Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

camner

macrumors regular
Original poster
Jun 19, 2009
245
18
I've read a number of the threads here about memory in the 2020 27" iMac (dual vs single channel, mixing brands and/or size modules resulting in lower clock speeds, etc.).

I want to equip my new iMac with 64 GB, and was originally thinking of getting 2 x 32GB and putting the stock 8GB in the other two slots, for a total of 72GB (and giving me the ability to easily increase the RAM later, if desired), but then I read about the lower clock speeds. But if only populate two slots (2 x 32GB), then I'm running single rather than dual channel, which led me to rethink the whole thing and perhaps get 4x16GB (which is a bit cheaper, anyway).

Are these differences noticeable in the real world, or are they small enough so that it doesn't much matter how I do this?
 
  • Like
Reactions: Wesmosis
Screen Shot 2020-08-20 at 1.27.24 AM.png
with just 2 stick your not losing duel channel

just make sure its either slots 1/3 or 2/4
 
Whiteman007 is correct you will not lose dual channel with two DIMMs as long as you put them in the correct slots. I would suggest 2 X 32GB leaving the two other slots empty allowing you to upgrade later if required (don't mix with the factory apple DIMMs)
 
Whiteman007 is correct you will not lose dual channel with two DIMMs as long as you put them in the correct slots. I would suggest 2 X 32GB leaving the two other slots empty allowing you to upgrade later if required (don't mix with the factory apple DIMMs)
Why do you suggest not mixing with the factory apple DIMMS?
 
In other words it doesn't make a difference. It may produce ever so slightly lower benchmark results but will be completely unnoticeable otherwise.

Also, what made you believe that installing TWO modules will result in SINGLE-channel mode instead of DUAL-channel mode?
 
In other words it doesn't make a difference. It may produce ever so slightly lower benchmark results but will be completely unnoticeable otherwise.

Also, what made you believe that installing TWO modules will result in SINGLE-channel mode instead of DUAL-channel mode?
I prefer to know that the machine I paid several thousand dollars work at full speed as expected.
 
In that case you better disable all power management features on your CPU, because that constant clocking down might cause a slight decrease in peak short-burst performance. Also, make sure to run your fan on full blast all the time so your CPU will never miss a beat and don't forget to disable dynamic graphics switching on mobile Macs, because otherwise they might be using the slower internal GPU instead of the faster external GPU, therefore not running "at full speed as expected" :p

Kidding aside: we're talking about less than 1% in real world applications. That's hardly worth even thinking about. In fact, keeping the old RAM around might even be beneficial to real-world performance because it increases the amount of cache available to macOS.
 
I would rather have 64GB at full speed than 72GB at a lower speed.

On topic though: get 2x32GB, unless 4x16GB is a lot cheaper and you're sure you'll never need more than 64GB.
 
In that case you better disable all power management features on your CPU, because that constant clocking down might cause a slight decrease in peak short-burst performance. Also, make sure to run your fan on full blast all the time so your CPU will never miss a beat and don't forget to disable dynamic graphics switching on mobile Macs, because otherwise they might be using the slower internal GPU instead of the faster external GPU, therefore not running "at full speed as expected" :p

Kidding aside: we're talking about less than 1% in real world applications. That's hardly worth even thinking about. In fact, keeping the old RAM around might even be beneficial to real-world performance because it increases the amount of cache available to macOS.
 
What's that supposed to tell me? This is a serious question, as I don't get it. Please enlighten me.

Here's what I gathered from this post: He ran two benchmarks. Both with RAM running at 2667 MHz, one in single-channel mode (16+16+4+4) resulting in 2,616 points and one in dual-channel mode (16+0+16+0) resulting in 2,655 points. In other words: the overall performance difference between dual-channel RAM at 2667 MHz and single-channel RAM at 2667 MHz is around 1.5%.

What has not been tested is dual-channel mode running at 2133 MHz (16+4+16+4), which is most likely going to sit right in-between those two benchmark scores.
 
  • Like
Reactions: Wesmosis and iMi
What's that supposed to tell me? This is a serious question, as I don't get it. Please enlighten me.

Here's what I gathered from this post: He ran two benchmarks. Both with RAM running at 2667 MHz, one in single-channel mode (16+16+4+4) resulting in 2,616 points and one in dual-channel mode (16+0+16+0) resulting in 2,655 points. In other words: the overall performance difference between dual-channel RAM at 2667 MHz and single-channel RAM at 2667 MHz is around 1.5%.

What has not been tested is dual-channel mode running at 2133 MHz (16+4+16+4), which is most likely going to sit right in-between those two benchmark scores.
RAM transfert speed takes a serious hit. For memory intensive applications, this is crucial. We don't increase RAM speed every generation for fun. There's a reason, and wanting to keep it as high as possible is meaningful in some use cases, especially when you pay a machine priced that high.
 
  • Like
Reactions: Wesmosis
Yes, agreed. In some cases it can make a difference. Like serious number crunching. Ballistic calculations. Statistical analysis. Protein folding. Genome decoding. Particle simulation. Weather pattern analysis. Climate prediction calculations. None of which are typical use cases for an iMac. Everything else will take a 1% hit.

Either way, I'm out of here. I've said all that needs to be said from my perspective and I don't see the point going down this rabbit hole. If you honestly think that there is a noticeable and noteworthy difference in overall everyday performance between running two sticks at 2667 versus running four sticks at 2133 MHz then so be it.
 
  • Like
Reactions: Wesmosis and iMi
Yes, agreed. In some cases it can make a difference. Like serious number crunching. Ballistic calculations. Statistical analysis. Protein folding. Genome decoding. Particle simulation. Weather pattern analysis. Climate prediction calculations. None of which are typical use cases for an iMac.
LOL.

In bold are these I do or a geneticist friend of mine do.

So yes. Some people use their iMac for these particular use cases.
 
  • Like
Reactions: mj_
Lol, you are funny.

You think this forum is full of noobs? You would be surprised how many talented, gifted and serious pro people you would find here. Filter out the trolls and fanboys and you have a lot of people who do serious stuff.
And iMac fills a certain gap now that its as fast or even faster than iMac Pro.

Your 1% is irrelevant and actually statistically incorrect. So, please don't bother responding, so far you just stirred the crap here for zero reasons.

Either contribute or not. The memory "issue" is new with this gen so its only fair that it gets discussed.




Yes, agreed. In some cases it can make a difference. Like serious number crunching. Ballistic calculations. Statistical analysis. Protein folding. Genome decoding. Particle simulation. Weather pattern analysis. Climate prediction calculations. None of which are typical use cases for an iMac. Everything else will take a 1% hit.

Either way, I'm out of here. I've said all that needs to be said from my perspective and I don't see the point going down this rabbit hole. If you honestly think that there is a noticeable and noteworthy difference in overall everyday performance between running two sticks at 2667 versus running four sticks at 2133 MHz then so be it.
 
Most people won’t notice the difference between 2667 and 2133, but most people also won’t notice the difference between 64GB and 72GB RAM...
 
  • Like
Reactions: Wesmosis
In other words it doesn't make a difference. It may produce ever so slightly lower benchmark results but will be completely unnoticeable otherwise.

It also has real world benefits for anything that is dependant on memory bandwidth and that is not just benchmarks! In fact a lot of these benchmarks are not even testing memory performance. Sure you're not going to notice it at all if you're browsing the web but perhaps one application that many users would take advantage of is gaming - Here is an illustration of how memory can affect game performance.

P4eeDtW.png
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.