I have decided to post some information here about memory latency. I came across this subject on a thread posted by me about my future Graphics Cards. It was sugested to me by Chapel that perhaps I should get Dual Channel Memory (May write an explanation sometime once I have figured it out). I found some and he then suggested that I should perhaps get some with lower latency. I was not quite aware what latency was (I had read one explanation in my gaming magazine, NAG). I have done some research and have posted all my findings here. Please notify me immediately, if anything I say is incorrect. Incorrect information is dangerous information.
Please feel free to add any information about Latency but please do not clog up this thread with reviews of different memory modules. If you want to give reviews of how good different memory is please make a new thread. Of course I can't force you to so be courteous. So to the review
My main source was Extreme Tech which is a great site for most things PC hardware related. They an excellent review of how much latency affects a computer which you can access via the above link. Now to the explanation.
When you see a stick of memory advertised you will likely see 4 or 3 numbers. The Rule is the smaller the number, the better the memory. The numbers on a stick of DDR1 memory will likely look something like 2.5-2-3-5. Each of these numbers has its own meaning. The explanation of these is a bit technical if you are not aware of the way your computer works. However all you have to understand is the fact that your computer does things in clock cycles. Each time your CPU does a computation this is called a clock cycle. However your memory also has a clock speed. This is what you see when it says DDR333. It means that the memory can be accessed at 333Mhz. This is substantially slower than your average CPU and 333 is a bit slow for memory these days. So if your memory, like mine, has a 333 MHz rate then your computer can access 333 million computations a second . This is large amount which means that a clock cycle does not take a long time. If I have worked this out right then it takes my computer 0.000000003 seconds to do one cycle. In other words it takes 30 billionths of a second.
The numbers on the memory all tell you how many memory clock cycles it takes for the memory to be accessed. So you may think that it takes barely any time at all to access data. And you would be right; it takes nearly no time to receive a single piece of data. However when you are grabbing the huge amounts of data vital for running of games it starts to slow things up. This is when having low latency memory modules can speed things up. There are different processes of retrieving memory and just to confuse you someone decided to reverse the order in which they are written. They way that it works is as follows.
The numbers stand as follows CAS-tRCD-tRP-tRAS.
CAS: Coloumn Acess Strobe. An important thing to understand is that memory is that individual pieces of memory are saved in column and rows. The first thing the computer does is select the row and it then selects the column that it is in. Because of this, the column is the last thing that the computer does. However it is also the thing that the computer that does the most. Because of this the speed at which your memory selects the column affects the speed of your computer. Basically you want as few cycles here as possible
tRCD: RAS to CAS Delay. This is the amount of time that it takes between the Row being selected and the first Column being selected again in clock cycles. This does not have a great affect on the overall performance on the memory as the column is not reselected very often
tRP: RAS Precharge. This is the amount of time that it takes for one row of memory to stop being accessed and another row being accessed. This only has a great affect when you have a program that spans a large number of rows. However most games span a large number of rows so this does come into play when playing games.
tRAS: Active Precharge Delay. This is a confusing process. Basically it is the amount of time that it takes from when the memory receives the signal to retrieve or write and for it to actually select the Row. This does not affect the performance that much unless there is a large change in the memory stored such as starting application.
So in chronological order tRAS – tRP – tRCD – CAS. Of course tRP does not necessarily need to be used if the memory is already accessing the right row. In that case tRCD won’t be used either as a column will already have been used. So really the only thing that truly affects the performance of the memory is the CAS. However how much this truly affects the performance is questionable. If you are interested in how much Latency affects your performance then check out the link that I have included at the top of this post to the extreme tech website.
I hope this helped someone and please ask any questions you may have or any mistakes you have spotted (I have included two, lets see if you can spot them).