The RAM banks are on package not on die. The size of the RAM is put deliberately on the package. No binning here.
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.
I am really torn between 8GB vs. 16GB RAM
- Thread starter Booji
- Start date
- Sort by reaction score
The Apple Store has a nice progress bar on local shipments. It gives you the delivery window and tells you the status of the order:
The Apple Store progress bar looks just like the progress bar from Dominos Pizza. I just leave a browser window up with the progress bar (one reason why three monitors is nice).
- Order in Progress
- Out for Delivery
- Arriving Soon
- Delivers
The Apple Store progress bar looks just like the progress bar from Dominos Pizza. I just leave a browser window up with the progress bar (one reason why three monitors is nice).
Although all of the M1s are on proven platforms, new designs might have 1st generation teething troubles, like flexgate.
Memory management is an inherently asynchronous process, especially when you have multiple cores running together side-by-side. Apple uses a "lazy" memory management technique: if you have 8Gb real memory and 48Tb virtual space for each task, why confine processes to strict use of the bottom end when you can easily walk across all of it leaving discarded fragments behind to be swept up at relative leisure?
Allocation is somewhat complicated, but retain allocates nothing, and release-to-0 probably just queues the object for deallocation whenever. I would not be surprised to learn that one of the Little cores get frequently called into service as a dedicated memory manager when loads go up or even most of the time so that the working cores can go in about their business.
retain and release are method calls, which is the most important point: method calls impose an overhead that is greater that mere function calls, so if those two are faster, it seems likely that all Cocoa/Swift method calls have been optimized in the Axx/Mx processors, which is a very good thing.
Yes but that queue is faster on Apple Silicon. Therefore, we can get to memory being cleared faster. If retain/release is faster on Apple Silicon, so is memory being cleared BECAUSE it can release faster it is able to clear out the memory faster.
And yes, it is essentially a method call which is why I asked if it was asynchronous since if you call MethodA then MethodB, you need to wait for MethodA to finish before MethodB gets called in synchronous code. So if it is NOT asynchronous, MethodB and similar code would take long on Intel vs Apple Silicon to be called in some scenarios.
I am still not seeing where you are saying I am wrong that you said memory transactions are not faster. If releases are faster, it gets to the queue to delete memory faster, which does delete memory faster, therefore cause and effect means faster releases can cause faster memory transactions/clearing memory.
MacBook Air 16/1 arrived. I left it on my table and it has been picked up. The only thing that I might have to do is to provide a power cord for it as it doesn't come with one. Also, maybe the NAS credentials; nobody seems to save them - they always ask me.
I've read this post four times and I still have no idea what you were trying to say.
Yup. They ship with a power brick and USB-C cable.
MagSafe Macs shipped with a power brick and cable (non-removable) to the laptop. Then they provided a cord that could plug into the wall and the brick along with a small piece that becomes part of the brick which you can plug in. The problem with the little piece is that the metal weakens over time so you plug it into the wall and it falls out of the wall.
You can buy the power cord separately for something like $19. I have a bunch of spares in one of my parts bins.
They do, just not the long power cord extension.
Well the power cable provides power to the brick. So they don't ship with a power cable. They ship with a power connector.
So this is something to keep in mind if you are considering 8GB and why I ultimately went with a 16GB version.
8GB of RAM is fine for my needs.....I have confirmed it with my 2019 i9 iMac by replacing my 128GB of RAM after market upgrade with the stock 8GB stick from Apple. I do 1080p video editing which, come on it should be no sweat for any computer these days. So I know 8GB of RAM is fine.
HOWEVER, and this is why I decided to go with 16GB. Something I was NOT thinking about initially was that my iMac has a Radeon Pro Vega 48, which is an 8GB GPU. Since the M1 processor now does not have dedicated RAM for system vs GPU, this means that the 8GB version the GPU needs to share that same memory.
SO - to match my iMac 8GB system RAM and 8GB GPU, I decided on the 16GB version. If you do something that uses the GPU a bit, it actually might be beneficial to get the 16GB version due to the shared memory pool - even if 8GB of system RAM is sufficient.
8GB of RAM is fine for my needs.....I have confirmed it with my 2019 i9 iMac by replacing my 128GB of RAM after market upgrade with the stock 8GB stick from Apple. I do 1080p video editing which, come on it should be no sweat for any computer these days. So I know 8GB of RAM is fine.
HOWEVER, and this is why I decided to go with 16GB. Something I was NOT thinking about initially was that my iMac has a Radeon Pro Vega 48, which is an 8GB GPU. Since the M1 processor now does not have dedicated RAM for system vs GPU, this means that the 8GB version the GPU needs to share that same memory.
SO - to match my iMac 8GB system RAM and 8GB GPU, I decided on the 16GB version. If you do something that uses the GPU a bit, it actually might be beneficial to get the 16GB version due to the shared memory pool - even if 8GB of system RAM is sufficient.
I'm curious - if 8GB meets your workflow needs, why did you spring for 128GB? Did your needs change substantially in the last year?
I went from 32GB in my outgoing Late-2015 iMac to 16GB in the new mini, but the big difference for me was that my employer transitioned a couple of years ago to Citrix for remote work. As a result, I no longer had to run Windows in a VM alongside OSX, which effectively halved my RAM footprint.
Because I was not sure if it would help, and upon asking for advice I was told to basically max it out! Which is why I sometimes wish people would be more careful with the "More RAM = better" or "get the most RAM possible" comments. $500 for 128GB after market. I make enough money where $500 wasn't a tough decision, but its still $500 that ended up being rather pointless. I took some online courses to get certified for video editing, and they mentioned 8GB was fine for 1080p, 16GB was fine for 4K. Since I only do 1080p, I was a bit curious if their comments were valid and sure enough, they were. 8GB version only took 20 seconds longer on a ~15 minute video. I did not run dozens or hundreds of tests so that 20 seconds extra could have been my system doing other things.
And honestly, placebo effect made me think my 2019 i9 iMac was better than the system it replaced - 2010 Mac Pro 8GB of RAM. But doing another test proved it does not actually improve much. Which is absolutely crazy that 9 years difference and my video editing experience is not any faster. The ONLY benefit my iMac has is working on HEVC files just for the file size benefits.
So I think the Mac mini would be great for my needs for quite a while. It can certainly beat a 2010 Mac Pro which is good.
I don't disagree with your overall reasoning, but would caution that with the shared video memory/unified memory architecture it is not a 1:1 relationship (eg 8gb of dDPU memory is not the same as 8gb of unified memory). It's also not magical in that unified memory architecture doesn't magically reduce total amount of memory needed either (although might in some circumstances).
The interactions between them are not straightforward, and will also depend on programming choices, use of different macos frameworks, use profile, programs used (how many would actually use the full 8gb of dGPU memory anyway?), etc. And of course as always how much the OS and programs use the memory to cache things (successfully and intelligently). Some programs will directly address GPU/CPU memory (reducing duplication/system I/O) and hence need less compared to intels with dedicated GPUs; some programs do things differently and don't use much GPU but will use all the system memory they can get.
Unfortunately that sort of leaves us back where we started - some will see more benefit from more memory than others, define "need more memory", tolerance of slowdowns (are they even noticed?), and whether the extra cost is significant to the buyer/user or not. Specific impact of on-board unified memory will vary a lot.
BTW I don't think pure benchmarks of time are the right measure in most cases (but for some, yes). When I do memory intensive stuff on my existing set-up, the main benefit I see with lots of memory is not increased speed for that task but that I can continue doing lots of other tasks without noticing. For the long-running intensive tasks I care less about raw speed but that the machine is 100% usable for the other stuff I'm doing. (This may not be representative, those tasks for me are mostly other I/O constrained - external disks - not CPU; they would be memory and hence swap constrained with less memory).
IMO you made the right choice for you. You could probably get by with 8gb but you're not that price sensitive, your needs may grow, etc.
The best way to check if you need more memory is to close everything and restart your computer in the morning. Then do your normal work that you do every day. At the end of the day open terminal and type in vm_stat and look at the Pageouts number. If it is significant, then you would definitely benefit from more memory. In my case it is not and that is over two days clearly.
Having said that nearly all people would benefit from the 8 to 16 jump. Modern operating systems will use as much memory as they can to cache frequently accessed files. This makes the OS experience feel more snappy and responsive.
Do I really need 32 GBs of RAM in this machine? Clearly yes, because even though I have not done anything intensive in the last few days, my cache is huge. It makes multi-tasking and having multiple applications and windows open a much more seamless experience. Would I benefit from more though, like 64 GBs? Probably not. As with everything there is a cost vs performance tradeoff that is first steep and then levels out.
Having said that nearly all people would benefit from the 8 to 16 jump. Modern operating systems will use as much memory as they can to cache frequently accessed files. This makes the OS experience feel more snappy and responsive.
Do I really need 32 GBs of RAM in this machine? Clearly yes, because even though I have not done anything intensive in the last few days, my cache is huge. It makes multi-tasking and having multiple applications and windows open a much more seamless experience. Would I benefit from more though, like 64 GBs? Probably not. As with everything there is a cost vs performance tradeoff that is first steep and then levels out.
Well I might not be using the full 8GB anyway either. But its not just 8GB of GPU I need to make up. Even if you only use 1 or 2GB of that GPU, that still means you LOSE 1 or 2 GB of actual system RAM for programs. Which means you don't have the full 8GB of RAM JUST for programs. Therefore, the comparisons we are making with Intel systems are actually the opposite to how we are thinking and in some cases we need MORE RAM on Apple Silicon than on Intel - like what I mentioned.
If you play games for example that use more GPU, that means you are sacrificing more of the system RAM because the GPU needs it.
Fact is, on my testing 8GB was fine on my iMac but I also had 8GB GPU. How much of that being used I do not know. If it was even 1GB being used, that would still mean I would only have 7GB in system RAM with Apple Silicon, and I do not know if that would be a tipping point for my workload or not. FCP does use the GPU for even light video editing for what I do, so it might utilize more than the 1GB in my example.
I would add that if you are running multiple monitors, that adds to the pressure on memory.
This is why I get Discrete GPUs with onboard RAM on my systems. No downside to more RAM.