Why 3*4GB kit costs more than than 3*4GB?

What do you mean by manual testing? Why untested modules won't work in multichannel? How testing can improve this?

> All memory is slightly different the companies have to physically put the memory in a computer and ensure it works with the other sticks in the kit.

All memories are different so companies need to put them into my computer to test it. Two questions here: how do they test the compatibility of their memories with my computer if they dont even have it and, secondly, I repeat my question: why multichennel is different? You know, the mem module alone may be incompatible with my controller. Why single memory module does not need the manual testing?

I thought that memories need to comply the standards for compatibility. Why it is sufficient to comply the standards for the single channel but not for multichannel? How testing solves the problems of missing standard in multichannel? How do their test with their non-standard controller ensures that it will work with my controller?


> Testing improves this because they have a billion sticks laying around (yes I'm exaggerating) and can find pairs or groups of 4 that work together. Sometimes they go though large groups of sticks to find 4 that will work together in a quad channel setup.

Again, doesn't compadibility depend on the memory controller? Do you know that memory sticks are is plugged into controller rather than together? If it is so delicate, why compatibility of sticks is checked against each other instead of (my) controller?

Which characteristics may fail the compatibility? I mean, I have a standard memory that responds at CL=9 @1666 MHz. This is all what the controller needs from the memory. What else needs to be "matched" and why?

As I suspected, this is a bigotry and cult of belief. Somebody is making money on spreading beliefs. Do you hear that I tell you that memory sticks are connected to controller rather than to each other? They can only be tested by attaching them to another controller. Wheter slightest variations in parameters make a big difference, you should look at the compatibility between you memory and controller. Instead of saying that [strike]"There is no two identical sticks"[/strike], you should repeat that "There are no two identical controllers" and fine-tune your sticks to exactly my controller. Adjusting for another controller may have opposite effect: the best combination for that controller may be worst for the mine (some unnoticable variations make a huge difference, remember?). Furthermore, the sticks may be plugged into different slots on by board - this also makes a big difference. The board delays are huge and plugging multiple sticks into one channel degrades it sighificantly, from 7-7-7 to 8-8-8. The board delays also vary with temperature. Controller adjusts to them dynamically. That is, controller understands that all sticks are different and adjusts to your sticks even dynamically. Which parameters are not accounted?



Which requirements are needed besides compatibility with controller? Why do you have have up to date specs for the controller (so that you do not need to test sticks against it) but old specifications for the new technology (multichannel is relatively recent)?



If they had my CPU, so its pretty simple to have my board to test, right?



I know that unwieldy erudition is demonstrated to hide information. Just one question: where is my CPU located if not on the board? Can we ask people to decide: "controller is in CPU, CPU is on the board, is controller on the board"? Ok, another question: what makes the difference (between having controller in CPU or in a separate chip, between CPU and memory)? Do you think that packaging makes the big difference for the circuitry?




I wonder: why such huge board delays make huge difference but memories, produced according to the same specifications as the boards, are not compatible, after they are compatible with the borads. How is this possible?



You can disregard the transistors, located in the memory controller, very well. Furthermore, do you remember that memory is attached to the memory controller rather than another memory (and lines are running over the board, which "does not matter")? If interface does not matter then why isolated internals are all important? Why billions of CPU transistors do not matter, why trillions of hard disk cells don't matter? There are other pieces in my PC, not certified to work together. Why to test only against some of the things, which are isolated from each other?