Overclocking Basics:

Posted: June 3, 2006 in How To

The theory behind it

Overclocking is basically making your hardware run faster than it was specified to do. All CPUs of a specific type are created pretty much equal. All that differentiates them is their ability to run at higher or lower speeds without getting too hot or requiring too much voltage. This is due to things like wafer purity etc when the CPUs are made.

However, companies like AMD are required to make available a complete range of CPUs (ie 1700+ right through to the 2800+). This is where they have problems. If they've created a batch of CPUs that are capable of running up to say 2400+ speeds safely but they have a high demand for 1800+ CPUs, they may 're-badge' some faster CPUs as slower ones to supply the demand. (this is an old example btw)

These CPUs are still capable of running at the faster speed however, which is why OCing can be so effective.

Also, the limits the manufacturers allow have to be quite generous. They can't have a CPU running at it's specified speed reaching really high temps, so they have to leave a little leeway as a safety margin.

This safety margin is based on using the standard retail heatsinks supplied with the CPUs, which in the case of AMD colers, is pretty poor. Merely by placing a better heatsink with better heat transfer paste on the CPU, the temperature will be greatly reduced thus increasing this 'safety margin'.

Overclocking your CPU basically takes advantage of the fact that the safety margin is there. The bigger you make this safety margin (ie, by having effective case cooling, big heatsinks, high CFM fans, watercooling, phase-change cooling etc) the further you can overclock your CPU.

The reason behind all this elaborate cooling methods is that the main thing that stops you from overclocking is heat.

The faster your CPU runs, the hotter it gets. Keeping the temperatures down increases the distance you can overclock.

There are a lot of other factors to consider with this though.

The techie stuff

The speed of your processor is determined by the Front Side Bus (known as the FSB) multiplied by the (aptly named) multiplier.

ie
FSB x Multi
in the case of a barton 2500+
166MHz FSB x 11 = 1826MHz (the actual speed a 2500+ barton CPU runs at)

When overclocking, you can do two things, increase the multiplier (no longer the standard method) or increase the FSB. There are some obstacles.:
1) Not all motherboards allow changes to FSB/Multi outwith the CPU specs.
2) Not all CPUs allow their multi to be changed (CPUs like this are referred to as 'locked').
3) Some CPUs are more receptive to increases in operating speeds than others.
4) Different motherboard have different 'OCing friendly' features.

There are other things to take into account too.

Voltages play a large part of OCing. The higher the voltage a CPU is recieving, the hotter it will get but the higher the voltage the more stable a CPU will be at a higher speed.

This doesn't mean you can just go a double the voltage to the CPU and then double the multiplier however!!  That is, if your motherboard even lets you set the voltage that high.
So this makes it clear that certain motherboards are more suited to OCing than others. A good board for OCing will have certain features that OCers will look for like:-

Voltage Control – Allowing them to increase the voltages to make higher OCs stable
FSB Control – Giving the user the ability to select higher FSBs in easy 1 MHZ increments is desirable
Multiplier Control – Alowing the user to change the multiplier
Temperature Monitoring – Allows the user to ensure that their CPU isn't getting too hot etc
PCI/AGP Lock – stops the PCI/AGP speeds changing when you change the FSB

Cooling

Like you said is a major issue:

CPU's (processors) have been getting hotter and hotter as they have become more and more powerful. This is mainly due to the increased amount of power it requires to run, which current flows through and experiences resistance -> produces heat.
The amount of heat produced by a particular cpu will also depend on the frequency that it is running at, and the voltage applied across it. The higher the voltage, the higher the frequency, the hotter it will get.

In light of this, to achieve reasonable overclocks, you need decent cooling. This counters the increased frequency of the core, and the increased volatage across it, necessary to maintain stability at that higher speed.

Cooling Methods:

At the moment there are three conventional methods to cool down your vital components:
Air cooling

This is the traditional, and by far the easiest method to cool your system. OEM manufacturers prefer it as it is cheap, and as they don't overclock, it doesn't matter how hot the CPU is (as long as its not melting yet )

However, this doesn't mean that this isn't an option. For the enthusiast, air cooling can still cool very efficiently.

For starters, you need to have good case cooling. 2-4 case fans are always good… after all warm air rises.

Heatsinks made by Thermalright, Zalman, and Swiftech are currently considered the best on the market for cooling your processor. Be warned, air cooling can soon raise noise levels significantly! [especially when including case fans!] Average Retail Cost – up to $50
Elaboration into air cooling your CPU:

There are many heat sinks out there, all at different prices with different designs. How do you know which one is the one for you? Here are some of the base points of heat theory and construction in order to help you choose.
Sections:

Thermal Conductivity

Thermal Contact resistance

Brief over view of fans and which one to use

Thermal Conductivity

You may have noticed that CPU heat sinks are made from different materials. Retail packages usually have the standard aluminium heat sink. So why is this a bad thing?

Check below:

Material Thermal conductivity (W•m-1•K-1 )
Air……………..0.026
Aluminium……..236
Gold……………320
Copper………..390
Silver………….430
Diamond………2500 (dependent on type)

In case you want to know:

W•m-1•K-1 – watts per meter-kelvin
g/m ³ – gram per cubic centimetre

The higher the number the better the material will transfer the heat away from our processor die. As you can see from the table air would not conduct heat away from the Processor die, which is why we use a heat sink. Most retail CPU coolers are made from aluminium. Aluminium is a cheap, readily available material. It is also very easy to work with and has a relatively good thermal conductance. This is why manufacturers use it; it is basically a good all rounder.

But being a good all rounder means it never excels at any one thing, which in our case is removing heat away from the processor as quickly and efficiently as possible. So the next step of heat sink manufacturers was to make a ‘plate’ of copper that would be in contact with the processor IHS or core. This copper plate is bonded to an aluminium heat sink. This does remove heat away from the processor at a greater speed than an all aluminium heat sink, but the main part of it is still aluminium.
Above this type of cooler is the all copper cooler. As you can see in the table, copper is quite good at conducting heat. Which is why it is the recommended cooler for over clocking situations? There have been a few designs of this, the more elaborate fin patterns of the Zalman ‘flower’ coolers which tout increased surface area, to heat pipes that will ‘transport’ heat all over the cooler for better cooling.

Water cooling

Has been growing in popularity in recent years, especially as components have been generating more and more heat. The main advantage of this is that it is significantly quieter than air cooling, thanks to the allowed use of perhaps just one low-speed, quiet 120mm fan for the whole system. Water is pumped to the CPU, gfx card etc, and then fed to a radiator where it cools.

Water cooling is still potentially risky (leaks are not a good thing!), but on the whole, time and money can be well invested in a water-cooled setup, which will produce a cool, quiet system. There is non-conductive "water" that you could buy so that IF it did leak you wouldnt have a problem at all, but it does cost more than well…water.

Average retail cost – starting from around $120 and prices can go very high

Phase change cooling

This works on the same principle as air conditioning systems, fridges etc. Heat is "pumped" from one area to another, so one area gets cold, the other hot. Of course, the cold part is attatched to the CPU, producing ridiculous temperatures for the best overclocks . Temperatures of -20°C through -150c  are common. The best "extreme" overclocks are achieved using phase change "nVENTIV" (now under a new owner and Vapochill being the main contenders), but there are some "home brew" creators out there like PC ICE ( a good friend of mine as well) who builds very nice units for a reasonable price. Single, Dual, and triple cascades, this man does it all. Triple cascades taking the cake when it comes to conventional err hardcore overclocking, since I would dictate that LN2 would be the most extreme.
Average retail cost can vary greatly : these days you can buy a used unit for 300bucks, and hopefully it will be regassed. New units can cost anywhere from 400-2k+ depending on how extreme you get with it.

Credit due to Airchie of DeviantForums 🙂 

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s