When I say "still", I'm not actually talking about current generation Macs. I'm specifically talking about 2012's. Is there still benefit to installing matched pairs or are those days long gone?

I have two 2012 Macs with 4 GB each. I just got my hands on 8 GB from a 2012 iMac. So the question is, which Mac do I upgrade to 8 GB?

...but I could also do 2 GB + 4 GB in each Mac, bringing each Mac to 6 GB. Way back when, it was always said to install in matched pairs, otherwise the RAM run's "slow". But is this still the case by 2012? Even if it is true, just how much slower is it? It still may be worthwhile to run unmatched pairs to get an extra 2 GB in each machine.

  • Matched pairs are ideal, but the difference is super minimal. The extra 2 GB of memory afforded by a 2G + 4G config will outdo any advantages offered by a 2G + 2G config. (A more important thing to know is that if you install two sticks of RAM which are rated for different speeds, your computer will operate all RAM at the speed of the slowest stick.) Jan 26 '19 at 5:47
  • Hey. You have permission to edit my answer significantly to add your benchmark. Also, feel free to add your answer and select it over mine. Great work documenting to others how to get an actual benchmark.
    – bmike
    Jan 26 '19 at 15:15
  • @bmike well, if I had saved the spreadsheet, I'd post the raw data. But I did not so people will just have to take my word for it, or run their own benchmarks.
    – l008com
    Jan 29 '19 at 8:34

No - even in the core 2 days, it was almost always faster to have more RAM that wasn’t matched compared to less RAM that was matched. I believe the penalty was 3% range - so you could test it in a benchmark, but real world - more RAM wins.

The only way matched ram would matter was in a perfect tie. Say two chips of 4 compared to four chips of 2. Always go with more RAM when you have capacity unless you are on a power budget - mobile devices on battery have to power all RAM and if you provision too much, you waste battery.

These days, the memory controllers are so good, I don’t think you can even benchmark differences due to a mix of memory sizes or locations.

  • 1
    Ok I downloaded XBench. I ran four memory tests with 4 GB and four memory tests with 6 GB. I put it all in a spreadsheet, did the math, and here's the result: The RAM speed was a statistically insignificant 1.4% FASTER with 6 GB (not slower). So that settles that, 6 GB of RAM all around.
    – l008com
    Jan 26 '19 at 8:44
  • Wonderful! Actual benchmarks rock and are so much better than “collective wisdom and opinion or even well reasoned speculation”. So for you the memory controller and CPU wasn’t the bottleneck, ram speeds was the bottleneck if you get faster rates with more RAM.
    – bmike
    Jan 26 '19 at 15:11

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .