I've read somewhere that apples m1 8 gigs a ram = 16 gigs of intel?

[Hoo Doo Dude]

Another long time IT guy here, decades as a tcp/ip network analyst and sys admin. I agree that the M1 systems are a tremendous advance in technology. The term "paradigm shift" gets overused a lot, has been reduced to corporate speak. But this is about as close to that as you can get in the context of the way the majority of personal computer users work. I've owned some serious Intel hardware in the past and in November reached the end of the line for my day to day system which I use for photo editing and music production. I had an i7 system with 32 GBs of memory, 3 TB hard drive internal and 6 TB external, and a high end discrete graphics card. I had a laptop I used for travel that was also long in the tooth. To simplify my setup I elected to try out just using a portable with a dock and external monitor with an exterior drive. Initial reports on the new M1 machines were extremely impressive so I thought I'd try something different. I ordered a MacBook Air M1 with 16 GB of memory and the 2 TB drive. I got my system in early December and have never looked back.

This machine is significantly faster than my previous desktop machine in every respect. It has been rock solid reliable, responsive, nothing about it makes me wish I had more. I use it with an LG Ultrafine 4K monitor, a Caldigit Element hub, and a Samsung T7 2 TB Ssd for external storage. There is no downside, photo and audio editing are nearly instantaneous, some video processing I've done is much faster than my previous system, and the 16 GB of memory is ample.

Everybody's experiences and uses are different. I'm sure some may find ways to hit the all with the 16 GB systems. But I suspect the vast majority of uses are going to find this is indeed a new era where our previous perceptions no longer apply. These are superbly designed and executed systems that should satisfy all but the most demanding users.

 

[JMacHack]

The answer is always “It depends” for technical questions like this.

“Yes” in the sense that the machine will feel fast and responsive even at the limits of 8gb. Opposed to an x86 machine which will slow down as it swaps memory less efficiently.

“No” because 8gigs is 8gigs, no matter the architecture. If you fill up the 8 gigs it will still start swapping from the ssd.

There’s other considerations like Apple’s memory compression that I’m not sure how it affects the memory usage, but the gist of it is that it’s not entirely true.

 

[m1maverick]

The answer is always “It depends” for technical questions like this.

“Yes” in the sense that the machine will feel fast and responsive even at the limits of 8gb. Opposed to an x86 machine which will slow down as it swaps memory less efficiently.

“No” because 8gigs is 8gigs, no matter the architecture. If you fill up the 8 gigs it will still start swapping from the ssd.

There’s other considerations like Apple’s memory compression that I’m not sure how it affects the memory usage, but the gist of it is that it’s not entirely true.

Can you provide details on how the M1 swaps more efficiently / less efficiently than an x86 system which makes such swapping noticeable on the x86 system?

 

[jdb8167]

Can you provide details on how the M1 swaps more efficiently / less efficiently than an x86 system which makes such swapping noticeable on the x86 system?

No one can really give details since Apple hasn't disclosed much.

One thing is that the M1 has decent memory bandwidth at about 68 GB/s and low latency (though I haven't seen the latency characterized in concrete terms.) Add that to decent SSD performance and the ability to move between RAM and SSD at about 3 GB/s you get pretty good swap speed. There is nothing radical about any of this except having it in the context of lower end machines with mobile processors. The overall performance is top notch.

Another thing is that the M1 has 4 high efficiency cores that provide the ability to compress and swap memory in the background without necessarily affecting whatever foreground task is running. The ability to compress memory in hardware probably helps too.

I suspect that it is a combination of features that allow what appears to the user as almost free virtual memory swapping and the ability to use much more RAM than actually is available with little perceived performance loss. This won't work for memory intensive applications but for many applications low memory performance will seem much better than expected.

 

[m1maverick]

No one can really give details since Apple hasn't disclosed much.

If details are unavailable then how can such a statement be made?

 

[jdb8167]

If details are unavailable then how can such a statement be made?

Some of it is known. Memory bandwidth I got from Anandtech. We know from system tools like Activity Monitor that memory is compressed. Other tools can give r/w performance of the NVMe SSDs. All the specific details are only available to Apple though as far as I know. I'm making educated guesses based on what is known.

We are only 4 months or so into the new architecture. More details will keep filtering out over time.

 

[leman]

If details are unavailable then how can such a statement be made?

Real-world comparisons show that M1 machines manage to stay considerably more responsive and more performant than similarly-speced x86 machines in high memory pressure scenarios. More agile swapping/memory management is really the only reasonable explanation. My bet is on a combination of larger memory page size, hardware memory compression, low-latency interrupts and some SSD controller magic.

 

[jdb8167]

Real-world comparisons show that M1 machines manage to stay considerably more responsive and more performant than similarly-speced x86 machines in high memory pressure scenarios. More agile swapping/memory management is really the only reasonable explanation. My bet is on a combination of larger memory page size, hardware memory compression, low-latency interrupts and some SSD controller magic.

I forgot about the larger memory page size. Good catch.

 

[Ethosik]

If one needs 16GB or RAM then one needs 16GB of RAM regardless of whether they're using an Intel or M1 based Macintosh. The paradigm shift from Intel to M1 hasn't changed that.

Sometimes people think they need 16GB but they don't. I recently went through this with my iMac. I have 128GB of RAM and all I do is 1080p video editing. Guess what....Adobe software uses as much as I give it - 118GB of RAM....for 1080p videos. This does not mean I need 118GB of RAM though. Sometimes software uses as much as you have.

Obviously there are workflows out there that need this type of RAM. We have a server with 4TB of RAM and its actually needed.

 

[ADGrant]

No. However, my 16GB M1 13” MBP renders PP videos faster than my 32 GB 16” MBP. That’s unscientific, I know, but the unified memory is showing it’s abilities well.

Or maybe it's the faster single core performance or the codecs? What resolution are these videos?

 

[jinnyman]

8 gig is 8 gig. The vastly improved performance is somewhat misleading. Based on my experience with M1 Mini with 8 gig, 8 gig is not enough, atleast for me.

Put it simply,

If you were the right target audiences that the previous intel Macbook Air(cheap price/light weight. not including lightest mobile users with heavy usage) and entry Mac Mini were aiming for, you probably should be ok with 8 gig.
Otherwise, if you are pro users, or use the stuff for living and has to go M1 because there's no option of better Apple Arm machine, go 16gig.

 

[sebfrey]

No way, my 13" Intel MacBook Pro with 16 GB runs many apps without issue, my 8 GB M1 Mac mini seizes up with only a few apps running.

 

[theluggage]

No. However, my 16GB M1 13” MBP renders PP videos faster than my 32 GB 16” MBP. That’s unscientific, I know, but the unified memory is showing it’s abilities well.

What you need to do is go look at the ”memory pressure” reading on both machines. Ignore “memory used” since MacOS will try to use up spare RAM for caches etc. If it is low on both then you’re not testing memory capacity at all. If it is high on the M1 then the M1 is being slowed down by lack of memory (and, possibly, you’re putting extra wear on the SSD).

“Unified memory” is mainly a time saver - stops data having to be copied between main RAM and VRAM/caches in GPUs etc. It won’t generally reduce the amount of RAM used, except in some cases where software has been specially optimised. What you’re probably seeing is the M1 being all-round more efficient and having hardware assist for video codecs etc.

 

[chabig]

Based on my experience with M1 Mini with 8 gig, 8 gig is not enough, atleast for me.

What measure are you looking at to make this statement?

 

[chabig]

my 8 GB M1 Mac mini seizes up with only a few apps running.

This I don't believe.

 

[Digitalguy]

The M1 Macs are more efficient with 8GB than Intel Macs, but this comes at the cost of more SSD wear (there are quite a lot of reports online about that). It's still a good trade-off since SSDs tend to last more than the time a person generally uses a Mac. Having said that saying 8GB is equal to 16GB is a big exaggeration, and the general rule should be if you need 16GB you are better off buying a 16GB Apple Silicon Mac too.

 

[Ratsaremyfreinds]

I recall when it was a good price to pay 20 bucks per meg lol

 

[m1maverick]

Some of it is known. Memory bandwidth I got from Anandtech. We know from system tools like Activity Monitor that memory is compressed. Other tools can give r/w performance of the NVMe SSDs. All the specific details are only available to Apple though as far as I know. I'm making educated guesses based on what is known.

We are only 4 months or so into the new architecture. More details will keep filtering out over time.

If some of it is known then why not answer my question? If all you can do is guess then I'm fine with that but then I would ask that statements be phrased as much.

 

[m1maverick]

Real-world comparisons show that M1 machines manage to stay considerably more responsive and more performant than similarly-speced x86 machines in high memory pressure scenarios. More agile swapping/memory management is really the only reasonable explanation. My bet is on a combination of larger memory page size, hardware memory compression, low-latency interrupts and some SSD controller magic.

Which real world comparisons are you referring to? Can you provide references to them?

 

[m1maverick]

Sometimes people think they need 16GB but they don't. I recently went through this with my iMac. I have 128GB of RAM and all I do is 1080p video editing. Guess what....Adobe software uses as much as I give it - 118GB of RAM....for 1080p videos. This does not mean I need 118GB of RAM though. Sometimes software uses as much as you have.

Obviously there are workflows out there that need this type of RAM. We have a server with 4TB of RAM and its actually needed.

That's a completely different topic.

 

[m1maverick]

The M1 Macs are more efficient with 8GB than Intel Macs, but this comes at the cost of more SSD wear (there are quite a lot of reports online about that). It's still a good trade-off since SSDs tend to last more than the time a person generally uses a Mac. Having said that saying 8GB is equal to 16GB is a big exaggeration, and the general rule should be if you need 16GB you are better off buying a 16GB Apple Silicon Mac too.

How are they more efficient?

 

[jdb8167]

If some of it is known then why not answer my question? If all you can do is guess then I'm fine with that but then I would ask that statements be phrased as much.

I'm sorry but I answered as well as I could. I certainly hedged many things that I am not completely certain of. Perhaps English isn't your first language? Colloquialisms like "I suspect" and hedges like "probably" should indicate the parts that I'm less sure of. I'm sorry that that wasn't clear to you.

If you want to ask specific questions about my post, I'm perfectly willing to answer to the best of my ability.

 

[MacModMachine]

Can you provide details on how the M1 swaps more efficiently / less efficiently than an x86 system which makes such swapping noticeable on the x86 system?

I actually covered this in another thread , I explained it in depth how it is more efficient. from a ARM developers prespective.

 

[leman]

Which real world comparisons are you referring to? Can you provide references to them?

There was for example this video on the Max Tech channel (I think

) where he does a heavy workload torture test with some interesting results. Now, the reviewer is not the most tech savvy, a bit over enthusiastic and his methodology is not necessarily very scientific, but it’s nevertheless a rather interesting impressionistic test that highlights the responsiveness of these machines in high memory pressure scenarios.

 

[m1maverick]

I actually covered this in another thread , I explained it in depth how it is more efficient. from a ARM developers prespective.

How about summarizing it here?