For XP system,
FSB is a parameter that
directly determines how fast the memory modules can commuicate with the CPU, or the
operating frequency of the memory modules.
The following describes a more detailed picture. Under the SYNC mode of memory operation (the preferred mode for XP systems),
CPU_frequency = FSB x CPU_multiplier
memory_frequency = FSB
In the formula, FSB stands for the FSB clock frequency.
The specification and quality of memory module and motherboard chipset determine how high in frequency the FSB can go.
Once the max FSB is determined, for multiplier unlocked CPU, the CPU frequency can be set according to the above formula, also as high as possible.
For system with PCI/AGP lock, FSB can be adjusted as high as possible, so is the memory frequency, independent of the PCI/AGP frequency.
FSB connects the CPU and the north bridge, in which the memory controller is located. Memory modules are connected to the memory controller where I/O and control for the memory are carried out. Data is transferred between the CPU and the memory via the FSB and the memory controller.
As FSB comunicates between CPU and the northbridge chipset, it also carries video data between the CPU and the northbridge which dispatches them to the AGP bus then the video subsystem (video card), as well as all other data between the CPU and the chipset northbridge, so FSB frequency is crucial not just to memory performance, but the
overall performance of the entire system.
For XP system, the memory bus is made up of 64-bit data path, i.e. 64-bit of data is moved each time or in each transfer of data.
For one FSB clock, there are two transfers of 64-bit memory data using both the rising edge and the falling edge of the clock pulse (called DDR or double data rate transfer), and so 2 x 64 = 128-bit of data is transferred. 128-bit is the same as 16 byte of data, as 1 byte = 8 bit by definition.
Typically, for a Nforce2 system rev 2 (aka Ultra 400), FSB is rated to run at 200 MHz, hence the max data transfer rate (or called
bandwidth) is 16 x 200 MB/s = 3200 MB/s. This is the max amount of data that can be passed between the CPU and the memory per second.
Typically, for XP Nforce2 system, the CPU can absorb about 90-95% of that memory bandwidth, or called effective bandwidth. That % number is usually called memory bandwidth efficiency and it measures how effective a system setup is on memory performance.
Related link:
Single Channel vs Dual Channel DDR Memory
Dual Channel, Nforce2, P4 and AMD FSB
What is cycle time and frequency
Frequency, clock, period of synchronous operations, latency
Latency
Analogy on Bus Speed, Bandwidth and Latency
Analogy for FSB, CAS2, CAS3 latency and bandwidth for DRAM memory