{"id":22745,"date":"2014-02-10T15:38:04","date_gmt":"2014-02-10T21:38:04","guid":{"rendered":"https:\/\/eshop.macsales.com\/blog\/?p=22745"},"modified":"2014-05-22T13:21:46","modified_gmt":"2014-05-22T18:21:46","slug":"mix-and-match-more-memory-faster-mac-pro-2","status":"publish","type":"post","link":"https:\/\/eshop.macsales.com\/blog\/22745-mix-and-match-more-memory-faster-mac-pro-2\/","title":{"rendered":"Mix And Match: More Memory = Faster Mac Pro"},"content":{"rendered":"

\"MacWhen using a new Mac Pro 2013, installing four matched memory modules will get you optimal performance. This allows the system to use quad-channel addressing, which is the most effective mode of operation.<\/p>\n

That being said, in the real world \u2013 as with the 2009-2012 model Mac Pros which supported up to triple-channel addressing and even with the 2006-2008 models that REQUIRED matched pairs \u2013 more memory trumps multi-channel addressing where you have an application or applications running that make use of that memory.<\/p>\n

Here is an example of a \u201cmemory hungry\u201d situation. We took a really large image file, in this case 40,000 pixels wide and 56GB in total size, and ran some scripted operations in Adobe Photoshop.<\/p>\n

\"DigLLoyd1\"<\/a>
Image used is 40,000 pixels across, 56GB, X-axis is in seconds.<\/figcaption><\/figure>\n

In this case, more memory always gives the most significant performance boost\u00a0since there was never a time when everything fit. But each additional gigabyte got us closer.<\/p>\n

Dropping back a bit and using a 20,000-pixel wide 15.7GB image gives the following results:<\/p>\n

\"DigLLoyd2\"<\/a>
Image used is 20,000 pixels across, and 15.7GB, X-axis is in seconds.<\/figcaption><\/figure>\n

You can see where increases in total memory become less significant above 16GB.\u00a0 Above 20GB, the biggest percentage improvement comes in going from 32GB as two 16GB DIMMs, to the same capacity but as four 8GB DIMMs for a 3.5% improvement.\u00a0 That is the ideal configuration for these systems; four matched memory modules, which allows full quad-channel operation.<\/p>\n

Taking it one more step, we ran a speed test that allowed us to stay in the memory-footprint of the smallest shipping configuration of 12GB (3x4GB).<\/p>\n

\"DigLLoyd3\"<\/a>
8GB speed test X-axis is in seconds<\/figcaption><\/figure>\n

Now all of the differences are driven by the mix of module sizes, which set the operating mode for the memory controller; dual-channel, triple-channel, or quad-channel. Again, the biggest difference is between the two 32 GB configurations where two 16GB modules underperform vs. four 8GB modules by about 3.5%.<\/p>\n

So, there you have it. Where you have more memory than needed, the real-world impact of mixing and matching memory is pretty limited, topping out at a little less than 4%. You can choose the right amount of memory for what you need to do, what you plan to do, and what best fits your budget.<\/p>\n

The 16GB modules deserve a bit of extra attention, as they are what are known as \u201cRegistered DIMMs\u201d or \u201cRDIMMs\u201d for short and are incompatible with the standard 4GB and 8GB products, all of which can be freely intermixed.<\/p>\n

Say that you have 12GB or 16GB now and want to go straight to 32GB. You can go with a 4x 8GB set or a 2x 16GB set. Knowing that you can\u2019t mix 16GB modules with 8GB or 4GB, the benefit of going with 2 x 16GB is that later on you can add another 16GB stick to go to 48GB and another to go to 64GB \u2013 all without losing your prior investment. It’s a small price to pay in performance to maintain flexibility.<\/p>\n

Here\u2019s the data from our testing:<\/p>\n

\"DigLLoyd4\"<\/a>
Digiloyd Photoshop Benchmark Data<\/figcaption><\/figure>\n

So, for best results, sets of four modules should certainly be utilized. If, however, you\u2019re on a budget, adding more memory will always provide the most significant performance bang for your dollar<\/b>.<\/p>\n

Don\u2019t be restricted to Apple\u2019s limited memory upgrade options. Click here to check out our current line-up<\/a>:<\/p>\n