A Brief History of Overclocking
This was written nearly a decade ago, but the methods haven't changed much. Sometime I'll post a followup covering the past decade.
Overclocking is the practice of running computer equipment faster than it was designed to run at. It is analogous to tuning sports cars. Instead of horsepower and ft-lbs. of torque, its units are FLOPS and megahertz, but the drive for performance is the same. Overclocking has various reasons, several methods, and opposing views by the computer industry.
Overclocking is done to extend the useful life of aging computers, to test the limits of new computer hardware, and to have usable but faster than available computers. Users who do it to extend the useful life of their computers generally do so only until they can purchase faster components, which may be beyond their means at the moment. They rarely do anything more significant than replacing a heat sink or fan with a more efficient one. These computers are analogous to modifying a typical four-cylinder car. Those who test the limits of new computer hardware are the opposite extreme, frequently using solid-state thermoelectric coolers or even liquid nitrogen in their quest for gigahertz. As of now (April 2002) they have already reached four gigahertz for brief periods of time using liquid nitrogen. These computers are akin to top-fuel dragsters; they're faster than anything else out there, but can only be run for brief periods. The third camp is the middle ground between the other two. These users run their computers faster than what is currently available, but close enough to the rated speeds to maintain stability. These computers are similar to muscle cars.
Since its birth in the 1980's, several methods have been devised for overclocking. These include changing the clock crystal, changing the bus speed, and changing the processor clock multiplier. The oldest method of overclocking is changing the clock crystals. All computers use a crystal with a resonant frequency of 14.618 megahertz, which is then used to set the various clock frequencies used throughout the computer. By changing the crystal used, the entire system may be run faster than otherwise possible. The price of this is that some components require a specific clock speed in order to function properly, or at all. One solution developed to avoid this problem is the use of two clock generators. The one on the motherboard is left at 14.618 megahertz, while a faster crystal is used with a separate device to generate the frequencies for the components to be overclocked. The appropriate connections on the motherboard are severed from the motherboard's clock generator, and connected to the external one. The difficulties with this generally limit it to individuals with training in electrical engineering.
In the middle part of the 1990s, it was possible to overclock by changing the multiplier for the processor via jumpers on the motherboard. This method eliminates most of the problems of the previous method, as only the processor is affected. However, unlike the previous method, it is limited to the multipliers supported by the motherboard. This only works when the processor does not have a hardwired multiplier, as was necessitated by a form of fraud called remarking, explained below. This leaves one method available: changing the system bus speed. This is between the other two methods in comprehensiveness. It affects many but not all components. The ones that require a fixed frequency stay at their specified speed, while the ones that can be run faster are. Recently, some motherboards have the capability of fixing the speed of some of these, as not all devices on in particular the PCI (personal computer interconnect) bus (especially hard drives) are tolerant of out-of-specification speeds. System bus overclocking is limited by the capabilities of the motherboard, but can be used with both processors with a hardwired multiplier and those without.
The computer industry has its own views on overclocking. In particular, the processor manufacturers are generally opposed to it, while motherboard manufacturers generally embrace it. The world's dominant processor manufacturer, Intel, was the first to hardwire the multiplier into their processors, starting with the Pentium II processors. This was in response to the aforementioned process of remarking, in which a processor that is proven to or speculated to be able to run at a faster speed than it is rated has its model number and other identifying characteristics removed by a reseller and replaced by the corresponding characteristics of a faster but otherwise identical component, which is then sold on to computer manufacturers or other customers. As the majority of Intel's customers are manufacturers, the end user has less use for overclocking capabilities in the processor, so Intel locked the multiplier on all their processors except those designated as engineering samples, which are quite literally viewed by many overclockers as worth more than their weight in platinum.
Intel's primary competitor, Advanced Micro Devices (AMD), is more receptive to overclockers, as they account for significantly higher percentage revenue for AMD than for Intel. AMD also instituted a multiplier lock several years ago, but made it relatively easy to defeat with, depending on the processor model, either a sharp pencil or conductive ink for closing several bridges on the chip. However, unlike previous processors that could be remarked, these processors are easily visually inspected for tampering. The majority of motherboard manufacturers, however, target the "performance" market, as the overclocking community is referred to by the industry. With very few exceptions, motherboard manufacturers have found it necessary to offer overclocking options on the motherboards designed for retail sales, as opposed to manufacturer sales. One benefit of targeting the "performance" market is that these motherboards tend to be more stable when running within specifications when compared to those that do not offer overclocking features.
Overclocking has a long history, dating back to the early days of personal computers. It is by and large fueled by the same drive for performance that leads car enthusiasts to modify high-performance vehicles for even more performance, often at a loss of reliability. Extensive information about overclocking can be found at Overclockers.com. More technical and formal information about both overclocking and the processors used is available in Ars Technica's CPU Theory & Praxis section.
Overclocking is the practice of running computer equipment faster than it was designed to run at. It is analogous to tuning sports cars. Instead of horsepower and ft-lbs. of torque, its units are FLOPS and megahertz, but the drive for performance is the same. Overclocking has various reasons, several methods, and opposing views by the computer industry.
Overclocking is done to extend the useful life of aging computers, to test the limits of new computer hardware, and to have usable but faster than available computers. Users who do it to extend the useful life of their computers generally do so only until they can purchase faster components, which may be beyond their means at the moment. They rarely do anything more significant than replacing a heat sink or fan with a more efficient one. These computers are analogous to modifying a typical four-cylinder car. Those who test the limits of new computer hardware are the opposite extreme, frequently using solid-state thermoelectric coolers or even liquid nitrogen in their quest for gigahertz. As of now (April 2002) they have already reached four gigahertz for brief periods of time using liquid nitrogen. These computers are akin to top-fuel dragsters; they're faster than anything else out there, but can only be run for brief periods. The third camp is the middle ground between the other two. These users run their computers faster than what is currently available, but close enough to the rated speeds to maintain stability. These computers are similar to muscle cars.
Since its birth in the 1980's, several methods have been devised for overclocking. These include changing the clock crystal, changing the bus speed, and changing the processor clock multiplier. The oldest method of overclocking is changing the clock crystals. All computers use a crystal with a resonant frequency of 14.618 megahertz, which is then used to set the various clock frequencies used throughout the computer. By changing the crystal used, the entire system may be run faster than otherwise possible. The price of this is that some components require a specific clock speed in order to function properly, or at all. One solution developed to avoid this problem is the use of two clock generators. The one on the motherboard is left at 14.618 megahertz, while a faster crystal is used with a separate device to generate the frequencies for the components to be overclocked. The appropriate connections on the motherboard are severed from the motherboard's clock generator, and connected to the external one. The difficulties with this generally limit it to individuals with training in electrical engineering.
In the middle part of the 1990s, it was possible to overclock by changing the multiplier for the processor via jumpers on the motherboard. This method eliminates most of the problems of the previous method, as only the processor is affected. However, unlike the previous method, it is limited to the multipliers supported by the motherboard. This only works when the processor does not have a hardwired multiplier, as was necessitated by a form of fraud called remarking, explained below. This leaves one method available: changing the system bus speed. This is between the other two methods in comprehensiveness. It affects many but not all components. The ones that require a fixed frequency stay at their specified speed, while the ones that can be run faster are. Recently, some motherboards have the capability of fixing the speed of some of these, as not all devices on in particular the PCI (personal computer interconnect) bus (especially hard drives) are tolerant of out-of-specification speeds. System bus overclocking is limited by the capabilities of the motherboard, but can be used with both processors with a hardwired multiplier and those without.
The computer industry has its own views on overclocking. In particular, the processor manufacturers are generally opposed to it, while motherboard manufacturers generally embrace it. The world's dominant processor manufacturer, Intel, was the first to hardwire the multiplier into their processors, starting with the Pentium II processors. This was in response to the aforementioned process of remarking, in which a processor that is proven to or speculated to be able to run at a faster speed than it is rated has its model number and other identifying characteristics removed by a reseller and replaced by the corresponding characteristics of a faster but otherwise identical component, which is then sold on to computer manufacturers or other customers. As the majority of Intel's customers are manufacturers, the end user has less use for overclocking capabilities in the processor, so Intel locked the multiplier on all their processors except those designated as engineering samples, which are quite literally viewed by many overclockers as worth more than their weight in platinum.
Intel's primary competitor, Advanced Micro Devices (AMD), is more receptive to overclockers, as they account for significantly higher percentage revenue for AMD than for Intel. AMD also instituted a multiplier lock several years ago, but made it relatively easy to defeat with, depending on the processor model, either a sharp pencil or conductive ink for closing several bridges on the chip. However, unlike previous processors that could be remarked, these processors are easily visually inspected for tampering. The majority of motherboard manufacturers, however, target the "performance" market, as the overclocking community is referred to by the industry. With very few exceptions, motherboard manufacturers have found it necessary to offer overclocking options on the motherboards designed for retail sales, as opposed to manufacturer sales. One benefit of targeting the "performance" market is that these motherboards tend to be more stable when running within specifications when compared to those that do not offer overclocking features.
Overclocking has a long history, dating back to the early days of personal computers. It is by and large fueled by the same drive for performance that leads car enthusiasts to modify high-performance vehicles for even more performance, often at a loss of reliability. Extensive information about overclocking can be found at Overclockers.com. More technical and formal information about both overclocking and the processors used is available in Ars Technica's CPU Theory & Praxis section.
