Moderate Overclocking of Processors

Moderate Overclocking of Processors

(Based on materials and with the permission of Russian magazine ComputerPress.)

CPU overclocking is based on the performance growth of the processor's core and integrated L2 cache memory (and, consequently, on overall performance growth), caused by an increase in the operating clock frequency. Both the core and L2 cache operate at the internal CPU clock frequency.

Click To expand
Figure 3.8: Core, cache memory, and buses of an Intel processor

Besides the CPU clock frequency, other parameters must be considered, such as the Front Side Bus (FSB) frequency and the memory-bus frequency.

The FSB frequency is the main operating frequency of the entire computer system. All other frequencies, which define the operations of computer subsystems and the data exchange among them, are synchronized with it.

In contemporary computers based on the Intel Pentium 4 processor, the FSB frequency can take values of 100 MHz, 133 MHz, and 200 MHz. It is expected that the FSB frequency will increase with technological advances and improvement of the architecture of semiconductor elements. On this bus, the duration of one clock is determined by rectangular voltage pulses. The arrival time of each new pulse is defined by its front (leading edge).

The processor bus, also known as the system bus, connects the processor to the North Bridge of the chipset. This bus is used by the processor for communication with all the other devices.

Click To expand
Figure 3.9: CPU and RAM connection to the North Bridge

In computers based on the Intel Pentium 4 processor, data travel on the system bus at the frequencies of 400 MHz, 533 MHz, and 800 MHz. This means that in the Intel Pentium 4 processor, the data-transmission frequency is four times that of the FSB frequency, and the address-transmission frequency is twice that of the FSB frequency. If the FSB frequency is 100 MHz, then the data-transmission frequency is 400 MHz, and the address-transmission frequency is 200 MHz. If the FSB frequency is 133 MHz, then the data-transmission frequency is 533 MHz. At an FSB frequency of 200 MHz, data are transmitted at 800 MHz. This value usually is provided in CPU and motherboard specifications.

Besides the frequency, the processor bus is characterized by its throughput (i.e., by the maximum amount of data that can travel via this bus per second). The CPU bus has a width of 64 bits, which means that it can transmit 64 bits (8 bytes) per clock. Consequently, if it operates at 400 MHz, its throughput is 3.2 GB/sec (400 MHz × 8 bytes). At 533 MHz, its throughput will equal 4.2 GB/sec.

Click To expand
Figure 3.10: Organization of data transfer via the SDR SDRAM bus
Click To expand
Figure 3.11: Organization of data exchange via the DDR SDRAM bus

The memory-bus frequency defines the speed of data exchange between the processor and the memory controller (located in the North Bridge). This frequency depends on the memory type and is synchronized with the FSB frequency. For the most common types of DDR memory, data transfer occurs twice per clock, both at the leading and trailing edges of the clock pulse. This means that the effective frequency of a memory operation is twice the clock frequency. The predecessor of DDR memory — SDR SDRAM — transmitted data at a frequency that coincided with the memory-bus frequency.

For DDR200, DDR266, DDR333, and DDR400, the effective frequency that defines the data-transmission speed is equal to 200 MHz, 266 MHz, 333 MHz, and 400 MHz, respectively. The clock frequency is 100 MHz, 133 MHz, 166 MHz, and 200 MHz, respectively. The memory-bus frequency is synchronized with the FSB frequency. For example, if the FSB frequency is 133 MHz, the memory frequency relates to the FSB frequency as shown in Table 3.2.

Table 3.2: Frequency Relationship between DDR Memory and FSB

DDR memory frequency (MHz)

FSB frequency (MHz)


Throughput (GB/sec)

















Like the system-bus frequency and the memory-bus frequency, the CPU clock frequency is synchronized with the FSB frequency. The coefficient that relates the CPU clock speed and the FSB frequency is known as the multiplier. For example, if the FSB frequency is 133 MHz and the multiplier equals 18x, Intel Pentium 4 processor will operate at a frequency of 2.4 GHz. For an Intel Pentium 4 processor with a clock frequency of 2 GHz, the multiplier will equal 20x, provided that the FSB frequency is 100 MHz.

At first, it might seem that the easiest way to increase the internal (core) clock frequency of the processor is to increase the multiplier value. If you have an Intel Pentium 4 processor operating at 1.6 GHz with a standard multiplier of 16x (and an FSB frequency of 100 MHz), you can turn it into an Intel Pentium 4 with a frequency of 2.4 GHz by simply changing the multiplier to 24x. This method is simple and reliable, but, in all contemporary processors, including those of the AMD Athlon family, the possibility of changing the multiplier is disabled. In AMD processors, it is still possible to sidestep this limitation using some tricks. (Information on this will be described in the appropriate sections of this book.) It is impossible to overcome this limitation in Intel Pentium 4 and Intel Pentium III.

Don't let this upset you. Recall that the CPU clock frequency is synchronized with the FSB frequency. Therefore, if you increase the FSB frequency, the CPU clock frequency will increase automatically. Motherboard manufacturers provide the capability of increasing the FSB frequency. For example, the nominal clock frequency of an Intel Pentium 4 processor is 2.4 GHz at an FSB frequency of 133 MHz (which means the multiplier is set to 18x). By increasing the FSB frequency to 180 MHz, the CPU frequency increases to 3.2 GHz (Table 3.3).

Table 3.3: Increasing the CPU and Memory Clock Frequencies with the FSB Frequency

FSB frequency (MHz)

CPU clock frequency (GHz), 18x multiplier

System bus frequency (MHz)

System bus throughput (GB/sec)

DDR266 memory frequency (MHz)

Memory throughput (GB/sec)





































Overclocking a system by increasing the FSB frequency also increases the memory frequency, because RAM is synchronized with the processor clock frequency. This circumstance is important, although it often is overlooked by PC users. You can never know what will die first: the memory or the processor. Furthermore, the memory usually is the main hindrance to overclocking; it can prevent you from increasing the FSB frequency. If the processor is capable of operating at an FSB frequency of 180 MHz but the memory is unable to support FSB frequencies that exceed 150 MHz, you are limited to the FSB frequency of 150 MHz. For this reason, overclocking strongly depends on the quality of the memory modules.

There are two main ways of overcoming the limited capabilities of memory overclocking. First, by tuning the BIOS settings, it is possible to change the ratio between the FSB frequency and the memory frequency, making the frequency of the memory bus as small as possible. Because overclocking results in a proportional increase of the FSB and memory frequencies, it is possible to create conditions under which the processor is overclocked to higher frequencies than the memory. Suppose that the system is designed to run at an FSB frequency of 133 MHz and to use DDR266 memory modules. This means that 266 MHz is the nominal frequency for the memory. If you specify the coefficient that relates the FSB and memory frequencies as 1.5, then, at an FSB frequency equal to 133 MHz, the memory frequency will be 200 MHz. This will be below the nominal value. When you overclock the FSB frequency to 177 MHz, the processor also will be overclocked, and the memory will operate at its nominal frequency of 266 MHz. This method of artificially slowing the memory is used quite frequently. However, it has drawbacks: When the maximum FSB frequency is reached, the memory still may run at a frequency below the nominal value.

For example, suppose that you have an Intel Pentium 4 processor with a nominal frequency of 2.4 GHz (and a multiplier value of 18x). The nominal value of the FSB frequency is 133 MHz, and your computer is equipped with DDR266 memory modules. If the coefficient that relates the memory and FSB frequencies is set to 1.5, you might manage to overclock the FSB frequency to 160 MHz. In this case, the CPU clock frequency will be calculated as follows: 160 MHz × 18x = 2.88 GHz. (This is not a bad result.) However, the memory will operate at a frequency calculated as follows: 160 MHz × 1.5 = 240 MHz (i.e., below its nominal frequency). Which is better — to raise the CPU clock frequency and decrease the memory frequency, or to try to overclock both the CPU and RAM?

This example that synchronizes the CPU and memory operations is not an artificial one. Overall system performance depends on the CPU and memory frequencies. Thus, true overclocking means finding optimal values. The conditions that provide a maximum increase in overall system performance often have to be defined experimentally.

Another popular method uses faster memory than recommended in the motherboard specification. For example, DDR333 or even DDR400 memory modules can be used with motherboards supporting DDR266/200 memory. If you combine both approaches, you can achieve high FSB frequencies without being limited by the memory capabilities.

In addition to frequency, other important parameters influence memory performance. These are memory timings. In most cases, tuning these will achieve significant performance gain. These will be covered in more detail in the appropriate sections of this book. For the moment, however, it is necessary to make some important notes.

Before you start processor overclocking, it is necessary to improve the cooling system. You should correctly install the cooler and the heatsink. At first, it may seem that there are no difficulties here, but this impression is wrong. To achieve adequate cooling, a layer of thermal paste usually is placed between the surfaces of the processor and heatsink. The only exception is the so-called box coolers, included with the kits supplied by CPU manufacturers.

Click To expand
Figure 3.12: Box kit comprising a CPU and cooler
Click To expand
Figure 3.13: Box cooler

Such coolers have a special heat-insulation layer; therefore, they do not require thermal paste. In all other cases, thermal paste is a must. If there is not enough thermal paste (for example, the CPU surface isn't covered entirely), or if the paste has been dried up, you must remove the paste using a special dissolvent and apply a new layer. Never try to scratch off the paste with the knife; scratches on the surface of the heatsink or CPU will degrade heat dispersion. The new layer of thermal paste must be distributed evenly over the entire surface of the CPU cover. This layer mustn't be too thick; a heavy layer also would degrade heat dispersion. The optimal thickness of the layer is 0.5 mm.

You should consider purchasing a high-quality heatsink with a turbo-cooler. Box coolers usually are not intended for extreme modes. They do allow moderate overclocking, because box coolers have some power reserve.

Besides this, it is necessary to consider installing an extra cooler into the system unit. Although this will make the computer a little noisy, it will improve the efficiency of the cooling system considerably.

The right choice of motherboard also is of great importance. Its design, manufacturing quality, and used components must guarantee stable operation at high frequencies. Built-in tools, as well as BIOS, must support overclocked modes.

The ability to change the following is standard:

  • FSB frequency

  • Coefficient relating the memory and FSB frequencies

  • Memory timings

  • Processor and memory core supply voltage

Different manufacturers have different opinions about the overclocking problem. For example, on most Intel motherboards, this capability is locked. Other leading motherboard manufacturers, such as Abit, Asus, Gigabyte Technology, and MSI Computer, not only enable their users to change the previously listed standard settings, but they welcome the overclocking capability. Some manufacturers even include special overclocking utilities with their motherboards.

The advantage of special utilities used for system overclocking is that main settings are changed programmatically, rather than via traditional methods such as BIOS or special switches on the motherboard. Usually, this takes place when the operating system is up and running. As a result, it is not necessary to reboot the computer after each change is introduced — provided that the system operates reliably and remains stable. Otherwise, reboot is inevitable.

The functional abilities of such programs partially duplicate BIOS. Despite this, BIOS often provides a richer set of functionality, such as memory timing settings. However, special utilities significantly reduce the time required for system overclocking. Programs that support several motherboards can be purchased separately.

In concluding this section, it is necessary to mention that the final result of overclocking strongly depends on the quality of the components used. It also depends on the overclocker's skills and experience.

The examples of extreme overclocking confirm this statement.