Video Demos Performance Differences Between 8GB and 16GB Apple M1 MacBook Pro

[mlykke]

8GB unified = 16GB standard (up to 4K workflow)
16GB unified = 32GB standard (for more than 4K workflow)

No, no and NO! Stop repeating that nonsense!

8 GB M1 = 8 GB Intel.
16 GB M1 = 16 GB Intel.

The M1 might have better performance and you might feel swapping slightly less, but swapping is still to be avoided. Swapping will ALWAYS be slower than a workload that has ram enough. Swapping is the last resort to ensure the computer doesn't have to run close down apps to stay running. Most tests of the 8 GB shows a fairly heavy use of swapping even in moderate workloads. So unless your workload is just a few tabs in a browser, some regular office stuff and watching Youtube, then you need to get the 16 GB version. And those of us which have a much heavier workload should really be hoping for a 32 GB and 64 GB option in future versions.

The M1 is impressive but you are not get a free launch. Your data and apps still require the same amount of ram - No ifs, ands or buts!

 

[Maximara]

Epyc servers have an entirely different purpose and usually run with 1TB+ of memory. They're designed for massive scientific simulations and feature film vfx and nobody cares what they cost or how much power they use.

Then why in the earth are they coming up in a discussion about consumer computers?! If anything this comparison of an Epyc series CPU to an M1 demonstrates how desperate the x86 crowd is.

 

[Kjs100]

A measly 8gb pfff. 😂

 

[Maximara]

No, no and NO! Stop repeating that nonsense!

8 GB M1 = 8 GB Intel.
16 GB M1 = 16 GB Intel.

No it isn't. Intel memory is accessed though a bus, M1 memory is on the chip. Ergo they are NOT comparative.

 

[mike...]

My Late 2013 15" rMBP has an NVMe SSD and 16GB of RAM. I can use 16GB of swap and it still runs fine. M1 is not doing anything special by using swap. But swap is always slower than RAM, and using too much swap will slowly kill your SSD.

No it isn't. Intel memory is accessed though a bus, M1 memory is on the chip. Ergo they are NOT comparative.

M1 uses LPDDR4X memory which is accessed just like DDR memory in any other computer. It is NOT part of the SoC, it's placed directly next to the M1 on the package but it is NOT part of the M1.

 

[TroyBoy30]

16 max is pitiful.

 

[amartinez1660]

That should be 13m 57s vs 5m 59s.

The present comparison of 8 vs. 16 leaves much to be desired. See my comments above.

A small note or message to them about the flaws and missing elements on their tests and they might revise it for the next ones.
They have been quite open and honest, the XCode build test is user provided for example and it seems to me they want their channel to get better and better the right way.
So much so that Vadim, one of the guys, dove as deep as he could into WWDC videos and extracted key info to the best of his understanding for the average user.

 

[mlykke]

No it isn't. Intel memory is accessed though a bus, M1 memory is on the chip. Ergo they are NOT comparative.

That affects the speed but not how much ram your apps and data require. So yes, they are still the same in terms of how much ram you need.

 

[Psychicbob]

Since these M1 Macs are essentially iPad Pros with extra memory, ports and cooling, I don’t see people complaining about memory usage and swapping with pro work on the iPads, so these M1s should do just fine!

 

[mlykke]

Since these M1 Macs are essentially iPad Pros with extra memory, ports and cooling, I don’t see people complaining about memory usage and swapping with pro work on the iPads, so these M1s should do just fine!

you are comparing a tablet/phone which runs a single app at a time with a computer that will usually run multiple things at a time. It's not the same.
Again, you need the same amount of ram with the M1 as you did your previous Intel based Mac. The M1 might be faster, but thats the main difference.

 

[mike...]

Since these M1 Macs are essentially iPad Pros with extra memory, ports and cooling, I don’t see people complaining about memory usage and swapping with pro work on the iPads, so these M1s should do just fine!

The difference between an iPad and a Mac is the OS. These Macs don't handle memory anything like iPads do. iPad OS will deliberately kill any background apps you have open to free up memory for your foreground app. macOS doesn't do that, instead it keeps everything running by using swap instead.

Addionally, iPad OS does not support true multi-tasking. Any app in the background that wants to do work in the background has to do so for a very specific reason, and use one of Apple's very specific APIs in order to achieve it.

On a Mac, apps are always running, in the foreground or background, and can do whatever work they like without any special APIs.

 

[andyyau]

No, no and NO! Stop repeating that nonsense!

8 GB M1 = 8 GB Intel.
16 GB M1 = 16 GB Intel.

The M1 might have better performance and you might feel swapping slightly less, but swapping is still to be avoided. Swapping will ALWAYS be slower than a workload that has ram enough. Swapping is the last resort to ensure the computer doesn't have to run close down apps to stay running. Most tests of the 8 GB shows a fairly heavy use of swapping even in moderate workloads. So unless your workload is just a few tabs in a browser, some regular office stuff and watching Youtube, then you need to get the 16 GB version. And those of us which have a much heavier workload should really be hoping for a 32 GB and 64 GB option in future versions.

The M1 is impressive but you are not get a free launch. Your data and apps still require the same amount of ram - No ifs, ands or buts!

This is the only and correct description for M1 memory. So many people never know how CPU and RAM work and give so many BS. And most YouTube review videos never address the ram limit but giving misleading points.

It is very easy to verify. Open Activity Monitor, after hour of use, if Swap Memory is 0 byte for M1 8GB model, than Apple has the magic. If you see many GB there, your M1 memory is not enough.

My Mac has 32GB ram, after 4 hours of light works, it shows ZERO byte swap memory. Any M1 8GB users are welcome to challenge me if your swap memory is 0 byte.

From those YouTube reviews, I am seeing 2GB to 7GB swap memory for M1 8GB models.

 

[crevalic]

No it isn't. Intel memory is accessed though a bus, M1 memory is on the chip. Ergo they are NOT comparative.

This is completely wrong. There is absolutely no RAM on the M1 chip, the RAM chips are separate from the SOC and are standard, off-the-shelf LPDDR4X chips which can be found in many laptops released in the last year, including the early 2020 Intel MBA. The SOC uses a completely standard bus to access the memory, more specifically a 128-bit bit bus split into 8x16-bit memory channels.

Here is a data sheet of a popular laptop Intel 10th gen chip released in 2019, which uses a 128-bit bus to access LPDDR4X RAM.

You can find a photo of the M1 chip with annotations in this great article, where you can use your own eyes to confirm that no memory can be found on the chip, however, you can easily see the standard memory channel interfaces.

 

[Cicnod]

This is the only and correct description for M1 memory. So many people never know how CPU and RAM work and give so many BS. And most YouTube review videos never address the ram limit but giving misleading points.

It is very easy to verify. Open Activity Monitor, after hour of use, if Swap Memory is 0 byte for M1 8GB model, than Apple has the magic. If you see many GB there, your M1 memory is not enough.

My Mac has 32GB ram, after 4 hours of light works, it shows ZERO byte swap memory. Any M1 8GB users are welcome to challenge me if your swap memory is 0 byte.

From those YouTube reviews, I am seeing 2GB to 7GB swap memory for M1 8GB models.

Sorry I don't buy it.

I have 32GB on my Mac mini currently and I do nothing heavy at all and I have used 1.06GB swap. I open spreadsheets, work on some pdfs, mail, internet, word documents, a little teams, and that is about it.

 

[theluggage]

8GB unified = 16GB standard (up to 4K workflow)
16GB unified = 32GB standard (for more than 4K workflow)

Sorry, no. 8GB of unified RAM = 8GB of standard RAM, if you have a workload that actually needs lots of data loaded into RAM. Plus "4k workflow" is a piece of string: how many sources, effects etc? Which formats?

What we're seeing in these tests is that for many single benchmarks, the combination of a faster CPU, faster GPU, other new gizmos like the Neural Engine and hardware codecs, significantly faster SSD, faster RAM access (on-package RAM and no copying between system RAM and VRAM) makes the 8GB model capable of things you previously wouldn't have even tried on a low-end 13" MBP. Many of the tests shown on that video probably wouldn't have stressed the RAM on an 8GB Intel Mac - they'd have been limited by CPU, GPU or SSD speed instead - and unfortunately the tester didn't know to look at the "Memory pressure" - rather than the almost meaningless "Memory used" - which would have shown which, if any, tasks were being limited by RAM.

Or, to put it another way, the old Intel 2-port, 13" MBP would have probably choked badly on the 8K video test even if it could have had 32GB of RAM.

...and none of those tests looked into the issue of typical "messy" workflows with multiple applications open, VMs running, multiple browser tabs etc.

It is clear that the 8GB M1 will be more than enough for most people. ...but then it is probably enough for most people with lower-end Intel machines which will run out of CPU/GPU grunt before they run out of RAM. Non-upgradeable machines force people to get extra RAM for "future proofing".

As for 32GB or more - people shouldn't be upgrading to 32GB anyway unless (a) it's a user-upgradeable machine and there's no inflated Apple RAM prices to make you skimp or (b) they know they have specific workloads that require lots of data loaded into RAM. If you've got that sort or workload, you should also be wondering how the Apple Silicon 16" MBP and iMac replacements are going to perform with (maybe) 12 cores and a better GPU, as well as other considerations like better multiple monitor support, more ports/I/O bandwidth...

 

[LiE_]

There doesn't seem to be too much of a performance hit when the M1 is using swap memory with the 8gb system

Not in my experience, when the system has a lot of swap and memory is full things start to get laggy. Animations are jerky as apps are pulled from swap, it destroys the buttery smooth feel. Pop in as apps are pulled from swap can be quite jarring.

 

[mlykke]

Sorry I don't buy it.

I have 32GB on my Mac mini currently and I do nothing heavy at all and I have used 1.06GB swap. I open spreadsheets, work on some pdfs, mail, internet, word documents, a little teams, and that is about it.

I don't know the exact details of how Apple and Mac OS handle swapping. It is definitely possible to see a very low amount of swap even though all ram is not used. But if you see 1 GB thats low and most likely, only rarely used data, IF all your normal ram is not fully used.
Experience tells me that if you're the type of user that often just puts the computer to sleep and there might be weeks or months between a restart, then I do notice the swap usage increasing a bit. The difference is the 8 GB M1 can only use around 6-6.5 GB total, before it needs to swap and that happens very fast if you run more than just a few simple apps.

 

[IllIllIll]

LOL @ "future proofing" a sub-$1500 device.

If you have to do that, you need a better job.

 

[theluggage]

Beware of the swap disk space!

Swap disk space is not an issue. Swap rate is an issue - i.e. how often data is having to be moved in and out of swap space - and Activity Monitor has a handy "Memory Pressure" readout that aggregates swap rate and various measures of RAM usage into a single indicator of whether performance is being constrained by RAM.

@ All the people bitching about 16gb not being enough RAM for a pro machine

We don't know how fast that 8K video test - the one that took 10 mins on a 8GB machine and 6 mins on a 16GB machine - would have run on a hypothetical 32GB M1 machine (especially since the video didn't look at the RAM pressure or other metrics to see what was limiting it). Maybe it would have shaved another 2 minutes off, maybe 16GB was enough to peg out the processor, maybe it would have at leas allowed you to get on with a bit of photoshopping while the render was finishing...

What we're currently seeing is that the M1s - at least on "tests" - can turn in similar performances to the Intel 16" MBP etc. That's a temporary hiatus until the Apple Silicon replacements for the higher-end Macs appear, which will hopefully have other "pro" features like more ports and support for >2 displays. Unless there is some urgency, or they can afford a M1 to experiment with (which is what the M1 Mini is perfect for), "pro" users should be waiting for the other shoe to drop.

 

[mlykke]

LOL @ "future proofing" a sub-$1500 device.

LOL @ somebody LOL'ing people discussing a tech device. Get over yourself.

16 GB is not for future proofing, but for being able to have more apps, tabs and files open without slowing performance. It will also help going forward if you expect to keep your device for a number of years. Not every has usage that makes them want to upgrade every single time a new computer comes out.

 

[Maximara]

A few others swap which the video the doesn't mention.
There doesn't seem to be too much of a performance hit when the M1 is using swap memory with the 8gb system, but its gonna wear down that SSD alot quicker which is not replaceable of course. I know there will alot who won't care and will sell the unit after a year anyway leaving it someone else's problem.

I knew it wouldn't be that soon but based on the 2016 Find out how much longer your SSD will last and the 2006 What is the life expectancy of the MBP's PCIeSSD unless the read/write cycles are insane the SSD will be lasting a long time.

"Yes, I found this Samsung 950 SSD Chart for comparison.
It appears that the some older budget 3.5" SSD are not very good some failing at 100tb.
So I guess you get what you pay for in most cases.
Most users will only write about 5gb. per day.
Samsung offers a 5 year warranty or 200tb on the 950-series.
At 200tb divided by 5gb per day = 40000 divided by 365 days = 109 years
Most SSD's even the New Budget 3.5 6gb. SSD can achieve and surpass 200tb.
So I would assume it is safe to say that not to worry about the PCIe SSD going bad in the Mac as it will most likely surpass the life of the computer in most cases for the average user.
TB writes appears to double with larger capacity SSD's.
So an SSD with 500gb. should last twice as long as an SSD with 256gb."

The TL DR comment is
"It is safe to say that purchasing a MB with a 500gb SSD is your best value purchase.
Where this 500gb. SSD should last twice as long as an SSD with 250gb.
At least 400tb writes. At 5gb. per daily usage this SSD should last up to 219 years."

If this is still correct then even if for some insane gonzo reason you were read-writing 1TB a day on a SSD with a 700TB read-write life span it would last 700 days or about 1.9 years. At the rate the author of the 2016 article used data very few are going to get to those levels and certainly not the average user.

 

[Cicnod]

I don't know the exact details of how Apple and Mac OS handle swapping. It is definitely possible to see a very low amount of swap even though all ram is not used. But if you see 1 GB thats low and most likely, only rarely used data, IF all your normal ram is not fully used.
Experience tells me that if you're the type of user that often just puts the computer to sleep and there might be weeks or months between a restart, then I do notice the swap usage increasing a bit. The difference is the 8 GB M1 can only use around 6-6.5 GB total, before it needs to swap and that happens very fast if you run more than just a few simple apps.

Yeah that is pretty spot on. I reboot maybe once a month.

 

[mlykke]

Most users will only write about 5gb. per day.

Try to take an 8 GB M1, open 30 tabs in Chrome and then check the amount of swap written at the end of the day. I absolutely guarantee that it will be way above 5 GB.

 

[mike...]

LOL @ "future proofing" a sub-$1500 device.

If you have to do that, you need a better job.

Maybe people just spend their money more wisely than you? Better to spend that $200 up-front and invest the money you'll save paying for an early upgrade.

Maybe people just care more about the environment than you? Making computers is very environmentally damaging, and upgrading regularly is a great way to increase your carbon footprint.

I'm both of these people. I have a £3500 desktop PC and a £2500 MacBook Pro, both over seven years old now. They're due to be replaced with the MacBook coming first as soon as Apple reveal.

Instead of skimping on a few hundred up-front I've been able to put thousands more in to investments. A quick result from the calculator of one of the products shows that £6000 invested in 2015 is worth £13,500 now. It pays not to be profligate, however rich or poor you may be.

 

[mike...]

The TL DR comment is
"It is safe to say that purchasing a MB with a 500gb SSD is your best value purchase.
Where this 500gb. SSD should last twice as long as an SSD with 250gb.
At least 400tb writes. At 5gb. per daily usage this SSD should last up to 219 years."

If this is still correct then even if for some insane gonzo reason you were read-writing 1TB a day on a SSD with a 700TB read-write life span it would last 700 days or about 1.9 years. At the rate the author of the 2016 article used data very few are going to get to those levels and certainly not the average user.

Nobody writes 4GB a day, even my idle work Mac can write 15GB a day.

If you exceed your available memory by a lot, you will write a lot to disk. I know this having already killed one SSD in my 16GB MBP. I write 200-300GB on an average day with 1TB on my worst.

Also Samsung's basic Evo 970 250GB drive has a warranty of just 150TBW. Apple ships these Macs with basic 256GB drives too. I doubt they have particularly high endurance.

Another factor to consider is how full a user's SSD is. The closer it gets to being full, the more writes are concenctrated in to fewer cells, making it very difficult for the SSD to do wear-levelling.

TL;DR: if you significantly exceed your available memory for long periods of time it is well within the realms of possibility that you will kill an SSD. And you probably don't want to pay Apple's prices for a logic board replacement. Buy more RAM.