The Overclocking Guide
Ahh... the grandmother of all overclocking articles has come into being. Aren't you excited? And as Trent Reznor (NIN) might say, "Doesn't it make you feel better?" Truth is, we at Tweak3D have only briefly gone over what goes into overclocking a computer while making sure it is as stable as granite, and I intend to change that - now!
Here's a basic outline of the kind of stuff you will find in this guide.
- Necessary tools
- Safety precautions
- Bus Clock Speed
- Bus Clock Multipliers and Multiplier Locks
- Chip Voltage and Stability
- The Overclocking Process
- Stability Testing Procedures
- Troubleshooting a Failed Overclock
- Electrostatic Migration and Burnout
- Overclocked Processor Lifetime
- Effect of Non-standard Bus Speeds on Other Computer Components
- Alternative Methods of Overclocking
There are several things you need to know before you begin trying to overclock a computer. Depending on the part of the computer you are trying to overclock and how far you are going to take the process, you may need any or all of the following things:
- Phillips head screwdriver
- Flat head screwdriver
- Thermal Paste
- Thermal Tape (FragTape)
- A flat razor
- Cooling fan/Peltier/etc.
- Application Specific Tools (ex: Peltier cooling systems may require insulation)
Make sure you have ALL the tools you need before you begin working (in some cases, you may not need any tools at all). Make sure you are in a clean, well-ventilated area, with plenty of workroom (if you will be taking on a larger project).
There are a couple of very important things to keep in mind when you are attempting to overclock a computer, so that you don't damage your equipment. The first of these things is to make sure you have adequate cooling to take on the project you are planning. As will be discussed later, cooling can make or break an overclock - but that isn't its only benefit. It also helps prevent damage being done to the chips due to excessive heat.
Ok, now that I have taken care of explaining the importance of cooling to you, on to the (second) most important safety precaution - which has to do with progressive overclocking. I know, I know, that isn't a term most people have ever heard of - and that's because I just coined the term. Progressive overclocking has to do with the process of slowly clocking your system faster and faster until it reaches its peak stable speed. This is frequently done with video card overclocks, because it is very easy to over do it and fry the card. The process with video cards is very easy - you simply overclock in 5 MHz increments until you reach an unstable speed, and then downclock the card in 1 MHz increments until you reach a stable speed. Then of course comes the obligatory testing to determine whether or not the card is stable even during system strain - and if it passes, you're gold.
However, the process with a CPU is more difficult - mainly because it is hastlesome to go back into the BIOS for every clock change. With bus clocks, bus multipliers, and chip voltages to contend with, things aren't always hunky-dory.
Bus Clock Speeds
The system bus clock is a very important concept when dealing with system overclocking, particularly when you are dealing with an Intel-based system. This is because Intel's processors are multiplier locked. More on that subject later, however. Right now I want to explain to you about the system bus and how it can effect your system.
The bus clock I am referring to is the system bus on which the processor communicates with the rest of the computer. It is derived directly from the computer's internal quartz crystal which runs at ~12 MHz (and subsequently also runs the computer's internal clock). This bus speed, when taken into account with the processor's bus multiplier, determines what speed the CPU runs at, as well as some other things. You see, the PCI and AGP slots derive their bus speeds from the system clock (33 and 66 MHz respectively) using a bus divider. These dividers have been set up specifically for the standard system bus speeds (66/100/133), but don't work quite as well for the non-standard bus speeds. That means that, unless you are jumping from 66 to 100 MHz, or from 100 to 133 MHz, you will also have to over or under clock your system buses - and sometimes they don't take kindly to the extra stress.
System Bus Multiplier and Multiplier Locks
The system bus multiplier takes the system bus * whatever the multiplier is to determine the speed at which the processor is running. That means that a computer running at the 100 MHz bus speed with a 4.5 bus multiplier would be running at 450 MHz. Simple enough to overclock your computer without messing with the system bus, right? Wrong. Why is that? It is because Intel had the audacity to lock the clock multiplier on its processors. That means that your computer HAS to use the 4.5 bus multiplier to derive the processor's clock speed, and dramatically limits the speed range of most processors. This, combined with the fact that most computer components don't function properly on non-standard bus speeds, makes overclocking most computers difficult (to get a completely stable system you have to jump up to the next standard bus speed - a mighty task for most processors).
Of course, AMD has (sort of) come to the rescue by not locking the system multiplier. However, to change the setting, you have to break the chip's casing open, hence voiding the warranty (overclocking voids your warranty anyway - so no big deal). They were even so nice as to include an edge connector to allow the connection of third-party jumpers to make overclocking a snap. Of course, you get the best results using a soldering iron... but that's an entirely different article.
Chip Voltage and Stability
Chip voltage can turn a not-quite-so-stable chip into rock hard granite. Most CPU's have some sort of way to change the voltage of the chip. Raising (and in some rare cases, lowering) the chip's voltage can create a much stabler chip, at the cost of more heat. Heat, of course, alternately lowers the overclockability of a chip, but it doesn't lower the chip's overclockability as much as upping the voltage raises it. And besides, there is always cooling. But more on that later.
The basic theory on chip voltage and how it affects the processor is this: a higher chip voltage increases the signal strength between transistors within the chip, allowing the signal to ignore greater discrepancies within the silicon core itself. You see, the silicon wafers used to make the chips aren't always pure, and they definitely aren't all of the same quality. A chip with a higher clock rate is generally going to have a core made of a higher quality silicon wafer (something that can't be determined until after fabrication, due to the fact that all the wafers are as pure as they can make them).
Now, the processor signal has two choices as to how to deal with a chip impurity (how it is dealt with has to do with quantum physics and really isn't imperative to this discussion). It can either jump the gap, or go around it. When the processor frequency is lower, the signal has the time to go around the defect if need be, but if the frequency is too high and the signal must go around, the signal doesn't get to its destination in time or at all (remember we are dealing with millionths or billionths of seconds), causing a miscalculation that usually will cause some form of software error (commonly it causes a crash).
However, upping the core voltage is like giving the signal a running start, it allows the signal to jump gaps within the chip with relative ease (sort of like a lightning arc), and the signal gets to it's destination in time.
Two of the parts of the overclocking process up the heat produced by the processor - upping the frequency (the actual overclock) and upping the core voltage. Excessive heat within the core creates more of those gaps that I was discussing above for the signal to cross, and too many of these gaps will weaken the signal to the point where it becomes non-existent and creates some more of those wonderful software errors. Here's the lowdown for you physically inclined folks - the extra heat energizes the particles within the silicon wafer. The pathways within the silicon wafer are approaching the size of light rays (read very small), so if the particles move too much, they break their connection with the other particles within the pathway. These temporary breaks do the same thing as the impurities mentioned above. Got it? Good.
Ok, now that you know all about why cooling is so important, here's the skinny on what kind of stuff is available to you hobbyist overclockers out there, and then maybe I'll do a little of the honorable mention thing to the more expensive cooling systems of the world. The simplest way to cool your chip is called passive air-cooling. Passive air-cooling is basically the use of the surrounding, cooler air to cool the chip, using some sort of ball bearing fan. This is the cheapest, easiest, and most common way to cool your processor - all it entails is attaching a fan/heatsink combo to the processor to cool the thing down.
Hard-core hobbyists, however, are never satisfied with simple 'air' cooling, oh no. Heck, I've even seen some guys go so far as immerse their systems into super-cooled glycerin (a non-conductive liquid) to cool their processors. But that, once again, is a subject for another article. There are two 'reasonable' types of active chip cooling. One, a Peltier system, basically uses a heat-transfer plate (called a Peltier) to conduct heat away from the processor, where it is then carried off by a standard fan/heatsink combo. The only extra stuff you need for this type of system is some form of insulation for the exposed portion of the cold side of the Peltier, because otherwise you will get condensation, and even frost (Peltiers are extremely efficient).
The other 'standard' form of active cooling is using some form of water cooling device. These devices are extremely complex, and on top of the mandatory insulation, you also need a pump and some form of condenser... for the average hobbyist, it would be easier to put your computer in the freezer and run the wires out through a self-drilled hole rather than set one of these bad boys up.
Of course, you always have the "professionally" overclocked systems from companies such as Kryotech. Kryotech uses a method of cooling called "liquid phase change cooling." It is extremely efficient but also extremely expensive - the special case alone costs $1000 US all by itself, not including the processor enclosure. Boy, what some people will do for a couple of extra megahertz. Anyhow, if you've got the cash, their systems are something you might want to look into.
To install a cooling device, first you need to remove the old fan/heatsink combo from your processor. This should be a fairly simple operation. Don't be afraid to use a little force to break the seal that was created by the thermal compound. You will then need to use a flat razor to remove the remainder of the thermal compound from the top of the processor. Once this is complete, apply either some more thermal compound or thermal tape (FragTape) to the top of the processor and attach the new heatsink on top of that. Simple enough, huh? Some setups may have other necessary steps to attach the cooling device (thermally insulating silicon caulking compound, etc.) to prevent condensation - but that won't be a problem with a standard fan/heatsink combo.
Well, there you have it - the first part of the overclocking how to guide. In the next part, I will be covering the actual overclocking process, and some other nice little tidbits. It should be posted tomorrow or the day after. And of course, feel free to email me with any comments or questions.
The Overclocking Process
Wow, we've finally gotten to the meat and potatoes of this guide - the actual process of overclocking your processor. Well, shall we begin?
The first step in this process is determining what steps you will need to take to overclock your system properly. For the sake of some minor brevity, I will assume you are using an Intel Celeron/PII/PIII processor and a BX motherboard of some kind. These systems are the ones that are most commonly overclocked, and besides, overclocking the Athlon takes more than a little technical know-how, it also requires quite a bit of manual coordination. I will also assume you will only be using some form or relatively simple passive air cooling, as opposed to a Peltier or similar unit.
For an Intel based system, the first thing to check is whether or not you can configure the processor speed and internal voltage within the BIOS. Most newer motherboards, as well as most older ABIT boards, support "jumperless" configuring. If you are lucky enough to have such a system, your job is simple. All you need to do is raise the processor speed to what you want, change the voltage if necessary, and you're gold. If you have to deal with the jumpers, however, you've got a chore for yourself.
To change the computer's settings using jumpers, you are going to need your motherboard manual, or a copy of a jumper map for the particular model of motherboard. Using the jumper maps, determine the proper settings for the wanted processor speed and core voltage. Then move the jumpers on the motherboard so that they match the jumper map. To do this, you may need a pair of tweezers and a pen light.
Woo-hoo, congratulations, now your system is overclocked. But wait, that was too simple - what gives? Why did I write this huge guide about overclocking if the steps were easy as 1-2-3? Well, we aren't quite done yet.
The next step in this process is to attempt to start up your computer. The first test is seeing whether or not the computer will post. Posting is the process of the computer initializing the BIOS and loading up the system settings. If your computer won't even do this, there is almost no chance you will ever get it to run at the configured speed. Go back and lower the processor speed and try again.
If the computer did post, however, but it won't boot up Windows, you may want to try going back and upping the chip's core voltage. Take this as a word of advice though, make sure that you don't raise the core voltage too much - generally no higher than 0.3 V above default. If you go much higher than this, you run a serious chance of permanently damaging your computer.
Assuming your computer starts to boot Windows, but crashes before the system begins to settle down, you have two options as to how to deal with the crash. You can either go back and change the system's core voltage or you can add more cooling. That may include adding more case fans, installing a larger and more powerful fan/heatsink combo, or both. At this point, if you are able to do both, you may even be able to reach a higher speed.
Ok, your computer boots. Great! But you aren't out of the dark yet, because you still have to check and see if the system is completely stable.
System Stability Testing
There are generally two types of testing that I perform on a newly overclocked system. One is an intensive integer/FPU test which keeps processor utilization up between 95 & 100% for upwards of a half an hour. If the CPU passes this test, the overclock on the CPU itself is stable. However, even if the system passes that test, I still run a gaming test. The gaming test determines how well the rest of the system responded to the overclock (this is particularly important when dealing with non-standard bus speeds and out-of-spec RAM. Another test that I recommend, if you own the software, is the SiSoft Sandra benchmarks, or alternatively, WinBench 2000. Both pieces of software do subsystem specific testing - something that can be very important, particularly if you are trying to determine which pieces of hardware within your system are causing a failed overclock.
The intensive integer/FPU test is aptly named Stability Test and can be downloaded at www.tweakfiles.com. To use this test, you need to configure it before you overclock. This test will not only keep CPU utilization up at 100%, it will also make sure that the system isn't making any mistakes. This test only takes about a half an hour to complete and is definitely worth the time. If your computer doesn't pass this test, first try to either add more cooling or up the chip voltage a little bit and see if it works - otherwise drop down to the troubleshooting section for a few tips.
The second test consists of either using Unreal in Flyby mode with everything turned on (OpenGL), or using 3Dmark 2000 in loop mode (D3D). You really should only use 3Dmark 2000 if your system's OpenGL drivers are less than satisfactory. If you don't have a copy of Unreal, but your computer has a robust OpenGL driver, another alternative would be to use the Q3 demo with xero's 'monkeycrusher' demo. If your system passes both the integer and gaming tests, you have got yourself a stable system. Congrats.
Troubleshooting a Failed Overclock
To do this kind of troubleshooting, your computer has to be booted into Windows. If it isn't, take the steps outlined above to increase the stability of your system. Once you are in Windows, you will need to either install a copy of SiSoft Sandra or Winbench 2000. Then take each of the subsystem tests and run them separately from each other. Make note of which subsystem tests cause the system to crash, and focus on those parts of the system. That may mean adding a hard drive fan, some RAM cooling, etc. Once you have focused on all of the parts of the system that cause a crash (this may include the processor itself as well), go back and go through the system stability tests once again. If you can't get the system to pass the tests now, you may need to go back and lower the speed of the processor.
If you still refuse to give up on your 'golden' speed, however, you may want to try the following things:
- Leave your computer's case open
- Put your computer closer to your air conditioner
- Move your computer farther away from any heating ducts
- Make sure the computer has at least 6" of breathing room (15 cm) between it and anything else.
- Put the computer closer to the floor and farther away from anything that creates heat (your subwoofer, monitor, etc)
- Immerse your system in supercooled Jell-O gelatin and avoid eating it's jiggly goodness (yes, that was a joke).
Electrostatic Migration and Burnout
Electrostatic migration and burnout are the two things most feared by overclockers. Burnout is very simple - an excess of heat builds up within the processor, permanently damaging the hardware, making it unusable. This is the main reason why people make $75 Celeron keychains and other such adornments. Electrostatic migration, however, is just as deadly, and much less known. Each transistor within the chip's core develops an electrostatic charge over time, much like the way iron can develop a magnetic charge that will linger after any electric current has subsided.
Within CPUs, this electrostatic charge is a dangerous entity if it begins to affect the other transistors around it. That is why it is so difficult for computer companies to develop new chips - they must take into account the need for buffering room between the transistors within the chip. As a particular transistor's electrostatic field increases, it can cause damage to the other transistors around it. Electrostatic migration shouldn't be a problem for most overclocked processors unless you exceed the company's maximum core frequency for that particular model of core.
If you are exceeding the maximum frequency for a particular model of core, you need to be careful as to how long you have your computer running and how many consecutive hours a day you have it turned off. The longer you have your computer turned off at one time, the longer it will take for electrostatic migration to affect your system. You can't stop electrostatic migration from occurring, but depending on how you use your computer will determine how fast it will begin to take hold. Don't worry about it too much though, because even in the worst cases, it still takes a few years before it starts causing problems.
Overclocked Processor Lifetime
Worst case scenario for the lifetime of a non-burned out overclocked processor is over two years, while most processors will continue to function after five or six years. Unless you don't upgrade your system except when it breaks (this is very uncommon among overclockers and tweakers alike), you should never run into a problem with your processor's lifetime.
Effect of Non-Standard Bus Speeds
Non-standard bus speeds can have a variety of effects on various types of computer hardware. Hard drives can miswrite data, CD writers can create even more coasters than usual, CD-ROM drives can refuse to function, RAM can refuse to work properly, etc., etc., etc. Some of these problems can be fixed by adding some rudimentary cooling, but many of them are simple limitations of the hardware which will limit a system's overclockability. Keep an eye out for these things because they can cause serious problems with a system.
Most of these problems occur when using bus speeds that exceed the PCI bus frequency. PCI is intended to run at 33 MHz. If you set it to ove 40 MHz or so (using 83, 124, or 133+ MHz bus), there is a fair chance your hard drive will lose everything. See page 3 of the Hard Drive Tweak Guide for more information.
Alternate Overclocking Methods
There are, of course, other ways to overclock your system. Most notably, SoftFSB allows for overclocking your processor from within Windows without restarting. This is an impressive feat, but it isn't always as stable as changing the settings using the hardware. When you decide to use SoftFSB to overclock your computer, take the following precautions:
- Save before attempting anything
- Progressively overclock the system - go up one speed step at a time to make sure you don't inadvertently damage the processor.
- Run the stability tests to make sure the overclock was completely successful.
Download SoftFSB here.
Well, there you go - you've got the low down on how to overclock a computer. If you are still not satisfied with the speed of your system, check out some of our other guides on how to overclock your video card or tweak your system. And as always, I'm available to answer any questions or receive any comments you may have.
All Content Copyright ęDan Kennedy; 1999