All About The Different Frequencies and Speeds Inside A Computer
Joined: 03 Apr 2008
Location: AVA HQ
Topic: All About The Different Frequencies and Speeds Inside A Computer
Posted: 20 May 2008 at 3:32pm
All About The Different Frequencies and Speeds Inside A Computer
Everyone seems to be concerned about the frequency of the processor these days. While a faster processor may bring about better performance, there are all sorts of different frequencies that your computer runs at. Matching these frequencies with each other helps make the computer use these clock cycles more effectively and efficiently thus bringing more performance to the total system. I plan on explaining the differences between them and how all these frequencies interact with each other.
Lets break down all the parts of system and explain them here. Later I will go over how they interact.
The motherboard has several frequencies which you need to be concerned about. There is the Front Side Bus (FSB) and then the maximum frequency ram supported that all motherboards state. However, how these speeds are calculated come from using the core clock frequency. This core clock frequency basically controls all the other devices frequencies (mostly CPU and ram), and this is mostly never stated. It is usually found in the bios and is somewhere around 200 MHz - 400 MHz depending on the motherboard.
Core Frequency - This frequency is usually never stated, but can be determined by referencing other frequencies stated. It determines the speeds that all the devices connected to the chipset/motherboard run at.
Front Side Bus - The FSB is the connection between CPU and the chipset. All Core 2 Family of processors are quad-pumped (sends four bits of data per clock), which means that the FSB is 4x the core clock frequency. A FSB of 800 MHz means that the core clock is 200 MHz. In a day and age when processors are being sold past 3.0 GHz, 200 MHz seems low. However, using technology to send more data per clock cycle and other tricks, we are able to squeeze more performance from the same old clock speeds. The width of the Core 2 FSB is 64 bits or 8 bytes (8 bits = 1 byte), so that 800 MT/s equals about
Memory Bus - Depending on the type of memory used, this bus can be stated at different frequencies such as DDR2-800 or DDR3-1600. However, the memory bus speed is usually rated in how many transfers it can do a second instead of the actual speed. Although it uses a MHz symbol, it is used so that people accustomed to older types of memory can reference that with the older and slower speeds. The memory bus speed is based upon the core frequency times a divider. The divider can be 1:1, 4:5 to slow it down, or even 5:4 to speed it up faster than the core clock. Most memories are limited to a real speed of around 400 MHz - 450 MHz. I will get into more detail later.
Direct Media Interface - The connection is the bus between the Northbridge and Southbridge of chipsets. Current generation of DMI currently is a serial connection running at 2 GHz.
PCI-Express - The PCI-Express standard uses a number lanes (described by x amount) to transmit data between device and controller. Version 1.0 had lanes clocked at delivering about 2.5 GT/s while version 2.0 doubled that to 5.0 GT/s. However, due to multipliers, the actual bus only runs at 100 MHz - 150 MHz. This bus is usually controlled by a separate clock, so it is usually unchangeable by changing the core clock frequency.
The processor also has several frequencies which you must be concerned about. These would be the FSB and Chip frequency, but also the multiplier. Lets go into how all these work.
Front Side Bus - As stated above, it is the maximum frequency at which the link between processor and chipset can operate at. Frequency stated is actually 4x the actual frequency.
Chip Frequency - This is the main frequency stated for the processor. The chip has a built-in multiplier which takes the mother core frequency and multiplies it by this number to achieve its rated frequency. Some lower-spec chips have locked multipliers, but higher-end chips have unlocked multipliers which are able to be changed in the bios. A chip frequency of 3.0 GHz will usually have a multiplier between 6 to 8, depending upon the type of chip family (consumer, enthusiast, extreme). So to get 3.0 GHz with a multiplier of 8, the motherboard will set its core clock to 350 MHz.
Multiplier - Although mentioned above, lets go into more detail here. This takes the core clock frequency and multiplies it to achieve the rated frequency of the chip. Chips which are manufactured to achieve high overclocks (Extreme Editions) usually have unlocked multipliers and are able so select in between multipliers (5.5, 6.5) to achieve more precision when overclocking. Normal chips targeted at consumers usually have locked multipliers or if unlocked, have multipliers that are low which are bad for overclocking.
The memory is one of the most confusing of the three components. Manufacturers continue to market higher and higher frequencies when in fact the same frequencies have been used for many years. The only thing changing is the amount of data being able to be moved and how many transfers a second can take place. Memory used in today's systems uses a bus that is 64 bits wide. Before we go into any more detail, lets go over what the difference between the DDR memories are.
DDR was the technology was able to double the data rate of the older SD Ram. It did this by making each memory cell retrieve two pieces of data at the same speed of the older SD memory. It then used a buffer to transmit the data on both the rise and fall of the core clock. This in fact doubled the rate at which data was transmitted but was still running at same frequencies.
The memory cells in DDR2 still run at the same frequencies, but achieve higher "bus frequencies" by retrieving 4 pieces of data on every clock cycle. The buffer would then again transmit this data on the rise and fall of every core clock frequency.
DDR3 is confusing to some people. It operates very much like the other DDR memories, but instead fetches 8 pieces of data per internal clock cycle. The buffer takes this data and transmits on both rise and fall of the clock.
Memory Frequency - The frequency of the memory is not its true frequency. It is stated again like the others as transfers per second, or effective frequency. The frequency of the memory can be calculated as 2 x Divider x Core Clock frequency. It is also the same thing as the memory bus speed in the motherboard section. The type of memory also plays a role.
Divider - The divider is very much like the multiplier in a CPU. The multiplier is stated as Memory Clock:Core Clock, so a divider of 4:3 means that the memory core runs 4/3 times faster than the core clock. The divider is able to be anywhere from 1:2 to 2:1 or in other words 50% to 200% the frequency of the core clock frequency. The divider has to be whole numbers and some motherboards only allow simple dividers such as 1:1 or 4:3 while some motherboards designed for overclocking may allow 8:5 or 5:3.
I/O Buffer - This is the true speed of the memory. This is directly related to the core clock by means of using the divider.
Timings or Latency - All memory being sold today includes numbers such as 5-5-5-15 or 7-7-7-24. These numbers refer to the number of cycles needed to perform a specific action such as reading data, sending data, preparing for another read. Although they do not influence frequencies, frequencies of the memory influence the timings. Usually the higher speed DDR2 and DDR3 memories have higher timings because it takes more time for a memory cell to read the higher number of bits needed. This means that DDR3 running at 1600 MHz is not running at double the speed of DDR2 at 800 MHz. The higher latencies will bring down the performance of the DDR3. However due to bursting technologies, sequential reads usually happen within the next clock cycle which makes DDR3 almost twice as fast with sequential reading.
How Do They All Relate?
Lets go over how all these relate to another. We will start with the core clock frequency (CF for short).
Now let us go over an example.
We will be using DDR2-800 memory, a 3.0 GHz processor, and mobo rated at 1333 FSB. The CF will be 1333/4 which equals 333 MHz. That will be our CF for the rest of our example.
Since the processor clock is rated at 3.0 GHz and we have a CF of 333 MHz, the multiplier for the processor will be 9.
The memory is rated to 800 MHz, which means that the I/O buffers are running at 400 MHz. Our divider for the memory bus would then be 6:5.
Overclocking and Performance
Now, the reason why anyone would want to care about this is either overclocking or looking to get the best performance from their system. Getting a motherboard that only supports 1066 MHz FSB will not able to push enough data that a 3.2 GHz processor can spit out. Also, the lowered core clock would limit the amount of memory that can be transferred as well.
Overclocking is another area where knowledge of how all the frequencies interact is very important. Changing the core clock will change all the different frequencies, while changing the multiplier or divider will only change the memory or processor. Raising the core clock from 333 MHz to 366 MHz (10% overclock) may be enough to cause instability, while raising the frequency of the processor by use of the multiplier may allow for an overclock from 2.6 GHz to 3.2 GHz (23%).
Edited by MattSlagle - 28 May 2008 at 10:43am
AVADirect Research and Developement
|Forum Jump||Forum Permissions
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot delete your posts in this forum
You cannot edit your posts in this forum
You cannot create polls in this forum
You cannot vote in polls in this forum