A question about dual channel RAM

vbimport

#1

Hi all,

I realize that for optimum performance in dual channel you need to have matching modules, but I also read that as long as the modules have the exact same specs they can been run in dual channel. Then I read from oneperson who says that the modules don’t need to be exact, specifically they are referring to the cas latency not needing to be the same.

Can someone here set me straight on just what exactly is required to run dual channel. How close do the timings have to be? If you could point me toward some good article that would be appreciated as well.

Thanks, dc3


#2

To avoid problems, they should be identical sticks.


#3

I would recommend you got get at least the same model…
//Danne


#4

I have purchased two of the same modules for my own purposes and they work well, that’s not the reason for my question.

I looking for documentation that I can use to answer once and for all a couple of question for doubting Thomases. Do the timings have to be exact? Can cas latency be different?


#5

I have heard of people even getting sticks of different brands to run dual channel. the point is, that if you have matched sticks, you are pretty much guaranteed it will run dual channel. If they are the same but not bought as a matched pair, they have a high chance of running dual channel but there is enough variation between batches, even if the same components are used on the memory, that there is a small chance they may not run dual channel. anything beyond that, it is a gamble weather it will work or not. some motherboards are pickier that others what they will run dual channel. You can have the same exact parts and it might work in one system but not another if the memory is not the same.
As far as timings, if I understand right, the two sticks have to be able to run the same timings (even if they are rated at different timings). A faster stick might be able to run slower timings to match the other slower stick, but again, its a gamble.

I just googled this if you want something in writing to show someone.
http://www.neoseeker.com/Hardware/faqs/kb/11,16.html

according to that kingston link in that post, this is what is required.

In this context, “matching” modules means:

  1. Both modules are the same capacity (e.g. both are 256MB, or 512MB)
  2. Both modules are the same speed (e.g. both are PC2700 or PC3200)
  3. Both have the same number of chips and module sides (e.g. both have the
    same number of chips on the module, and both are either single-sided or
    double-sided).

I think that experience tells us that the above conditions are sometimes, but not always enough, particularly with some motherboards (asus had a few boards a while back that were insanely picky from what I hear).


#6

Thanks Ripit, btw…you can emachines to that list, I found reference to them being picky as well. I’ll never let one in the house, so that’s not a problem for me.lol

I will continue to search for a definitive answer to my questions, there has to be some article somewhere that breaks it down.


#7

Latency, as I understand it, is not a “hard” specification of the chip but an operating parameter. As such, it is a range of operation. If both chips operate at the setting you use you should be OK. If you set latency outside of the range of one of the chips, you will encounter problems. I have run into matched sets of dual channel memory where one chip was picky concerning latency and the other was not. Both would run within the specs but one was a little “hotter” and able to run at a lower latency.


#8

If your board and BIOS allow you to set all the timings manually, the odds increase for getting un-matched sticks to play nice. IF the board is trying to set timings automatically, you can easily run into problems. So it really boils down to the specific board and BIOS.
No, there’s no hard and fast rule or specification about this. But you won’t find any board or RAM makers that will recommend using anything other than identical sticks. If one stick is spec for 2.5 CAS, and the other is 3, for example, success will depend on the board running at 3.


#9

As I understand it, what controls timings is how fast the memory chips capacitors can discharge and recharge. That is why overclockers sometimes look at the brand and model of the actual chips on memory, not just the brand and model of the whole memory module. The module will contain information that it gives to the computer, a default set of timing for each particular operational speed setting, but like chas0039 said, its a range. Like CDan said, it depends on what your motherboard bios will let you set. Some oem and cheaper aftermarket motherboards will not let you set the timings manually. they will only run the default settings that the modules identify to it. A few oem and most beter after market boards will let you set the timings wherever you want though. If you set the timings too fast, the capacitors in the memory cannot charge and discharge as fast as you are trying to make them. This will cause data corruption, and thats reason you use memtest 86 to test for data corruption when overclocking memory.
Even if the rated/default speed of the modules are different, if you can set them to the same timings manually, they may run in dual channel mode. If they will not run the same timings stably (without data corruption), then they will not run dual channel.

Lastly, always remember, even if the modules are the same speed, chip configuration, and can run the same timings, they may or may not run dual channel, the reason they have matched sets. It larglly depends at that point on how picky your motherboard is and how it likes the ram (some boards like some rams better than other, regardless of dual channel operation).


#10

True, as far as it goes. But there’s a lot more to it. The number of physical addresses on the DIMM (how much RAM is there) is also a factor. The memory controller itself is also a factor, and believe it or not the physical distance from the controller to the memory is also a factor. Incorporating the memory controller on the CPU was a big step forward, at least for AMD, and made RAM fiddling a much more stable experience.

It’s also true that all these numbers have exponential relationships, so the higher the clock speed that the RAM is running at, the smaller the tolerance for error. Hence, whatever timing doesn’t work at 200MHz might work at 166 MHz. This is part of the reason why some boards will default to a lower clock when more RAM, or more sticks, is installed.


#11

thanks…very, very good to know! NOW…if i have a matched pair say…, pc5300 2x512. .will adding 1 more 512 stick improve,…hinder…,or do nothing? .is matching the pair then , also req.? .i don’t see much about 1/2gbs. i.e… 1.5gig… 2.5gig…ect .UNRELATED but, on a 4tr. old dell 2400…i added a 512 pc3200 to max out the 1 gig. mobo & the mb jumped down to 266MHz. .both 512 sticks , though not paired , are both rated 400MHz. . b4 i added the 2nd. stick the board read 333MHz. .still trying to wrap my head around that one.


#12

Most memory controllers crap out with more than 4 rows of memory. (2 sticks = 4 rows). This is why they downshift to a lower clock speed, shift to 2T timing, or both. The end result is always significantly slower RAM, and potential instability on some boards. Whether this is an issue to you really depends on the system and your needs. Best option is always 2 sticks in dual-channel mode. There are such things as single-sided RAM sticks that allow you to use 4 sticks and still have just 4 rows. These are very expensive and hard to find.


#13

:bow: i just know enough to get in trouble. .my boards max is 2 512 sticks. .not that old…i was surp. 1gig is max & not 2. .not unstable at all. i gotta’ go back inside anyway so i’ll ck. if sticks are single or dbl. sided. i don’t even know :o ! any thoughts on that 1/2gb thing i asked? ,thanks