memory timings

[gunsmoney]

ok so, memory is definitely not one my strong points. I hear people talking about timings all the time, and i am not sure what the heck they r talking about. when someone says to tighter their timings or loosen their timings, what does that mean they r doing to the RAM? i would just like to find out more information, cuz i am always lacking in this area. thanks

[Phase]

There are probably four or five timings which most people tweak. To tighten them is to try to give them lower values, to loosen is to set higher values. It is a performance factor kind of like memory speed and what may work with timings at one speed for a certain memory voltage, might not work at another. For overclocking the memory to find the highest speed, people may want to loosen timings. After getting to a certain speed that they want, they might then work at trying to tighten the timings.

[Archer]

http://img11.imageshack.us/img11/1681/reallty.jpg
This gives an idea about what timings do.

[GeorgeStorm]

It depends, both increasing the freq and decreasing the timings increase the bandwidth that your RAM provides
Int eh past, tighter timings were more important, now, program generally prefer higher freq.
Use Eversest Mem benchmark to test various settings, and see which gives you the best bandwidth.
Raising the FSB wilil also dramatically increase the bandwidth.

[Archer]

This being true my main concern is I personally don't think all that bandwidth is necessary without the timings to all the cpu to keep the pipes full. Everest is great but is it going to help me encode faster? Sandra is great but the numbers are not generally backed op by the data from actual tests. I am just glad to see people paying attention to some of these threads as I personally believe that AMD and fast DDR3 will be a bad combination at this time and unless the DDR3 controller yields some benefit then there may be a net performance loss. I heard a lot about loosening timings to get a better oc well that may impact the CPU but some of that gain is negated by the memory timings.

[GeorgeStorm]

Well, in my opinion, you can never have too much bandwidth, :D
But, you are right
Unless you have a heavily clocked quad, you are unlikely to be able to use all the bandwidth available to you, however, I dont think it will actively harm performance, but it is definitely not being all used, and so, I suppose, wasted power. (I just like having lots of bandwidth coz I can get it, and coz im only running 2gb RAM, :P)

[Archer]

Not a problem as you may need the bandwidth it you do a lot of non cpu intensive memory work or as you put it to paraphrase "I like my bandwidth". Bandwidth is good, timings are good, both are GREAT!!!!

[GeorgeStorm]

Yeah, I mean obviously both freq and timings = bandwidth, and it depends on the program whether it gets more of a boost from either.
I dont need all of the bandwidth, its just coz I can get it, so I do get it, :P

[KTE]

Read over this article, research it, spend more than a month learning and testing your understanding, ask if unsure but not until you've put plenty of effort in and hopefully confusion will be a thing of the past. You can't beat the articles this guy wrote in his expertise: http://www.lostcircuits.com/mambo//index.php?option=com_content&task=view&id=35&Itemid=60

If you can understand that, RAM becomes a piece of cake to understand. To comprehend the various intricacies of memory in a particular system, have a skim over the technical documentations from now and then (i.e. from Intel or AMD). That will fill in the specific reasons, limitations and blanks applicable to each architecture. ;)

To start off, for RAM, two key words are 'throughput' and 'latency', just like for HDDs. Throughput is measured in Bytes per second whereas latency is measured in nanoseconds (for RAM).

[DrNip]

Read over this article, research it, spend more than a month learning and testing your understanding, ask if unsure but not until you've put plenty of effort in and hopefully confusion will be a thing of the past. You can't beat the articles this guy wrote in his expertise: http://www.lostcircuits.com/mambo//index.php?option=com_content&task=view&id=35&Itemid=60

If you can understand that, RAM becomes a piece of cake to understand. To comprehend the various intricacies of memory in a particular system, have a skim over the technical documentations from now and then (i.e. from Intel or AMD). That will fill in the specific reasons, limitations and blanks applicable to each architecture. ;)

To start off, for RAM, two key words are 'throughput' and 'latency', just like for HDDs. Throughput is measured in Bytes per second whereas latency is measured in nanoseconds (for RAM).


Nice post. I will be reading this.

[gunsmoney]

yea thanks for the replies guys, i will definitely be checking that article out