Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

djjeff

macrumors 6502
Jun 10, 2020
318
162
It relates to how a system feels in performance. There is a delta between SSD and RAM access but Apple is apparently hiding it really well in IPC and other means.
How? I expect your answer to be architecture related and not operating system related.
 

lightfire

macrumors regular
Aug 10, 2017
143
30
One of the early systems I worked on back in the mid-1970s was a PDP-11 (might have been a 45). This system had 28 KB of core. The operating system ran in 20 KB and user space was 8 KB. The OS swapped out a user to disk from memory and then swapped in a user into memory from disk, gave that user some time and so on. The user interface was ASR-33s which were teletypes at 10 characters per user and you generally felt like you had the whole computer to yourself. I think that the maximum number of users configured was 8. So you ostensibly had people that would normally need 64 KB of user memory sharing 8 KB of actual memory.

Paging became popular later on in Virtual Memory Systems.

Wow...It’s been a while since I have heard of those. I remember the company.I was working for bought me a PDP-11/70 to use. Those were the days. This is one reason I am always amazed at these little bitty devices we have now.
 

burgerrecords

macrumors regular
Jun 21, 2020
222
106
If you do a dummy purchase of a Macbook Air, after you click Buy you should see a 16GB upgrade option. With the Air it seems clicking 16GB also automatically ups you to 512GB storage (so no 16/256 option).
This is not my experience I have a 16/256 option
 

smoking monkey

macrumors 68020
Mar 5, 2008
2,363
1,508
I HUNGER
The problematic claim (which AFAIK Apple *hasn't* made themselves) is that there is some kind of magic that makes 8-16GB RAM in an M1 system work like 32-64GB of RAM in an Intel system.
Yeah. Apple never made that claim and never would.

And it was just something that somebody postulated and then mass hysteria kicked in and you know the internet... And I think the original postulation (from as far as I can remember as I've been following ARM threads very closely) was 8 would be like 12-16. But is it now 8-16 = 32-64? Haha. These computers are awesome, but not that awesome!
 

aajeevlin

macrumors 65816
Mar 25, 2010
1,427
715
Buddy, no need to get worked up, I'm just sharing my experience. I think you may be in the minority in being confused about what I meant by base air though. Not sure what it was about my post discussing memory pressure and swap that made you think I had no clue what laptop I just purchased, where I am or what I'm talking about.
Yea base air is pretty clear. multiple website and almost all the reviewer uses that terms.
 

Ethosik

Contributor
Oct 21, 2009
8,142
7,120
You're using reason again ...

No need for that, I hope. It was a good discussion until a certain dynamic entered into it with one person in particular. Some trimming might do, and then people can return to reasoning.
I agree. I was liking the discussion until all of this nonsense happened. I asked many times for both of them to stop.
 
  • Like
Reactions: Sanpete

dmccloud

macrumors 68040
Sep 7, 2009
3,146
1,902
Anchorage, AK
Yet testing shows that the M1 Macs consistently outperform their Intel counterparts while using less RAM and swap space and having a lower memory pressure. The testing (from multiple sources) consistently indicates that the M1-based systems stretch that 8GB of RAM much further than anything Intel has made (regardless of OS being used). When these head to head comparisons are also showing the M1 outperforming the 16" MBP that comes with a MINIMUM of 16GB RAM (compared to the 8GB on the M1s being tested in most cases), it becomes clear that Apple has done something in terms of memory management that nobody else has done. Your claims are in terms of absolutes, but the numerous articles, videos, reviews, etc are showing that there is not an absolute position to be had. Your position is also based on the flawed assumption that if you simply replace the underlying platform, all other factors that affect RAM requirements will remain unchanged. If a new processor handles memory differently and uses a completely different code base, then the RAM requirements can change drastically based on that alone. Granted, the new system could actually require more RAM than its predecessor depending on the exact changes between the platforms, but in this case all indications are that the M1 with 8GB RAM is running neck and neck with many systems running 16GB or more RAM in the vast majority of cases (barring incompatibilities with certain applications or plugins).
 
Last edited by a moderator:

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
I was cleaning up my office the other day and found my old MacBook Pro. Out of curiosity, I plugged it in and started it up to look at the specs.

I found that it had a whopping 1GB of RAM. Yet, I managed to open browser windows, run Microsoft Office applications, and do all sorts of things which people now claim couldn’t possibly be done with only 8GB.

I have a 1977 HP-67 with 240 Words of RAM and I can play games I wrote for it back in the 1970s. That amount of RAM was fine for that operating system back then. I also have Dell Inspiron 4000 and 4100 that run Windows XP though I haven't booted them in a couple of years. Those were Pentium 3 models I believe.

I used a system with 8K of RAM in 1973 and that ran fine too. There is a reason why Apple doesn't support newer operating systems on older hardware. Microsoft is actually a lot better at supporting old hardware compared to Apple.
 

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
Yet testing shows that the M1 Macs consistently outperform their Intel counterparts while using less RAM and swap space and having a lower memory pressure. The testing (from multiple sources) consistently indicates that the M1-based systems stretch that 8GB of RAM much further than anything Intel has made (regardless of OS being used). When these head to head comparisons are also showing the M1 outperforming the 16" MBP that comes with a MINIMUM of 16GB RAM (compared to the 8GB on the M1s being tested in most cases), it becomes clear that Apple has done something in terms of memory management that nobody else has done. Your claims are in terms of absolutes, but the numerous articles, videos, reviews, etc are showing that there is not an absolute position to be had. Your position is also based on the flawed assumption that if you simply replace the underlying platform, all other factors that affect RAM requirements will remain unchanged. If a new processor handles memory differently and uses a completely different code base, then the RAM requirements can change drastically based on that alone. Granted, the new system could actually require more RAM than its predecessor depending on the exact changes between the platforms, but in this case all indications are that the M1 with 8GB RAM is running neck and neck with many systems running 16GB or more RAM in the vast majority of cases (barring incompatibilities with certain applications or plugins).

The two places where they could make a difference are hardware architecture and the operating system as I wrote before. Hardware support has been added over the decades to improve Virtual Memory performance but you usually need corresponding changes in the operating system.

It would be interesting to see what Apple did here.
 

dmccloud

macrumors 68040
Sep 7, 2009
3,146
1,902
Anchorage, AK
LOL! I read this while waiting to pick up my new Mini (one cannot just walk into an Apple store these days, you have to make an appointment). Let's just say I was not impressed. Just because some blogger posts information to the Internet does not mean that blogger knows what they're talking about. This is a prime example.

Then you didn't read it in full:

But I think it’s also the case that for most tasks for most people, these M1 Macs need less RAM to perform equivalently to, if not better than, their Intel-based counterparts. This sounds like ******** but it’s not. It’s the result of a hardware system architecture whose design aligns with Apple’s software architecture.


First, an intriguing benchmark from David Smith, an engineer at Apple:


Fun fact: retaining and releasing an NSObject takes ~30 nanoseconds on current gen Intel, and ~6.5 nanoseconds on an M1
… and ~14 nanoseconds on an M1 emulating an Intel.

In short, the OS and processor are designed to work together in unison, as opposed to the x86 versions of MacOS or even Windows, where the x86 architecture is being used because "that's what most people use." Also, I would point out that while John Gruber (and Daring Fireball by extension) does have his critics, he is by no means just "that blogger". If that was the case, he wouldn't be interviewing Craig Federighi and Greg Joswiak during WWDC to talk about what they had just announced.



 

Frixos

macrumors 6502
Nov 17, 2020
253
281
I’m not sure how people max out 8GB “running a few tabs and YouTube videos”
I opened a dozen Firefox pages, a few excel sheets, word docs, ppts, Zoom, Music, Photos, and a handful more apps - and the memory pressure was definitely still in the green with memory usage being around 6gb. Swap was also only a few hundred mb

(on a 2012 MBP)
 

MacBH928

macrumors G3
Original poster
May 17, 2008
8,738
3,895
To the guys saying 8GB good for few tabs on a browser:-

8GB M1 is a Logic Pro monster! Can handle over 1000 plugins, 128 samples, 195 channels ... till it runs out of resources. With a responsive UI the whole time! (The YouTube Video)


I was cleaning up my office the other day and found my old MacBook Pro. Out of curiosity, I plugged it in and started it up to look at the specs.

I found that it had a whopping 1GB of RAM. Yet, I managed to open browser windows, run Microsoft Office applications, and do all sorts of things which people now claim couldn’t possibly be done with only 8GB.
exactly what I am trying to explain in this thread.
I’m not sure how people max out 8GB “running a few tabs and YouTube videos”
I opened a dozen Firefox pages, a few excel sheets, word docs, ppts, Zoom, Music, Photos, and a handful more apps - and the memory pressure was definitely still in the green with memory usage being around 6gb. Swap was also only a few hundred mb

(on a 2012 MBP)
I think most people understand RAM like its storage space that once its filled you need more. They don't understand RAM is loaded data, once its filled the computer can swap it out and bring new data in and swap back to the prior data
 

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
I’m not sure how people max out 8GB “running a few tabs and YouTube videos”
I opened a dozen Firefox pages, a few excel sheets, word docs, ppts, Zoom, Music, Photos, and a handful more apps - and the memory pressure was definitely still in the green with memory usage being around 6gb. Swap was also only a few hundred mb

(on a 2012 MBP)

I used to do Firefox performance testing, back in the day when it used about 30 MB of memory. One way to get the memory use way up was to open a lot of very large JPEG files.

Another way to allocate a lot of memory is to write a C program which malloc's a lot of memory. I think that I could malloc 1 GB back in the Windows XP x64 days. Running several processes that allocate 1 GB should increase memory pressure.
 

Beefie

macrumors newbie
Jan 14, 2019
7
5
I had to put my 2c in.

RAM usage is a function of software design, and not hardware design.
Apple transitioning into its own hardware design does not mean software will be using less RAM.
RAM is RAM is still true, whether on Apple silicon or Intel.

The reason software/app developers use more RAM in their software designs is because in the Intel paradigm, there is a huge disparity in speed between RAM and SSD/disks. Since data is stored on SSD/disks, running operations on that data will be much faster if that data is copied to RAM. Thus, to allow a great user experience, more data on high speed RAM is the way to go.
Now, on apple silicon paradigm, the speed disparity is much much less. Thus, running operations on data stored in SSD/disks is a much more reasonable proposition than it is on Intel’s.
Software devs can then re-design the software to copy less data to RAM, and still be able to provide great user experience.

Less RAM is a decision from software devs. Apple simply made the decision making easier.
 

LonestarOne

macrumors 65816
Sep 13, 2019
1,074
1,426
McKinney, TX
I think most people understand RAM like its storage space that once its filled you need more. They don't understand RAM is loaded data, once its filled the computer can swap it out and bring new data in and swap back to the prior data

Yes. Virtual memory was invented for a reason.
 

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
I had to put my 2c in.

RAM usage is a function of software design, and not hardware design.
Apple transitioning into its own hardware design does not mean software will be using less RAM.
RAM is RAM is still true, whether on Apple silicon or Intel.

The reason software/app developers use more RAM in their software designs is because in the Intel paradigm, there is a huge disparity in speed between RAM and SSD/disks. Since data is stored on SSD/disks, running operations on that data will be much faster if that data is copied to RAM. Thus, to allow a great user experience, more data on high speed RAM is the way to go.
Now, on apple silicon paradigm, the speed disparity is much much less. Thus, running operations on data stored in SSD/disks is a much more reasonable proposition than it is on Intel’s.
Software devs can then re-design the software to copy less data to RAM, and still be able to provide great user experience.

Less RAM is a decision from software devs. Apple simply made the decision making easier.

The software didn't change though. Software companies just recompile the same software for Intel or Apple Silicon. Or users are running things through Rosetta.

Virtual Memory usage is a function of software design. Software actually doesn't see RAM - it sees Virtual Memory. It could be RAM, SSD or a combination of that. The hardware architecture and operating system determine page/swap efficiency and it looks like what they've done in the hardware and operating system is a lot better than what we see in Windows and macOS/Intel.
 

Beefie

macrumors newbie
Jan 14, 2019
7
5
The software didn't change though. Software companies just recompile the same software for Intel or Apple Silicon. Or users are running things through Rosetta.

Virtual Memory usage is a function of software design. Software actually doesn't see RAM - it sees Virtual Memory. It could be RAM, SSD or a combination of that. The hardware architecture and operating system determine page/swap efficiency and it looks like what they've done in the hardware and operating system is a lot better than what we see in Windows and macOS/Intel.
“Recompiling” is not similar to “redesign”

A “redesign” is needed.

Virtual memory is not decided by app developers. Virtual memory usage is managed by the OS. If apps aren’t requesting huge amounts of memory, OS wouldn’t bother with virtual memory.
 

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
“Recompiling” is not similar to “redesign”

A “redesign” is needed.

Virtual memory is not decided by app developers. Virtual memory usage is managed by the OS. If apps aren’t requesting huge amounts of memory, OS wouldn’t bother with virtual memory.

Recompiling is not design at all. You just flip a switch and get the binaries for a different operating system/hardware platform. The software engineer should not have to worry about RAM. That's a modern software paradigm. They allocate memory as they need it and free it when they are done with it. Though you don't even have to do that with some languages as you have automatic garbage collection.

The operating system provides you with virtual memory. It may page out some of your process even if there is plenty of free RAM. Your working set parameters may work to adjust how much memory you are using. Though I don't think that macOS has the concept of working set. At least not the client version. Windows certainly does.
 

Beefie

macrumors newbie
Jan 14, 2019
7
5
The software engineer should not have to worry about RAM. That's a modern software paradigm.
Which sadly, is the point im trying to make on redesigning software - because modern software paradigm assumes RAM is unlimited, when in reality it never is.
The modern software engineer does not care about RAM anymore. Which is why apps need more and more RAM each passing year.

Again, with architectural changes on apple silicon, the decision to keep requesting RAM shouldnt be the norm anymore. Because other I/O channels have much improved throughput.
Hardware folks have done their part in improving user experience. I think software folks should too.
 

pshufd

macrumors G4
Oct 24, 2013
10,151
14,574
New Hampshire
Which sadly, is the point im trying to make on redesigning software - because modern software paradigm assumes RAM is unlimited, when in reality it never is.
The modern software engineer does not care about RAM anymore. Which is why apps need more and more RAM each passing year.

Again, with architectural changes on apple silicon, the decision to keep requesting RAM shouldnt be the norm anymore. Because other I/O channels have much improved throughput.
Hardware folks have done their part in improving user experience. I think software folks should too.

In the really old days, we did overlays. Or program chaining. You programed program section overlays and brought in code before you needed to run it. Another approach was program chaining where you wrote out intermediate data to a file or buffer, dumped the current programs and fired off the next program.

Labor is expensive and hardware is cheap and getting cheaper so it makes a lot of sense to reduce software engineering costs to push the memory costs to the customer. This is similar to car companies replacing assemblies instead of replacing of repairing components. The obvious example is alternator brushes. You bring your car in and the mechanic says that they need to replace the alternator when only the brushes need replacing but the shop only does a full alternator replacement.

I applaud Apple's VM performance improvements. I don't think that software engineers are going to tighten up on VM usage.
 

Beefie

macrumors newbie
Jan 14, 2019
7
5
In the really old days, we did overlays. Or program chaining. You programed program section overlays and brought in code before you needed to run it. Another approach was program chaining where you wrote out intermediate data to a file or buffer, dumped the current programs and fired off the next program.

Labor is expensive and hardware is cheap and getting cheaper so it makes a lot of sense to reduce software engineering costs to push the memory costs to the customer. This is similar to car companies replacing assemblies instead of replacing of repairing components. The obvious example is alternator brushes. You bring your car in and the mechanic says that they need to replace the alternator when only the brushes need replacing but the shop only does a full alternator replacement.

I applaud Apple's VM performance improvements. I don't think that software engineers are going to tighten up on VM usage.
Indeed. In some industries, hardware is cheaper. And that notion is reinforced since hardware innovation is still happening at a great pace.

Some industries live on the fact that every cent saved on hardware is gold. And software folks suffer for it.

I’m sure Apple will do everything to make software development as easy and simple as possible, software design idealisms notwithstanding. They still need to expand their mac userbase.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.