Does using a matched pair of memory sticks provide any benefit?

Question

Apple recommends installing pairs of memory sticks with matched sizes to allow for dual-channel memory interface.

However, some people believe that it's also important not only to match the sizes of the two RAM sticks, but to install what manufacturers sell as "matched pair kits". Others disagree.

http://forums.macrumors.com/showthread.php?t=392843:

The tolerances of two single memory modules can be different. With a matched pair the tolerances are supposed to be close to the same if not the same.

I don't think their is a difference, at least in the real world. Two modules of the same part number should work just fine. Matched pairs seems to be little more than marketing and paying a 'tax' for intangible 'peace of mind'

You can get a matched pair by buying 2 singles of the same ram too..

http://www.dslreports.com/faq/9231:

Two modules of the same model/brand purchased from the same vendor at the same time is essentially as likely to work properly in a dual channel configuration as is a dual channel kit.

http://wiki.answers.com/Q/What_are_the_advantages_to_buying_DDR_RAM_in_packaged_pairs?classic=true:

Typically the chips are the same production batch and that increases the likelihood that the paired modules will work approximately at the same speed, therefore being detected (and usable) as dual-channel.

To summarize:

• Some people believe that buying two individual sticks of RAM from the same manufacturer with the same specifications is no different than buying a paired kit.
• Some people believe that buying a paired/matched kit provides some performance or reliability benefit as compared to buying two individual sticks of RAM from the same manufacturer with the same specifications.

Which is true?

Answer 1

I am going to assume that the memory in question has a manufacturer's warranty, as all memory I have ever seen does, and that the memory is warranted to perform within the specifications given by the manufacturer, which I think is reasonable.

So, to answer this question, we need to look a bit closer at those specifications.

Aside from the quantity of memory, and the memory type (such as DDR2, DDR3, etc.), RAM is sold with varying speeds and varying timings. These timings specify to the computer's memory controller how quickly the RAM is able to interact with the system in various ways. For the speed, which is expressed in MHz, a higher number is faster, and for the timings, which are expressed in clock cycles, lower numbers are faster.

Each stick of memory you buy will have an SPD chip in addition to the memory. The SPD contains the speed and timings that the manufacturer has rated the RAM to run at. It may contain more than one profile, a set of speed/timings, for use in enthusiast or high performance systems. The computer's BIOS uses these to automatically configure the system to use the RAM, though many systems allow the user to override these values, e.g. to overclock.

Where things get interesting is what happens when you put memory in a system, where each stick of RAM has different speed or timings. In this case, the system will run all of the RAM at the speed and timings of the slowest stick.

This is probably what is at issue with the claim that matched pairs may have performance or reliability benefits over otherwise identical single sticks. Matched pairs of sticks are guaranteed by the manufacturer to have identical speed and timings. But pairs purchased separately may not have identical speed and timings, especially if the buyer is not careful when making the purchase. By way of example, consider these two 8GB RAM sticks:

• CMZ8GX3M1A1600C10
• CMZ8GX3M1A1600C9

At the time of writing, these are the same price, run at the same speed (1600) and look virtually identical except for the last bit of the model number. But one has slightly slower timings (10-10-10-27) than the other (9-9-9-24). It's easy to make a mistake and buy the "wrong" one if a person buys them separately. If one used one of each of these sticks, the computer would run both sticks at the slower timings.

But if one does purchase two of the same stick, it's perfectly reasonable to expect them to perform exactly the same as if they were sold in a pair. There's no credible evidence that I can find to suggest otherwise.

Where this expectation breaks down is with overclocking. As noted earlier, manufacturers only warrant their products to run at their rated speeds/timings, and if someone chooses to run the product outside those specifications, all bets are off.

The only real advantage I was able to find to buying RAM in matched pairs was that it can be slightly cheaper than buying the same sticks separately. For example, at the time of writing, a matched pair of the 1600C9 sticks shown above is $15 cheaper than two of the single sticks, and a matched pair of the 1600C10 sticks is $10 cheaper.