Memory upgrade troubleshooting in 2020 27-inch iMacs

[Juzo Fuwa]

Put in 32GB (I suggest 2x16GB sticks) and toss Apple's 8GB.
There are different opinions on RAM brand: I suggest Crucial if available, otherwise OWC or Timetec are both fine (IMO). Just make sure they say they are "made for Mac," as they then have lifetime guarantee to be compatible. Crucial make sticks some of which are for Mac some not. I don't know if there is really any difference between them, but better to get the ones they recommend for Mac.

Thanks Wilberforce ????.

 

[wilberforce]

ups... my mistake

just edited to :

CPU single core- 1285
CPU multi core- 6957

OK, multicore 6957 is poor for the i7. If that was with the 4-4-32-32 RAM it is understandable. It should be about 8900 for empty-32-empty-32 RAM configuration

 

[getrealbro]

Put in 32GB (I suggest 2x16GB sticks) and toss Apple's 8GB....

When adding 32GB of RAM to a an 8GB 2020 iMac…

I’m guessing that there are benchmarks that show that 40GB of RAM running at 2333MHz is less desirable than 32GB of RAM running at 2667 MHz? But do we have any real life examples that confirm that putting the 8GB of Apple RAM in the drawer is preferable to just leaving it there, yielding 40GB of RAM running at 2133 MHz?

GetRealBro

Edit: fixed 40GB RAM speed

 

[jasnw]

Has anyone seen a good discussion of why the 2020 iMac is so finicky?

 

[wilberforce]

When adding 32GB of RAM to a an 8GB 2020 iMac…

I’m guessing that there are benchmarks that show that 40GB of RAM running at 2333MHz is less desirable than 32GB of RAM running at 2667 MHz? But do we have any real life examples that confirm that putting the 8GB of Apple RAM in the drawer is preferable to just leaving it there, yielding 40GB of RAM running at 2333 MHz?

GetRealBro

In normal use both options will still work fine, and subjectively will probably seem identical. It is only when doing a long processor-intensive task, like rendering video, would a difference be noticed. I would suggest for those that really need more than 32GB, just get 64GB.
btw, it drops to 2133MHz, not 2333Mhz.

 

[wilberforce]

Has anyone seen a good discussion of why the 2020 iMac is so finicky?

My personal guess is that it is a POST bug.
Like it makes sense to check DIMMs in slots 1 and 3 are identical, and slots 2 and 4 are identical, but it is incorrectly coded to check that slots 1 and 2 are identical, and slots 3 and 4 are identical.
But I really have no idea.
I would be interested in any theories

 

[PBG4 Dude]

My personal guess is that it is a POST bug.
Like it makes sense to check DIMMs in slots 1 and 3 are identical, and slots 2 and 4 are identical, but it is incorrectly coded to check that slots 1 and 2 are identical, and slots 3 and 4 are identical.
But I really have no idea.
I would be interested in any theories

The weirdest thing was Apple had to upgrade their RAM page. Seems like they were caught off guard too with this change.

 

[TiffMy]

My personal guess is that it is a POST bug.
Like it makes sense to check DIMMs in slots 1 and 3 are identical, and slots 2 and 4 are identical, but it is incorrectly coded to check that slots 1 and 2 are identical, and slots 3 and 4 are identical.
But I really have no idea.
I would be interested in any theories

Don’t know if Apple could fix it by software update...

 

[wilberforce]

When adding 32GB of RAM to a an 8GB 2020 iMac…

I’m guessing that there are benchmarks that show that 40GB of RAM running at 2333MHz is less desirable than 32GB of RAM running at 2667 MHz? But do we have any real life examples that confirm that putting the 8GB of Apple RAM in the drawer is preferable to just leaving it there, yielding 40GB of RAM running at 2133 MHz?

GetRealBro

Edit: fixed 40GB RAM speed

So here perhaps is a "real life" comparison from MaxTech:

 

[TiffMy]

So here perhaps is a "real life" comparison from MaxTech:
View attachment 953075

The job requires not much memory that greater than 32GB, so the speed of RAM makes the difference between the 32GB and 40GB@2133MHz.
However I would like to see any difference between 40GB on different speed, on works that require less than 16GB of RAM.

 

[mikehalloran]

…
However I would like to see any difference between 40GB on different speed, on works that require less than 16GB of RAM.

You will find comparison tests in the first few pages of this thread.

Real world tests that can be repeated showing an actual difference? Yea… Good luck with that.

Who are the AV pros running Lightroom Classic on a consumer grade iMac? For that matter, who are the hobbyists doing the same?

 

[wilberforce]

The job requires not much memory that greater than 32GB, so the speed of RAM makes the difference between the 32GB and 40GB@2133MHz.
However I would like to see any difference between 40GB on different speed, on works that require less than 16GB of RAM.

Well, anybody can run their own personal "real life" workflow test, by trying it with, versus without, the 8GB of Apple RAM included.
I am just relaying found information to aid in making an informed decision, in case it might be helpful.

 

[mikehalloran]

Well, anybody can run their own personal "real life" workflow test, by trying it with, versus without, the 8GB of Apple RAM included.

Really?

How do you propose doing that and getting a precise measurement of the results?

 

[TiffMy]

Well, anybody can run their own personal "real life" workflow test, by trying it with, versus without, the 8GB of Apple RAM included.
I am just relaying found information to aid in making an informed decision, in case it might be helpful.

Well yes. But I am not interested in doing experiments anymore. I am afraid the RAM door will break someday, I will not open it unless upgrading to 32GB RAM or having any kind of RAM related errors.

 

[wilberforce]

Well yes. But I am not interested in doing experiments anymore. I am afraid the RAM door will break someday, I will not open it unless upgrading to 32GB RAM or having any kind of RAM related errors.

I agree, me too. I got concerned about shoving RAM in and out of the slots over and over again trying different arrangements. I don't know how many times the slots can take that, but I don't want to find out. So I am done with trying different arrangements.

 

[Bechi]

I'm doing what @gtosnipey was doing: not using Apple RAM, but mixing two types of third party RAM, which is why I told him to re-order them to maximise results.

Sorry to get your hopes up if you thought I was mixing Apple and third party.

It still shows that each stick does not have to be identical.

Evidence attached.

I've been going through this thread again in preparation of the arrival of my new iMac and Crucial DIMMs. So far, (and I haven't reached the end of this thread yet), it seems that issues are associated with the iMac20,2 build and Geekbench is showing your build as iMac20,1. Could that have something to do with your ability to maintain dual-channel? If so, sucks for us with the later builds...

 

[getrealbro]

So here perhaps is a "real life" comparison from MaxTech:
View attachment 953075

Thanks. So it looks like leaving the 8GB of Apple RAM in, could slow a 2020 iMac by roughly 20% on a long job.

I think I'll recommend that my photographer friend stick with 40GB at 2133MHz and hope this gets cleared up with a firmware update. If a firmware fix doesn't happen, it will be easy enough to yank the 8GB later. FWIW my friend's new 27" iMac i7 w/ 5500XT is replacing a 27" 2011 iMac. So I'm guessing my friend won't notice the 20% slowdown right away

GetRealBro

 

[Juzo Fuwa]

Thanks. So it looks like leaving the 8GB of Apple RAM in, could slow a 2020 iMac by roughly 20% on a long job.

I think I'll recommend that my photographer friend stick with 40GB at 2133MHz and hope this gets cleared up with a firmware update. If a firmware fix doesn't happen, it will be easy enough to yank the 8GB later. FWIW my friend's new 27" iMac i7 w/ 5500XT is replacing a 27" 2011 iMac. So I'm guessing my friend won't notice the 20% slowdown right away

GetRealBro

Lol, that’s the same year my new iMac will be replacing ??.

 

[AZREOSpecialist]

So far all that I have seen post who have installed 4 of the same size and spec RAM sticks by putting the matched pairs in slots 1 and 2 and the other matched pair in 3 and 4 have kept full speed and dual channel.

So if the person above puts the 2x 4 GB sticks of Apple RAM in slots 1 and 2 and the 2x 4 GB sticks of third party RAM should maintain 2667 MHz and dual channel.

I am all ears if you want to find a person who did the config I mention and see it drop in speed or lose dual channel.

Please stop repeating and spreading misinformation unless you are willing to try it yourself and report back. I have a new 2020 iMac and I can confirm that no combination of Apple and 3rd party RAM will work together at 2667 MHz and dual channel. You can make them work at 2667 MHz single channel, but not dual.

 

[quagmire]

Please stop repeating and spreading misinformation unless you are willing to try it yourself and report back. I have a new 2020 iMac and I can confirm that no combination of Apple and 3rd party RAM will work together at 2667 MHz and dual channel. You can make them work at 2667 MHz single channel, but not dual.

So you have tried all 4 sticks being the same size but one being the stock Apple RAM and the other third party?

If so and it is running in single channel, what did you use to confirm it was running in single channel?

I can’t try it myself because my Apple RAM is the stock 8 GB and I picked up 64 GB. So my tests would be mismatched sticks which I know won’t work.

You would be the first person to have 4 sticks of the same size not able to get full speed and dual channel. Everyone else I have observed in this thread has been able to get 4 sticks of the same size to run full speed and dual channel.

 

[AZREOSpecialist]

So here perhaps is a "real life" comparison from MaxTech:
View attachment 953075

There is something seriously wrong with this chart. How can mixed 40 GB at 2667 MHz be over 40% slower than the same mixed 40 GB at 2133 MHz? I don't think this chart is correct.

 

[wilberforce]

There is something seriously wrong with this chart. How can mixed 40 GB at 2667 MHz be over 40% slower than the same mixed 40 GB at 2133 MHz? I don't think this chart is correct.

It is correct, because that arrangement loses dual channel (which is much worse than lower frequency)
See 7:02 in the video for more explanation

 

[wilberforce]

Please stop repeating and spreading misinformation unless you are willing to try it yourself and report back. I have a new 2020 iMac and I can confirm that no combination of Apple and 3rd party RAM will work together at 2667 MHz and dual channel. You can make them work at 2667 MHz single channel, but not dual.

anthony13 confirmed that you can combine Apple and 3rd party RAM and get both 2667MHz and dual channel, provided all the sticks are the same size, by putting them in order A-A-B-B. See post #257 and prior posts.

1. The prerequisite for 2667MHz is that sticks in slots 1 & 2 must be identical, and sticks in slots 3 and 4 must be identical.
2. The prerequisite for dual channel is that sticks in slots 1 and 3 must be the same size, and sticks in slots 2 and 4 must be the same size.

The only combo of 4 sticks of two different brands (A and B) that satisfies these two prerequisites simultaneously is the order A-A-B-B, where all sticks are the same size.
I also have tested and confirmed this.

 

[vinchensoo]

Hi guys!
I just received my new imac 2020 with 8 gb ram.
I checked apple ram and found they use this model (in my case) https://www.skhynix.com/products.view.do?vseq=3321&cseq=75 Hynix HMA851S6DJR6N
I compared it with default ram from imac 2019 and found, that this memory has ECC and ODT support.
If I'm not mistaked, Apple recommends use memory without ECC for Imac2019.

Can this cause the issue with third-party RAMs? I'd like to buy 32 additional gb too

 

[wilberforce]

I've been going through this thread again in preparation of the arrival of my new iMac and Crucial DIMMs. So far, (and I haven't reached the end of this thread yet), it seems that issues are associated with the iMac20,2 build and Geekbench is showing your build as iMac20,1. Could that have something to do with your ability to maintain dual-channel? If so, sucks for us with the later builds...

I believe the iMac 20,2 are those with the 5700XT GPU (or 5700 GPU?). Don't know why, maybe the logic board is different to accommodate the different GPU.
It's not as if 20,2 is a later version of 20,1 as they both came out the same time.
I may be wrong, which would be proven if someone has a 5700XT GPU and their model identifier is not 20,2.