Gomeler.com

Overclocking Series: Basics

02.24.06 - 08:09am

Disclaimer: By reading this article you agree that any damage that is sustained via modifying your computer will be solely your fault and hold the author liable for nothing. The purpose of these guides are to simply explain the processes of modifying a computer to run out of default specifications (over/under clocked).

On the internet there are a plethora of resources available that describe the process of over and under clocking a computer. Unfortunately much of this information is either terribly out of date or located within a forum which makes for a very hard read. I hope that the following guides will be very clear and easy to understand for the novice modifier yet still contain some information that will interest those that already understand how to overclock their computers. Now, let’s get into the actual purpose of this article.

Why?

overclock

/oh’vr-klok’/ vt. To operate a CPU or other digital
logic device at a rate higher than it was designed for, under the
assumption that the manufacturer put some slop into the
specification to account for manufacturing tolerances.

This is the definition of overclock as provided by Dictionary.com which does a good job of saying what overclocking is, but it is lacking in details. Generally when somebody overclocks they are trying to exact more performance out of their current hardware at no expense, sort of like a free lunch. An example of this would be buying a relatively cheap CPU, say an Athlon64 3000+ that runs at a respectable 1800 megahertz (MHz) and try to overclock it to about 2200mhz which is the speed that the Athlon64 3500+ operates at. This gain of 400mhz is considerable when you look at the prices of these two chips, as of today, Febuary 24st, 2006, the 3500+ is running at $223 shipped while the 3000+ is running at a cool $118 shipped. So basically you save yourself $105 that can be put into something else, like a video card, or saved for the future.

Besides saving money, overclocking provides some people a challenge, how far can you push chip X before it becomes instable, or how many more frames per second can you achieve with video card y? These are just two examples, but the principle is clear, it is a challenge, just like any other activity that people pursue in their lives.

Looking back, we have saving money and the challenge as reasons for overclocking, but one other reason would be extending the lifetime of a machine. I personally can relate to this, my main machine runs on an Athlon XP 2400+ CPU that ran stock at 2000mhz and currently runs 2200mhz but with a 400mhz FSB instead of the stock 333mhz. If these terms don’t make sense, basically I’ve made my computer run more efficient and slightly faster and have seen significant gains in many applications, and it has kept me from upgrading my computer for 4 years now. That rounds out all the reasons I can think of, Money, Challenge, Longevity. On a side note, overclocking decreases processor lifetime, we’ll get into that later, but I might have lowered the CPU’s lifetime from 12 years to a lengthy 8.

The Basics

Being able to overclock your computer depends less on the processor in your computer and more on the motherboard. If you are sitting on a Dell or Hewlett Packard, chances are you won’t be capable of overclocking due to the restraints placed on the BIOS. For those that don’t know, the BIOS is the control center that sets all the frequencies and values for your motherboard. If you are sitting on a “custom” computer that was built by a non-OEM manufacturer then you’ll be enter your BIOS and tweak things. Now one thing to remember is that not all motherboards are created equal. In most cases, you get what you pay for so skimping on your board will hurt you in the long run. Research your individual motherboard and figure out if it is a good platform to start with, you can cause some serious damage or be limited if your motherboard isn’t up to the task. Your motherboard should let you alter the multiplier, Front Side Bus/Clock generator, memory dividers, memory timings, CPU/Chipset/Memory voltages, and HyperTransport multiplier. Because this is a very general guide, some things mentioned may not be part of your computer, such as the HyperTransport multiplier or chipset voltages. If you have all of these, then you are set to go. One additional thing that I like to see on boards are CPU shutdown temperatures. I usually set mine around 50 Celsius so that incase I seat my heatsink incorrectly I don’t fry my chip immediately. Now how about I demystify some of this stuff and start explaining how this stuff works.

Processor Frequency

Your CPU frequency is determined by the following equation.

CPU Frequency = Multiplier * Front Side Bus (FSB)

For a 3500+ that would mean that 2200mhz = 11 * 200mhz since A64’s operate with a 200mhz, 400mhz effective, FSB. Most modern processors, excluding the Extreme Edition Pentiums and Athlon FXs come with their multiplier toplocked. This means that any multiplier above the default, such as 12 for the 3500+ are not possible, but anything below stock is possible such as 10. The FSB however is a different story, this is where your overclocking will come from. If you increase the 3500+’s FSB from 200 to 220 this will net you 2420 MHz and a free 220 MHz. To show how this works on Pentium 4s, here is a quick sample. Pentium 4 651 runs at 3400 MHz = 17* 200 MHz. Increasing from 200 MHz to 220 MHz gives you 3740 MHz and a free 340 MHz.

As you slowly increase the FSB you are going to hit a wall where you can go no further. This could be due to many things including heat, voltage spikes/drops, and FSB limitation. I’ll tackle heat first since it is the main problem in overclocking.

Heat

When a processor is created by Intel, AMD, IBM, or any other chip fabrication company, it has to fit within many specifications, some being physical size, power consumed, and power dissipated. For the overclocker, power consumed and dissipated will be what concerns you the most. If a processor dumps a lot of heat from stock, overclocking it will only increase your problems. This is one of the reasons that people attribute to keeping Intel from breaking 4000 MHz with their processor line. The heat that a processor puts out is measured in watts and luckily I have found a few websites for AMD and Intel . The Intel and AMD whitepaper links are at the bottom of this article. For this review, I plan on using the Athlon 64 3500+ and the Pentium 4 651 as my examples for the remainder of the review unless otherwise mentioned. The stock Thermal Design Point (TDP) for the A64 is 67 watts at 2200 MHz and the P4 has a TDP of 86 watts at 3400 MHz. Just a word of warning, Intel and AMD rate their TDP’s differently with AMD giving worst case scenario and Intel giving the average TDP for average use. So with these two pieces of information, let’s see how an increase in processor speed changes the TDP. The following equation is what I have been using for years and it has been accurate enough for my cooling solutions.

TDP_new = (TDP_stock)*(Frequency_new/Frequency_stock)

For both processors, let’s increase the FSB from 200 MHz to 220 MHz and see how that changes things up.
Pentium 94.6 = 86*(3740/3400)

A64 73.7 = 67*(2420/2200)

As you can see, there is a slight increase with the change of frequency, but it doesn’t hurt you that much. Now there is another issue to deal with, what happens when you hit a point where you processor is unstable. You have two options, back down the FSB, or increase the voltage and charge forward.

Processor Voltage

In computing, your system operates by detecting the difference between a low and a high voltage. The low voltage is usually somewhere around 0 and the high can range from .8 volts up to 4 volts depending on the part in question. For our purposes, we’ll stick to our trusty 3500+ and 651 and their voltages. The A64 gets along on 1.35 volts and the Pentium zips around at 1.3375 volts. Now some people may say “Oh but Chris, Intel said their 651 runs between 1.2 and 1.3375 volts” and that is ok, but I’m going for the worst case scenario which is 1.3375 volts. So with that said, let’s discuss voltages. When your computer becomes unstable, adding voltage will help to make it stable again and increase the overclock that you can reach. The only downside is that increasing voltages does wonders to increasing your TDP, as you’ll see when we add a bit to our handy formula.

TDP_new = (TDP_stock)*(Frequency_new/Frequency_stock) *(Voltage_new/Voltage_stock)^2

As you can see with the addition of the Voltage to the formula, a small addition of voltage is squared, possibly putting you into a world of hurt. Let’s say that by increasing your FSB to 220 MHz your computer became unstable so you bumped up the voltages to 1.4 for the A64 and 1.4 for the 651. This might be overkill, but this is an example.
Pentium 103.6 = 86*(3740/3400)*(1.4/1.3375)^2

A64 79.3 = 67*(2420/2200)*(1.4/1.35)^2

Woah! Now that hurts, for the 651 that is 17.6 watt increase and the A64 suffers a 12.3 watt increase. So if you are using the stock heatsink provided with your processor, it might be possible to dissipate this extra heat and the temperature not change by much, but it is guaranteed your temperature will rise a little bit. The higher you go and the more voltages you pump, the more trouble cooling your chip will become. Here is an example of something extreme just to show how bad this can get.

On December 12th 2005, an overclocker managed to get an Intel 670 up to 7532mhz, so lets see how high the TDP went.

Pentium 419.8 = 115*(7532/3800)*(1.9/1.4)^2

420 watts is a lot of heat to cool, thankfully they were using liquid nitrogen to cool the processor, which we will get into later on when we hit exotic cooling methods.

I think this covers the basics of the processor and what overclocking does to it. The next article will cover memory frequency, memory timing, chipsets, and how all of this fits together. Here is a quick recap and hints.

Extra Stuff

TDP_new = TDP_old * (Clockspd_new/Clockspd_old) * (Voltage_new/Voltage_old)^2

Clockspeed should be in MHz to retain accuracy and use the TDP given by the chip manufacturer.

Heat is your enemy, make sure you can dissipate your heat output before trying something crazy. Running a 3.8 GHz Pentium on 2 volts on the stock heatsink would not be wise, use your head.

The heavier the heatsink, usually the more effective it will be. Copper should be chosen over aluminum in all instances, if anything a copper baseplate should be part of the heatsink.

Besides your heatsink, your powersupply needs to be beefy and stable. If you have a 500 watt powersupply that is very light, chances are the rails aren’t going to be stable under load. I highly suggest before building a new machine that you do some research into getting a proper powersupply. Some companies that I suggest looking into are Antec, OCZ, PC Power and Cooling, and Ultra Computer Product. Shop around for the right supply, there are many reviews available on the internet. Here is a link to a very good power supply review.
Overclock slowly in steady increments, you don’t practice for marathons by running them. Slowly bump up timings by 2-5 MHz at a time, and increase voltages by the minimum value.
Record everything on paper, when your machine refuses to boot from a heat or stability issue, you’ll lose your bios settings, potentially losing hours of work.

If you notice something that needs correcting, please leave a comment.
Whitepapers

A64 and FX, Opteron, 6×1 series, 9×0 series, and a general Intel TDP sheet

Category: Articles, Hardware, Overclocking