I had read somewhere on here that some people are adding 3rd party RAM to their new systems but taking out the stock 8GB of Apple RAM as it is faster than mixing it. Is that actually true and, if so, is it true with respect to all 3rd party RAM? I have a new system coming and was planning to purchase 32GB of Crucial or OWC RAM.
Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
Mixing RAM in 2020 iMac
- Thread starter jazzer15
- Start date
- Sort by reaction score
There are several threads discussing this, but the short answer is yes, in your case you are best to put in the 32GB RAM and toss the 8GB Apple RAM (or sell it, you might get $20 for the two Apple sticks). It does not appear to make any difference which 3rd party RAM.
The answer is a bit different for different cases. For instance if you are adding only 16GB, it might be better to keep the 8GB, despite it running at a lower clock speed. Or if all four sticks are the same size, definitely keep all four sticks, as they can be arranged to keep full speed.
The answer is a bit different for different cases. For instance if you are adding only 16GB, it might be better to keep the 8GB, despite it running at a lower clock speed. Or if all four sticks are the same size, definitely keep all four sticks, as they can be arranged to keep full speed.
Thanks for the response. For some reason when I searched the other threads didn't come up, but I found them now. If I install the extra RAM and it shows as running at the proper clock speed, I assume I am safe in assuming that everything is fine?
The answer is summarized on page 11 of this thread...
https://forums.macrumors.com/thread...hooting-in-2020-27-inch-imacs.2249254/page-11
FWIW Yesterday I added a matched pair of 16GB sticks of Crucial RAM to a friend’s 2020 iMac bringing the total to 40GB BUT the RAM speed dropped from 2667MHz to 2133MHz
We decided that 40 GB of RAM at 2133MHz was preferable to 32GB at 2667MHz -- IOW Space over Speed.
GetRealBro
https://forums.macrumors.com/thread...hooting-in-2020-27-inch-imacs.2249254/page-11
FWIW Yesterday I added a matched pair of 16GB sticks of Crucial RAM to a friend’s 2020 iMac bringing the total to 40GB BUT the RAM speed dropped from 2667MHz to 2133MHz
We decided that 40 GB of RAM at 2133MHz was preferable to 32GB at 2667MHz -- IOW Space over Speed.
GetRealBro
No that is not safe to assume.
If you add the 32GB RAM to the 8GB Apple RAM, you will likely get one of two results, depending on the order you arrange the sticks:
Sticks in order: A-B-A-B: 2133 MHz, running in dual channel
Sticks in order: A-A-B-B: 2667 MHz, running in single channel - this is even worse!
If you only put in the 32GB:
Sticks in order: empty-B-empty-B: 2667 MHz, running in dual channel
There is always a chance you will get a result different from everyone else, but it is unlikely.
The best way to check is to run Geekbench CPU score and Novabench RAM speed test for each arrangement, and see which gives the best result.
btw, do not freak out when swapping sticks around in your machine: it takes a LONG time to boot after a stick swap, as it runs extra POST tests
In the interests of "science" I tried a real world (for me) example for 32GB third party RAM, vs 40GB RAM (32GB third party + 8GB Apple RAM), which was importing and rendering 1:1 previews in Lightroom for 44 photos, and also ran the benchmarks again for each. This is what I got (for an i7 CPU and 5500XT GPU):
32GB at 2667MHz:
Lightroom import and render: 63 seconds
Geekbench CPU: 1262 single, 8940 multi
Novabench RAM speed: 29080 MB/s
40GB at 2133MHz:
Lightroom import and render: 65 seconds
Geekbench CPU: 1256 single, 8460 multi
Novabench RAM speed: 25900 MB/s
I was careful to delete the Lightroom previews file before each import.
You can make your own conclusions and decisions. (It is interesting to me the Lightroom import was only slightly slower with 40GB @ 2133MHz.)
Hope this helps
32GB at 2667MHz:
Lightroom import and render: 63 seconds
Geekbench CPU: 1262 single, 8940 multi
Novabench RAM speed: 29080 MB/s
40GB at 2133MHz:
Lightroom import and render: 65 seconds
Geekbench CPU: 1256 single, 8460 multi
Novabench RAM speed: 25900 MB/s
I was careful to delete the Lightroom previews file before each import.
You can make your own conclusions and decisions. (It is interesting to me the Lightroom import was only slightly slower with 40GB @ 2133MHz.)
Hope this helps
Last edited:
Thank you. Very helpful. That's the same machine that I have coming and the same amount of RAM I plan to run.
Just to be clear, if I only have 32GB (2 x 16) and don't use the original 8 (which it seems like I should do), do the sticks go in slots 1 and 2 (both in channel A) or 1 and 3 (which the summary in the thread getrealbro cited me to states). I have a friend that has his set up like the attached (which seems to be what you suggested above by stating "empty B ... empty B), but the other thread (which you also commented on) says that with 2 sticks putting them in either the odd or even slots was better. Obviously I could test both when the machine gets here to see which gets better results.
Sorry to be a bit dense.
Last edited:
Just a thought... but the one factor that no one seems to be considering is the fact that the Apple RAM is the only RAM that is creating havoc when mixed with other non-factory installed Apple RAM. It could be that there is a very slight difference in their RAM that isn't found in the other RAM which is making it incompatible. Isn't the Apple RAM supposed to be eco-friendly? Perhaps it draws a slightly different amount of power or has some other internal coding that makes it behave this way.
Pulling the RAM and replacing it or using additional RAM that was also found in the new line seems to be the only thing that works with it, which seems to support the theory that the RAM isn't really identical (third-party) even though by specification it appears to be the same.
It would be interesting to see if Apple would respond to this issue as clearly third-party RAM doesn't play well with the Apple RAM.. which suggests either the RAM is different than previous models or something is telling the machine to behave this way. As in, it happens on all the new iMacs just released, which suggests something has changed since the last release (2019).
Pulling the RAM and replacing it or using additional RAM that was also found in the new line seems to be the only thing that works with it, which seems to support the theory that the RAM isn't really identical (third-party) even though by specification it appears to be the same.
It would be interesting to see if Apple would respond to this issue as clearly third-party RAM doesn't play well with the Apple RAM.. which suggests either the RAM is different than previous models or something is telling the machine to behave this way. As in, it happens on all the new iMacs just released, which suggests something has changed since the last release (2019).
That screenshot is the wrong way to do it, and loses dual channel.Thank you. Very helpful. That's the same machine that I have coming and the same amount of RAM I plan to run.
Just to be clear, if I only have 32GB (2 x 16) and don't use the original 8 (which it seems like I should do), do the sticks go in slots 1 and 2 (both in channel A) or 1 and 3 (which the summary in the thread getrealbro cited me to states). I have a friend that has his set up like the attached (which seems to be what you suggested above by stating "empty B ... empty B), but the other thread (which you also commented on) says that with 2 sticks putting them in either the odd or even slots was better. Obviously I could test both when the machine gets here to see which gets better results.
Sorry to be a bit dense.
The machine comes with the Apple sticks in slots 2 and 4. You should take these out and put your sticks in the same slots from which you removed the Apple sticks. Like this:
It should work also to put them in Slots 1 and 3, but when I tried that it wouldn't boot (maybe for a different reason). The safest is to put them in the slots that Apple did.
OK. That's what I thought. I'll do that when mine comes. I'm not sure why he did that. I just assumed the original RAM came in those slots and he just replaced them, but obviously that's not the case. I'll suggest that he change the configuration.
Thanks again.
Looking at your Memory Slots screen dump reminded me that when I added 2 16GB sticks of Crucial RAM they had a Manufacture code of 859B while the Manufacture code for the 2 4GB sticks of “Apple” RAM was Micron.
I’m wondering if this memory configuration speed glitch can/will be fixed with a firmware update?
GetRealBro
It almost certainly could. It could be set to accept and run faster RAM too since Windows recognizes the faster RAM (at least according to TallyHo on YouTube). I’m generally doubtful that Apple would issue a firmware update to fix the RAM stuff — it doesn’t seem to be in their best interest. That said, since Apple ships machines with different RAM manufacturers (some are SKhynix and some are Micron), it’s possible that this may create problems with the RAM modules Apple sells directly (at insane prices). As a thought experiment, if I decided to spend $800 on 32GB of RAM at my local Apple Store (2 x 16GB sticks) and then it turned out to be a different RAM manufacturer than what my iMac came with (let’s say I had Micron and the store modules are SKhynix), that’s a problem when the ensuing installation either leaves me with RAM running slower or not in dual channel.
Now, sadly, I no longer actually trust the Genius Bar to employ people who a) know what dual-channel memory is, b) know how to check for dual-channel, or c) understand general RAM speeds — but assuming you did manage to find someone who understood those things, sold the RAM from the store, and did the upgrade in store, that could create some frustrations if it meant the Genius had to source other RAM to either replace the initial RAM with (so they’d match) or to have to cold-call other stores to see who made what modules are on the shelf. Do I actually think that hypothetical would be enough to encourage a firmware update for the logic board to be less picky about RAM? Honestly, no. But it’s the only thing I could see that might encourage some sort of fix.
Alas, I think we're just going to need to remind our friends and loved ones that the 2020 iMac is sensitive and that you need matching brands. Oh, and not to follow the terrible advice from 9to5Mac on disabling dual-channel just so clock speeds check out.
I find it fascinating, that people are essentially breaking dual channel support in order to obtain the 2166Mhz speed all over the web and touting it as a success. No one seems to care about dual channel? Apple themselves don't even have a single document online that mentions how to configure RAM modules in the 4 slots, just plop any old module in any old slot apparently so long as it meets the RAM specs provided.
I would venture to say that the majority of Macs are not running in dual channel mode even though it supports it because of the lack of documentation about it and the lack of compatibility with it. I will have to admit, in the past, all I ever looked at was that it said x amount of RAM installed on the system information page... never once looked at the clock speeds or even if it was in dual channel. My machine shipped from Apple with 3 out of slots installed, slots 1-3... this is on an older Mac Pro 5,1.
Such a conundrum.
Some interesting testing over on Bare Feats: Link
Pretty much what I expected given the discussion here. The only unexpected thing is how slow the 2x4GB DIMMs are, even when configured correctly.
Pretty much what I expected given the discussion here. The only unexpected thing is how slow the 2x4GB DIMMs are, even when configured correctly.
The fact that 32 x 4 was slower than 32 x 2 (dual channel) or 16 x 4 surprised me too but this obviously shows what we’ve all bee saying: match brands and ideally match size of the module too
interesting I just ran the 32x4 tests and 32x2 tests here:
forums.macrumors.com
and geekbench is reporting conflicting results with BareFeats. Though I can't seem to figure out how to replicate their benchmark tests. (never mind their benchmark tool cost $$)
iMac 2020 Reviews And Benchmarks - The Last Intel iMac
updated the i7 8 core w/ 128gb of ram: Single core: 1381 multi core: 8777
forums.macrumors.com
and geekbench is reporting conflicting results with BareFeats. Though I can't seem to figure out how to replicate their benchmark tests. (never mind their benchmark tool cost $$)
The Mac Pro 5,1 supported triple-channel memory and you were encouraged to install in groups of three so that would be the reason for that setup.
I don’t know if apple’s public documentation mentions dual channel anymore but I seem to recall that it did at one time. The repair manuals for Genius’s/authorized repair people certainly do.
I think more people probably run dual channel than you think because it’s configured that way out of the box and most users don’t upgrade — and those that do usually buy RAM kits, meaning they’d ideally install another pair or set of pairs. I’m sure some folks mix stuff the wrong way but for Mac users, it’s probably the minority, since Apple installs RAM the right way. But for sure plenty of people don’t know and misconfigure stuff, and that’s why I’m angry that sites that make a business out of service journalism and content for regular users gets it wrong.
I was a technology journalist for a decade and it’s just frustrating to see the wrong info go out there, especially when I know the impact news sites (and now YouTube channels) have on regular users.
In…
https://forums.macrumors.com/thread...s.2249254/page-11?post=28833756#post-28833756
I suggested that “for most people who buy a 2020 iMac, the original 8GB of Apple RAM is only really useful for testing & configuring their new iMac prior to upgrading the RAM.” The Barefeats benchmarks are more punishing. The best use of the original 8GB of Apple RAM is in a work drawer, just in case you decide to return/repair the 2020 iMac
GetRealBro
https://forums.macrumors.com/thread...s.2249254/page-11?post=28833756#post-28833756
I suggested that “for most people who buy a 2020 iMac, the original 8GB of Apple RAM is only really useful for testing & configuring their new iMac prior to upgrading the RAM.” The Barefeats benchmarks are more punishing. The best use of the original 8GB of Apple RAM is in a work drawer, just in case you decide to return/repair the 2020 iMac
GetRealBro
Apple will never refuse to repair an iMac with certified third party RAM with the part numbers listed here https://forums.macrumors.com/threads/memory-upgrade-compatibility-list-for-imac-2019-2020.2248911/
I am still unclear about mixing ram. I know that mixing two sizes will either drop speed or lose dual channel. At present, there is no way to mix, for example 2x4 and 2x16 and retain speed and dual channel. However, those who have two different ram manufacturers in ram that is identical in speed, capacity, etc., (such as 2x16 and 2x16) still have speed drop to 2133 if it is in ABAB. Putting them in AABB increases speed to 2666. In this case is dual channel dropped? Is there anyway to determine this for certain.
Dual channel is dropped. I know it. Republicans know it. Democrats know it. Everybody knows it.
More seriously, yes, it's dropped. As a former computer hardware tech, it's dropped. It needs to be ABAB.
I understand. But if at all possible when using Apple for repairs…I prefer to wipe the internal drive, install a fresh copy of the most current OS and put the original RAM back in how it came. This way they have no excuses
GetRealBro
Just wait until you have to optimise the memory configuration for servers with Intel Scalable Xeon (2 memory controllers with 3 memory channels each, and 2 DIMM slots per channel per CPU) or AMD Epic (8 memory controllers with 1 channel each and 2 DIMM slots per channel per CPU) with up to 8 CPUs per system. It make i7 and i9s with 1 dual channel memory controller seem pretty straight forward.
Just wait until you have to optimise the memory configuration for servers with Intel Scalable Xeon (2 memory controllers with 3 memory channels each, and 2 DIMM slots per channel per CPU) or AMD Epic (8 memory controllers with 1 channel each and 2 DIMM slots per channel per CPU) with up to 8 CPUs per system. It make i7 and i9s with 1 dual channel memory controller seem pretty straight forward.
Indeed! And don’t even get me started on memory clearance and heatsink sizes...