RAM is expensive so trying to determine an appropriate RAM capacity for one's workflow is somewhat important to determine.
Our office is currently using an iMac Pro 10-core with 128 GB RAM. Today I captured several screen shots (40 seconds apart) of the Activity Monitor's Memory tab display to see what RAM use was. Two of the screen shots showed ~24 GB and ~27 GB. This on the face of it does not seem excessive at all.
However, RAM is used for multiple things besides Application's RAM footprint, and quite likely an important use is for cacheing data read from storage; this is shown in the Activity Monitor's "Cached Files" box. The two "Cached Files" amounts corresponding to the two screen shots was ~11 GB and ~12 GB. So within the 40 seconds elapsing between the two screen shots the cached data grew around 1 GB. If this cached file data is reused or reused frequently the responsiveness will be far superior to having to read the data from the storage devices, be they SSDs or Spinners. This aspect is another use of RAM besides what the Applications' RAM footprints are collectively.
So IMO, a baseline for the amount of RAM needed is an estimate for what the typical workflow Applications need. Having more than this amount is what I would call 'gravy' and should be as large as the pocket book can withstand.
Another point to consider is that as Apple releases newer macOSes the need for RAM by the basic system can vary, and IMO typically increases over time. For example, on my late 2016 15 inch rMBP13,3 with 16 GB RAM running Catalina 10.15.2 I typically see some 4.5 GB being used by the kernel_task whereas in earlier macOSes this kernel_task RAM use has been substantially less.
At this point I'm of the opinion my best approach for RAM capacity with a starting point of having the stock 32 GB in my configured MP7,1 12-core is to add 8 more 8 GB RAM modules to the existing 4, for a total of 12 RAM modules. This will provide a RAM size of 96 GB with all 6 memory Channels being used/available. The Apple memory documentation does state that to get max memory performance all slots should be populated with the same sized RAM modules. So having 12 x 8 GB RAM modules meets this advice.
So if the workload Applications on the MP7,1 can churn along using, say 32 GB, then 64 GB can be used for the kernel buffer cache (Cached Files) and/or more RAM needs by the kernel_task and other system processes as the future unfolds.
Adding 8 x 8 GB RAM modules is also the best way to minimize the RAM upgrade price as the original stock 32 GB can be kept and not disposed of. According to the Apple RAM population mix, it's possible to mix RAM module size, but will cause a lowering on the memory speed/performance some (the Apple doc discussing mixing RAM sizes does not seem to indicate a huge drop in performance). Thus if more than 96 GB of RAM is needed replacing some of the 8 GB RAM modules with larger sized RAM modules can done - keeping in mind not to mix the R-DIMMs and LR-DIMMs.... see https://support.apple.com/en-us/HT210103#mixed.
I found that OWC offers 8 x 8GB R-DIMMs for some $590.
Our office is currently using an iMac Pro 10-core with 128 GB RAM. Today I captured several screen shots (40 seconds apart) of the Activity Monitor's Memory tab display to see what RAM use was. Two of the screen shots showed ~24 GB and ~27 GB. This on the face of it does not seem excessive at all.
However, RAM is used for multiple things besides Application's RAM footprint, and quite likely an important use is for cacheing data read from storage; this is shown in the Activity Monitor's "Cached Files" box. The two "Cached Files" amounts corresponding to the two screen shots was ~11 GB and ~12 GB. So within the 40 seconds elapsing between the two screen shots the cached data grew around 1 GB. If this cached file data is reused or reused frequently the responsiveness will be far superior to having to read the data from the storage devices, be they SSDs or Spinners. This aspect is another use of RAM besides what the Applications' RAM footprints are collectively.
So IMO, a baseline for the amount of RAM needed is an estimate for what the typical workflow Applications need. Having more than this amount is what I would call 'gravy' and should be as large as the pocket book can withstand.
Another point to consider is that as Apple releases newer macOSes the need for RAM by the basic system can vary, and IMO typically increases over time. For example, on my late 2016 15 inch rMBP13,3 with 16 GB RAM running Catalina 10.15.2 I typically see some 4.5 GB being used by the kernel_task whereas in earlier macOSes this kernel_task RAM use has been substantially less.
At this point I'm of the opinion my best approach for RAM capacity with a starting point of having the stock 32 GB in my configured MP7,1 12-core is to add 8 more 8 GB RAM modules to the existing 4, for a total of 12 RAM modules. This will provide a RAM size of 96 GB with all 6 memory Channels being used/available. The Apple memory documentation does state that to get max memory performance all slots should be populated with the same sized RAM modules. So having 12 x 8 GB RAM modules meets this advice.
So if the workload Applications on the MP7,1 can churn along using, say 32 GB, then 64 GB can be used for the kernel buffer cache (Cached Files) and/or more RAM needs by the kernel_task and other system processes as the future unfolds.
Adding 8 x 8 GB RAM modules is also the best way to minimize the RAM upgrade price as the original stock 32 GB can be kept and not disposed of. According to the Apple RAM population mix, it's possible to mix RAM module size, but will cause a lowering on the memory speed/performance some (the Apple doc discussing mixing RAM sizes does not seem to indicate a huge drop in performance). Thus if more than 96 GB of RAM is needed replacing some of the 8 GB RAM modules with larger sized RAM modules can done - keeping in mind not to mix the R-DIMMs and LR-DIMMs.... see https://support.apple.com/en-us/HT210103#mixed.
I found that OWC offers 8 x 8GB R-DIMMs for some $590.
Last edited: