Isn't that mostly an example of pitting AWS EC2 without swap (it was turned off to force you to pay for more RAM) vs a Mac with swap? Still, an interesting situation you were able to solve with a base model Mac. Compression and swap go a long way to getting the job done, or to add buffer performance, rather than shutting a job done.
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.
My new MacBook Air m2 - 8gb is great
- Thread starter colonelbutt
- Start date
- Sort by reaction score
No this is wrong.
Suggest you read some comp sci textbooks, particularly around virtual memory implementations. Marshall Kirk McKusick's Design and Implementation of the 4.4BSD operating system and Tanenbaum's Modern Operating Systems cover it. Also some later material on mmap (memory mapped files) and sparse data structures. Not sure if Knuth's later books cover that or not if I'm honest these days - has been about 30 years since I last looked.
I used to work on multi-user Unix machines with 32MB of RAM. And that was four VME cards worth of RAM that cost as much as a car each. In that case it was impossible to buy more RAM because (a) it didn't fit and (b) it was really expensive so it'd take 6 months to get a PO through to buy it and we'd probably have to wait for the next budget cycle. So we overcommitted and used swap. And we didn't need to do it that often so it wasn't worth spending on.
We actually designed and built operating systems with virtual memory systems so we could remove the cognitive load of spending time optimising running workloads and remove spending so much money on RAM. The trade off is of course wall clock time because we have to drop down the cache hierarchy which increases latency. Meh.
Incidentally I run workloads at work which have about 2TB of sparse allocation and we do this on 16Gb machines backed with NVMe. The tradeoff is wall clock time which is fine for that workload.
So really, if you hit memory pressure once a month, should you go and buy a 32Gb machine or just wait a couple of more minutes for your workload? The ROI is probably quite low if you have to spend 4x as much on hardware.
Well it's pitting two allocation methodologies against each other. The platform is irrelevant. In the case of the windows side of things, the allocator bundled in the package we use plays with the ancient HeapAlloc API instead of VirtualAlloc so it doesn't support overcommit. The macOS port of it, via homebrew, hits the native libc implementation of malloc which supports overcommit by default. But we can't modify the allocator on windows so moving it to somewhere that supports overcommit even with much less RAM means you still get over the line at the end of the day.
The point is if you use overcommit, you can get as much RAM as you need as long as you have swap for it. Disk is cheaper than RAM, even at Apple prices, and if you only use it occasionally then the tradeoff on wall clock time may be acceptable.
So basically it's not that you can't do the job with 8GB of ram, you might just have to wait a little longer. Boo hoo!
Passmark is a CPU benchmark test, right? If it's not saturating 8 GB RAM at all, or not by much, then it's not going to sell the idea that 16 GB RAM is better. The RAM debate is about saturation and bottlenecks, so one needs to be multitasking multiple apps and/or using a performance-based app that opens more than 6 GB of data on top of the 2 GB of running macOS.
The assertion that the SSD will fail if used too much for swap is a myth today based on information that is woefully out of date.
www.howtogeek.com
"When held up against real-world use for your average consumer, TWBs are astronomically high. Most 1TB drives have a TWB of 600TB. To hit that TWB value in 5 years, you'd need to write 328.8GB to disk every single day."
That is more than a lifetime for almost any Mac computer, or Windows PC for that matter. Those that are concerned about SSD wear, and present their opinion as fact, are woefully misquided and out of touch with reality.
8 SSD Myths and Misconceptions People Still Believe
These SSD myths have hung on for far too long.
www.howtogeek.com
"When held up against real-world use for your average consumer, TWBs are astronomically high. Most 1TB drives have a TWB of 600TB. To hit that TWB value in 5 years, you'd need to write 328.8GB to disk every single day."
That is more than a lifetime for almost any Mac computer, or Windows PC for that matter. Those that are concerned about SSD wear, and present their opinion as fact, are woefully misquided and out of touch with reality.
The assertion that the SSD will fail if used too much for swap is a myth today based on information that is woefully out of date.
8 SSD Myths and Misconceptions People Still Believe
These SSD myths have hung on for far too long.
"When held up against real-world use for your average consumer, TWBs are astronomically high. Most 1TB drives have a TWB of 600TB. To hit that TWB value in 5 years, you'd need to write 328.8GB to disk every single day."
That is more than a lifetime for almost any Mac computer, or Windows PC for that matter. Those that are concerned about SSD wear, and present their opinion as fact, are woefully misquided and out of touch with reality.
Yeah this.
Only SSD failure I have ever seen was in a DB server that was a write master on a huge cluster that was sucking up 5TB a day on transaction logs and that took three years. It exceeded the quoted TBW capacity by an order of magnitude. The disk only got replaced because the reallocation count exceeded the safe number. It could still be written to and read from.
500 desktop machines? No failures ever. 250 macs. No failures ever.
Thats why—when helping people configure a purchase—I ask what they do, for how long and how often.
You can double performance in some tasks by buying the extra RAM, but if you're only doing that once-per-month, then slower performance isn't a big deal.
But then again, for example, there was a video editor with a base model M1 that was swapping by over 20 GB all the time. There's more people like that, than we realize. They don't look at Activity Monitor, they don't really understand computer systems, so they just think, "This is how long it takes" or "this is how slow things are" and don't realize performance would have improved had they had sufficient memory.
My position is it's a case-by-case thing what people should buy.
That is more of an issue with SSD bandwidth, not memory usage. The 16GB, and higher models, have larger bandwidth because of the SSD chip configuration. Exporting photos does not consume large amounts of memory.
That's incorrect because both test systems had the same 256 GB drive. The only variable was RAM amount (8 vs 16).
Here's the video of the test results with timestamp.
Here is the timestamp showing the two test systems.
OK. I have done thousand image exports on my M2 Air with 16GB and it has never taken more than 5 minutes. RAW to JPG, cropping, color balance and exposure compensation. To have an export take 18 minutes something else is going on with those systems or the tests were biased in some way. 36 minutes with 8GB, something is seriously wrong. In fact I sometimes run two exports at the same time (Lightroom can do that) and both exports together took less than eight minutes.
You are right, that’s why MBA is the wrong machine for these tasks, instead MBP with minimum 36GB RAM would be more suited.
See this is what I call speculation. Can anyone point to any SSD they have had ever failing. I think this is just conjecture.
Art Suwansang is a pro photographer and Global Ambassador for BenQ and other photo companies. He's been making these test videos forever. He's legit so I trust he knows what he's doing in Lightroom... but maybe there's something I'm not thinking of. Are you exporting 45 megapixel RAW to full quality JPG?
Also, Art explained in the video that, at that time, Adobe supposedly hadn't yet optimized for the M2 so it wasn't concurrently utilizing the GPU for some reason (unlike the M1), so at the time of the test, this was simply a CPU-only performance test. Still, it illustrates how the CPU has to stop/slow during swap, which slows down conclusive results, which is the point I'm making.
Using Apple or Samsung tradein is always poor. I sell on eBay and make a decent amount for each one, as I do my phones when I upgrade.
Let’s make this simple . An Apple Mac for 1200. Sold after 3 years will be 600, including inflation. That is £200 a year, simple. The higher depreciation is on higher spec models. If you pay for more memory and SSD you never get it back relatively in resale. It’s why I stick to the cheapest Apple laptops. In addition, I go for a discount on the laptop by buying refurb or when Amazon or Johnlewis does a discount. My power machine is my windows desktop, which I build myself and upgrade every three years or so.
Although I have found that specs have slowed over time. Memory demands have hardly changed for more than a decade. The second fastest moving item is the cpu, as new ones are released every year, but by far the fastest is the graphics card.
The reason why graphics card progression is important, is because the games need more power. Battlefield 5, cyber punk, etc. needed a whole new generation of cards. PC memory is irrelevant. An 8gb PC with a RTX4xxx will run way way way faster then a 32gb PC with a 2 series or even 3 series card.
Last edited:
Congrats. I had a Lenovo 4g 128gb which lasted 6-7. Only ever used for web and videos. Most people are fine until it falls apart. I actually think 8gb is excessive for what more people use their laptops for
What if you have ~10 tabs open in Safari, not browsing, but working on some of them, Skype open for at least 8 hours, Some image and pdf editing in Preview, sometimes in GIMP, sometimes Word or Excel open for few minutes, maybe a YT video (some music) going on in the background, audio streamed out through Bluetooth to an external sound system for few hours, while working on Safari tabs, etc, would an 8GB m2 air be enough? All the time, the Mac would be plugged into electricity.
Actually, I am using a 2018 16GB MBP atm, which is doing very well, on battery or otherwise. The memory pressure is green all the time, it uses ~ 12GB, maybe just because it is there. Temperature~ 27°C most of the time, sometimes going up to 31.5°C.
Not thinking yet to replace this MBP, but lately had been playing with 13.6" M2 air in the shop. It is sort of affordable, but has just 8GB RAM. In my former Lenovo Ideapad S340 Windows (and Linux) with just 8GB, I could do a lot more than I mentioned above at the same time. No heating, no fans running and it never lagged. With macOS being much lighter than Windows, shouldn't 8GB enough?
Actually, I am using a 2018 16GB MBP atm, which is doing very well, on battery or otherwise. The memory pressure is green all the time, it uses ~ 12GB, maybe just because it is there. Temperature~ 27°C most of the time, sometimes going up to 31.5°C.
Not thinking yet to replace this MBP, but lately had been playing with 13.6" M2 air in the shop. It is sort of affordable, but has just 8GB RAM. In my former Lenovo Ideapad S340 Windows (and Linux) with just 8GB, I could do a lot more than I mentioned above at the same time. No heating, no fans running and it never lagged. With macOS being much lighter than Windows, shouldn't 8GB enough?
No, 46 megapixel RAW to compressed JPG (80%). Which should take more time than not compressing a JPG, I would think.
Yes, let's spend another thousand dollars for processing a task perhaps once a month. Why not go all the way and get 128 GB of memory? Or better yet, a Mac Studio with 192 GB of memory. Makes perfect sense, if it were not my money.
Lightroom classic
Depends how big the photos are and what her workflow is. The worst part of Lightroom is the catalogue, so I keep a folder structure with under 100 in each folder. After a certain amount of time I might start a new catalogue. I also keep my photos either on a micro sd with card reader when travelling or on a nas when at home.
It also depends on her workflow. For example I normally take small numbers of photos (say less then 50-100 a day when travelling) and spending time selecting ones I like. I then spend a few minutes editing each one. Also the new enhance functionality takes some time to run, almost a minute.
If she is processing a lot more and doing multiple or batch processing, she might want to have 16gb.
*Sigh*
Whatever if 8 GB is enough for tasks or not, the point is that RAM is cheap, faster than SSD swapping, and most importantly... it doesn't degrade with time.
I will for the moment accept SSD degradation only happens under significant use over a prolonged period of time, but the real question is if $200 for 8 GB extra is worth it. Our argument is that adding an extra ~$10 USD atop an already $1100 computer is worth it. This is what Apple would spend for an extra 8 GB without their outrageous upgrade costs attached.
The arguments presented by the 'pro-16 GB' crowd is and has always been regarding Apple's huge upgrade costs. If someone is happy with only 8 GB then good for them. However I wonder if they would have settled on those stats if the cost for upgrading were only $10 more.
Whatever if 8 GB is enough for tasks or not, the point is that RAM is cheap, faster than SSD swapping, and most importantly... it doesn't degrade with time.
I will for the moment accept SSD degradation only happens under significant use over a prolonged period of time, but the real question is if $200 for 8 GB extra is worth it. Our argument is that adding an extra ~$10 USD atop an already $1100 computer is worth it. This is what Apple would spend for an extra 8 GB without their outrageous upgrade costs attached.
The arguments presented by the 'pro-16 GB' crowd is and has always been regarding Apple's huge upgrade costs. If someone is happy with only 8 GB then good for them. However I wonder if they would have settled on those stats if the cost for upgrading were only $10 more.
I don't know why the discrepancy between his test and what you're saying your experience has been, without asking them personally. Maybe there's effects being applied during their conversion test. I think he performs the exact same test with every Mac, and he's been making videos for several years (I think 5 or 6).
Either way, the principle of the matter stays the same: swap slows down the CPU, and the more swap used, the slower the process.
Just picked up my Midnight MBA 15" 8GB / 512GB and absolutely in love with this machine. For my workflow, I notice absolutely no difference between this and my MBP M1 Pro. Loving the less weight and bigger screen. Couldn't be happier!
Thanks! You nailed the points. I started become bored discussing these topic with RAM police here.
"Enough" is relative to the individual.
Some people are very concerned about their time on a computer. For them, waiting an extra 3 seconds to open a Safari YouTube tab might be too much.
Each of us have different priorities.
For me, longevity of the system is most important, as is ease of use (meaning reliability.)
If it takes my computer 3 seconds to open a window when some more expensive configuration could do it in 2 seconds is not something that is really, really important to me.
Only you can assess what you need.
One thing that will be true for any laptop user, though: the internal SSD is much more convenient than having to schlep around USB SSDs to hang off the side. So if you are buying a laptop and intending to carry it most days of the week, the $200 upgrade tradeoff between more RAM or more SSD is likely better spent on the latter.
Last edited: