experience people get 4.2GHz.


Details about hardware you HAVE to know before you OC

Your CPU's VID
The average OC for your CPU on high end cooling by an experienced OC'er
Your MOBO's VDROOP
All the basics about your CPU: FSB, MULTI, MAX VCORE
If you decide to go for the complex OC, you should also know: RAM timings, Max FSB of CPU, max FSB of MOBO, cache size, fabrication proccess.

That may sound complicated, but a handy piece of software called CPU-Z can tell you 90% of those things. The other 10% can be easily worked out and/or found on all good overclocking websites


THIS IS NOT A DETAILED "HOW TO." This is more of a "How you should start, or go about overclocking if you wish"


The Simple OC

What skills/knowledge do you need: you should be acomplished in using the times and possibly divide function of a calculator. You also need to be good at pushing the + and - keys on your keyboard. (yes, that is how hard basic OC'ing is)
What hardware do you need: anything except a dell/HP. no need for any fancy coolers, particular CPUs or particular RAM.
What software do you need: CPU-Z. Some form of temp reading software (e.g. CoreTemp) and some form of stress testing (recommend OCCT because it is the fastest)

Method: turn up the "FSB" (this is one simple setting in the BIOS) run a stress test program. up FSB. run a stress test program. Up FSB, run a stress test program. Oh dear it's crashed, turn it back to previous setting and you are done.


The intermediate OC

What skills/knowledge do you need: said skills for "simple OC." Just need to know a few more simple terms and what they mean/do. Oh you may need the arrow keys as well as the + and - keys depending on the motherboard...
What hardware do you need: Same as before, but likely hood is you probably have a higher performence PC to start with (i.e. this means your MOBO will have more options, case will have better cooling etc)
What software do you need: same as before. Maybe speedfan.

Method: same as simple OC, but this time you get to play with VCORE (give the CPU more voltage so it doesn't crash at higher speeds) and you get to change one other RAM setting to make the CPU OC a bit further.


Average enthusiast OC method what some of you regular C4F computer section people should be doing

What skills/knowledge do you need: Everything before AND patients. If you can leave the PC on overnight then that may be helpful, as final stress tests should be about 5 or 6 hours. You need to know what a variety of the more advanced terms/abbreviations mean and do. You also need to have quite a good knowledge of what other people with the same proccessor OC'd to etc
What hardware do you need: need a decent MOBO, after market cooling.
What software do you need: speedfan

Method: Use 3 main sub methods: "Max FSB, max CPU, max RAM." Suprise suprise these each find the highest value each component can take before it crashes. however you can't have all components at max at the same time, so at the end of it all you need to choose an optimum setup.


Advanced OC

What skills/knowledge do you need: even more patients. You definately need to know a fair bit about your own hard ware.
What hardware do you need: at this level you should have top of the line air coolers, MOBO born to be overclocked, good case cooling, G.Skill Geil etc RAM.
What software do you need: same as before

Method: Use 3 main sub methods again, but you make use of all the advanced options in the BIOS to achieve higher clocks (e.g. RAM ratios, ACC, RAM latencies, NB & FSB voltages etc) Again it's just about tweaking and tweaking till you find the perfect setup.