Everything About [Computers :: Overclocking]

Underlining Principles

The most basic of basics that you need to understand in "overclocking" is: ALL CPU's, RAM, Video Cards are exactly identical... that is within the same generation or "architecture". Manufacturers can add things across revisions such as SSE4 & the higher ratios for faster memory, they usually also "shrink the die" which just plainly makes everything smaller using a more precise manufacturing process... but the basic architecture is identical across all revisions. Die Shrinks allow for things such as less power consumption & subsequent heat because everything is smaller, doesn't travel as far, or take as much to get there.

Yields

Ok, well... if the Athlon64 3000 and the Athlon64 FX/57 are identical processors... You should be able to overclock a lowly 3000 up to FX speeds, right? Nope, this is where manufactures come into "yields", specifically their quality. Imperfections in the materials or even the manufacturing process goes through maturing so to speak... perfecting the manner in which the product is made. Processors are then rated based on how good their individual yield is... so a CPU that tends to start overheating at a given speed using stock conditions will be rated at a low speed then say one that yields so well that it will be given a much higher yield. Imperfections in the ondie cache may also cause the manufacturer to disable that 1/2 the cache and subsequently give it a rating better suited to that lesser amount of cache.

Manufacturers have also been known to mark higher yields with lower ratings due to overall better yields than expected. The reason being that they can sell more CPU's in the $200 range than then $1000 range... this is of course an overclocker's dream come true. The manufacturer usually simply steps up on releasing faster products which in turn gives them a competitive advantage. So there aren't nearly as many of these lower marked products than we all wish, but it does happen more than you'd think.

Underclocking

All of the above is true for RAM as well... there is (relatively) no difference between memory rated at DDR400 & DDR500 outside of their quality in the manufacturing process. Memory, like CPU's, are rated to operate at a given speed w/ the manufacturer's full warranty. Underclocking would be akin to taking DDR500 & underclocking it 2 DD400 speeds because your motherboard isn't capable of those higher speeds or to keep heat down in a small chassis. Some people also do it to extend the expected life of their components as there is a MTBF (Mean Time Btween Failure)... doesn't last forever, and eventually wears out. This is also why servers run such things as error correction at "registered" speeds... usually much slower than the consumer counterpart, but all about reliability. Especially when a server runs 24/7/365 while your PC does not.

Requirements

So now we come to actual overclocking... overclocking a compenent higher than it's given "rating" will often lead to higher power consumption and greater heat. This is because it takes that component more juice to achieve a similar speed as a higher rated component remember.

So the 2x main things you worry about is:
1.) Supplying sufficient power (usually raising the voltage)
2.) Better-than-Stock cooling (whether it B liquid/phase/air)

Traditionally w/ CPU's, their speed is determined by 2x figures...
1.) Front Side Bus (typically the speed @ which the memory operates)
2.) Multiplier (An internal figure 2 multiply the FSB 2 get actual speed... more indepth)

Conclusion

I know it's a little confusing, but you'll get it... overclocking is simply the luck of the draw, some component are simply better at overclocking than others (w/in any given rating).

Links

Overclocking the AMD Athlon64 x2 4200+:
http://www.anandtech.com/cpuchipsets/showdoc.aspx?i=2452&p=3

Overclocking the Intel Core 2 Duo e6300:
http://legionhardware.com/document.php?id=606