Enter The Matrix
The memory controller first sends the row address for the cell it wishes to address to the module logic. After a certain period of time, tRCD (RAS-to-CAS delay), the module makes the contents of the row available in interim storage. On modern RAM chips, this process takes two to three clock cycles. You can even have fractions such as 2.5 clock cycles (CL 2.5), since DDR RAM can send control and data signals on both the rising and falling edges of the clock signal, i.e., twice per clock cycle.
Once the contents of the row have been sent to interim storage, the controller will send out the CAS signal (column address strobe) that transmits the column address for the memory cell. It takes an amount of time equal to tCL (CAS latency) until the contents of the selected cell have been sent to the output register of the memory chip.
In BIOS, you can set the number of clock cycles available for the timings tRCD and tCL . The lower these values, the better your performance. A CL setting of 2.0 or even 1.5 is only possible on the fastest of modules.
If you are reading out adjacent data from the same memory row, the only factor determining the speed of the access is the CL timing, since the controller already knows the row address and doesn't have to query it again. Whenever the controller has to address different rows in a RAM chip, the time tRAS (row active time) will pass before it can move from one row to the next. The time tRAS is increased by the time tRP (RAS precharge time), which is needed to charge the circuits up to a higher voltage level. In other words, even fast memory modules need at least seven clock cycles for the entire process.
Modern DDR RAM chips are subdivided once again into four segments (banks), each of which represents a separate memory zone. Bank interleaving allows zones in different chip banks to be addressed simultaneously, thereby increasing the data rate. While data is being read out of one memory bank, a new data zone can be addressed in another bank. You can specify in BIOS how many RAM banks of the chip can be addressed at the same time. The fastest setting is "four."
Top Performance With 1 GB Of RAM Or More
Another important performance criterion is the amount of RAM installed. Image and video-processing applications get an enormous boost from more memory. Readings taken with Content Creation Winstone prove that Windows 2000 and XP systems don't really take off until they have 1 GB or more of RAM. The benchmarks show how heavily system performance depends on the amount of memory. Indeed, 512 MB RAM is the bare minimum for fast Windows XP systems. Long gone are the days of Windows 98 and Me, when 512 MB was the most memory that the majority of systems needed.
The maximum amount of RAM depends solely on the motherboard and its chipset. For more information, go to the "Memory Support" table below. In x86 systems, however, the maximum memory allowed is 3.5 GB, no matter how much RAM has been installed. The CPU simply cannot address any more memory. The remaining capacity is reserved to control the PCI circuits.
You should install as few RAM modules as possible. Reducing the number of chips on the module will also enhance performance and stability. The modules generally consist of eight or 16 chips.
The number of memory modules you use will have a direct impact on your command rate. The command rate specifies the number of clock cycles the memory controller needs to activate the modules and chips. If you've filled all your memory banks, you'll generally have to increase the rate from one to two clock cycles to keep your system stable. Unfortunately, that will also impair performance by up to three percent.
