overclocking is making it run faster than it normally does.
say you get your Athlon 64 3500+ which runs at 2.2GHZ - don't let the low frequency against Intel's CPU's fool you, its performance is roughly equivalent to a 3.4-3.6GHZ Intel. often performing better in gaming
anyway, so you have your 2.4GHZ Athlon (3500+)
if I were to overclock it, I would make it run higher. let's say 2.6GHZ.
doing this will make it faster, however can make it produce more heat
AMD has made a new core called the Venice which so far outranks all other Athlon 64's in overclocking ability. it runs at low voltage and produces little heat, so on its standard cooling most people can get it from 2GHZ (3000+) to 2.8GHZ.
the 4000+ and FX-55 (high end) actually run at 2.6GHZ, so for the cost of a low-end CPU, you now have one that runs faster than those much more expensive ones
a CPU communicates to RAM at a certain speed, normally 200MHZ
DDR400 actually runs at 200MHZ, but transfers twice the data per clock cycle of the older SDRAM, so people say it runs at 400MHZ
your CPU has a multiplier, which is a number, that multiplies the CPU-to-RAM communication speed (normally called the Front Side Bus, but Athlon 64's technically don't have one, I'll explain after) to get the core speed
basically, to overclock your CPU, you increase your Front Side Bus (FSB, or HTT in Athlon 64's)
just say you had a 3000+, it runs at 2GHZ (2000MHZ) with a multiplier of 10
you increase the FSB (or HTT) to 250MHZ and your CPU speed is 250x10=2500MHZ
because you are increasing the CPU-to-RAM speed, you also overclock your RAM when you do this. this means your CPU can communicate to your RAM faster.
so what people often do, is they find the core limit of their CPU. let's say it can't get past its stock speed of 2000MHZ, they can increase the FSB/HTT to 250MHZ and decrease their multiplier to 8, so that 250x8=2000MHZ
the core speed is not increased, however their CPU communicates to RAM faster, meaning your system can be faster overall when parts need data from RAM (even if not the CPU itself, which is very often)
typically, when a CPU wants to communicate to RAM, it actually communicates through a memory controller chip (MCC) which is normally on the motherboard. the CPU itself does not normally 'know' how to talk to RAM. it is the link between the CPU and MCC which is called the FSB.
because the CPU has to talk to the MCC, it creates a bottleneck.
when AMD designed their K8 CPU's (Athlon 64's, Opterons, Semprons) they put the memory controller on the CPU itself. the MCC being just next to the CPU, not through the motherboard, means it can communicate much faster to RAM
because of this, there is no real need for a 'bus' to communicate with the MCC, it communicates almost directly to memory. AMD called the link between K8's and RAM "HyperTransporT" or HTT (don't confuse it with HT, which is Intel's HyperThreading)
------------------------------------
PSU's are important. they transform the 110/240V (differs between countries) AC power from the wall into the lower voltage DC power for your computer. if it doesn't supply the right amount of power, it can cause all sorts of problems.
a lot of cheap PSU's are only tested in certain conditions, unrealistic to normal day-to-day use, and usually not for a great length of time.
they often rate the PSU according to its maximum output, rather than what it can continuously output (which is usually considerably lower)
also, they often do not supply stable current. if a power supply causes fluctuations, it can damage your parts
so you can see, a good PSU is important.
Antec, Enermax and Thermaltake are trusted brands, there are a few others, but there are really only a few ways you can tell a good PSU from a bad one:
brand
price
weight (a heavier PSU usually being a higher quality one)
------------------------------------
PCI-E vs AGP
the main expansion slot before AGP was PCI. although still used today, it is far too slow for today's video cards.
AGP was made as a dedicated graphics card port, to give video cards the extra bandwith they needed.
however, we are finding AGP still has its limits. although current video cards still do not take up the full bandwith of AGP 8x, the increase in video card technology means thaqt in a few generations of video cards, it won't be one day.
PCI-E is the serial version of PCI, which is much faster than even AGP is.
PCI-E 16x is what's used for video cards. however, often people think that the 16x speed is against AGP's speed. it's not.
people think "okay, here's AGP 8x, and PCI-E 16x. 16x is twice that of 8x"
PCI-E is measured differently. AGP 8x means 8x the bandwith of the original AGP. PCI-E 16x means 16x the bandwith of PCI-E 1x
PCI-E is measured in how many serial connections there are to and from the card. 1x being one connection to and from the card, 16x meaning 16 to and from...
currently, there's nothing wrong with AGP, cards still do not utilize the full bandwith, and there is virtually no difference between an AGP and PCI-E card with the same GPU.
however PCI-E is more future-proof, and offers SLI to those with money.