Will 16GB of Memory Finally Come Standard Alongside the M3 Chip?

[Stevenyo]

For the first half of the MacBook Pro's history, from 2006 until 2014, the base memory on Apple's premier laptop doubled five times, from an introductory 512 megabytes to a whopping eight gigabytes. In the second half of the MacBook Pro's history, from 2015 until today, the base memory has doubled...zero times. Even today, the base 13" MacBook Pro comes with the same 8 GB of memory that has been standard since 2014. Way back in April 2016, the last 4GB Mac notebook was discontinued. And yet, all these years later, the next big doubling of memory has never come.

That's not to say there haven't been improvements, Apple Silicon is a revelation, with those 8 GB of memory now integrated onto the chip and complemented by speedy SSDs. Nonetheless, this base RAM has grown too long in the tooth. Later this year, the iPhone 15 Pro is rumored to receive 8 GB of RAM, and its Mac siblings should take the hint that this standard is no longer acceptable in a full-fledged computer.

Apple should use the opportunity presented by M3 to increase their base memory to 16 GB. The M1 Pro, unveiled in late 2021, came alongside 16 GB of RAM as standard, in addition to a base storage of 512 GB. Two years and an innovative three-nanometer process later, the M3 will likely be competitive with the M1 Pro, if not running laps around it. These necessary upgrades will solidify Apple's standing in the computing space.

With the M3 Air launching later this year, Apple will have the choice as to whether keep the $999 M1 Air in their lineup or to discontinue it. If they let the M2 Air take its price, few will spend two or three hundred dollars extra simply to upgrade the specs of their chip. However, if this upgrade comes alongside an included increase in base RAM (and perhaps even more storage), the upgrade would be a no-brainer. Apple is all about price-ladders, and this one would make sense, pushing consumers towards the newest, greatest, and more-expensive model.

The earnings for the Mac sector this previous quarter were a jarring decline, and Apple is counting on a big Fall release of M3 to flip the script for the Mac in the year ahead. Though the M3 will undoubtedly be faster and more efficient, this overdue upgrade to RAM will push it from good to great. Let's see if we'll finally have the doubling we've been waiting for.

Having too little base RAM is part of the ladder seling strategy that helps apple be so insultingly profitable. Giving enough or upgradeable options in the base configuration would stop people from "climbing the ladder" and buying a device for many hundreds of dollars more than the advertised starting price. Giving 16/512 in the base MBA would mean 32/1TB in the base MBP (the 14", the 13" "Pro" doesn't count). Then you've got to convince people they really need 64GB to be able to sell an upcharge in the MBP, but the current "Pro" chip doesn't even support that, so the whole ladder would be ruined by the huge step up to M2 Max just to get a bit more RAM.

Someday the base number will climb, of course, but as computing needs aren't going up that quickly these days, I bet it's a long while before the entire Mac product line is redesigned around more RAM. Storage might be a different matter. With how insanely cheap SSDs have been lately, upping the storage capacities across the board for a marginal cost to Apple of like $30 and using it to justify a price increase several times that to starting prices could happen.

 

[Digitalguy]

I don't understand their Surface line either -- I'd LOVE a Windows tablet like the latest Surface Pro, plus the keyboard, but the specs are just plain bad. Older processors, not enough RAM, and *expensive*. Give me 32G RAM minimum, a big fast SSD, and a modern processor!

I owned a Surface Book a few years ago and it worked well for me, if a bit heavy, but I just can't think of paying their prices with those specs.

Surface Book is overpriced and I would have never bought the 15" at $2500, but I got the SB2 new (display model) at $1200 and the SB3 used, but in mint condition, at $1000. At those prices it's a heck of device.
Similarly I would have never paid full price for a maxed out 12" MacBook, but at $500 used it's a different story (and even then only the 2017 version with the improved keyboard and 16GB RAM).
Currently the Surface Laptop Studio with 32GB RAM 1TB SSD is still overpriced at $2000, although not as much as the Book was, but at $1500 it would be a different value proposition, and at $1000, even used if in mint condition, it would be a no brainer if you can make use of it's capabilities.

 

[Digitalguy]

It's also funny when people think they need a lot of horsepower because they drive long distances...

That aside, it's worth noting that you're making your assessment based on a Microsoft Surfacebook 3 running Windows. I don't know how one system compares to another, but they aren't the same-- we really can't draw any conclusions at all from your comments.

Because people confuse horsepower (CPU/GPU) with things like RAM. For instance, for what I do the the power of M1 is plenty, as is an 11th gen Intel chip or later or a mobile Ryzen 5000 or 6000. However even 16GB is tight (and 8GB is unusable).
And I do have Macs too. And my M1 Mini with 16GB is swapping more often than not (again, it's on for weeks), maybe less that Windows devices, but still enough that 16GB feels tight.

 

[kpluck]

Memory upgrades are a key component, especially on lower priced machines, of Apple's industry leading profit margins. I don't see them giving that up any time soon.

-kp

 

[CalMin]

Because people confuse horsepower (CPU/GPU) with things like RAM. For instance, for what I do the the power of M1 is plenty, as is an 11th gen Intel chip or later or a mobile Ryzen 5000 or 6000. However even 16GB is tight (and 8GB is unusable).
And I do have Macs too. And my M1 Mini with 16GB is swapping more often than not (again, it's on for weeks), maybe less that Windows devices, but still enough that 16GB feels tight.

I'd be interested in knowing what your workflow is. It seems more demanding than most given your statements about 16GB being tight and 8GB being unusable.

I have a 16GB M1 Pro and there is nothing that I can do that slows this bad boy down. In a typical day I use MS Word, Excel and PowerPoint, a Parallels Windows 11 VM, and a million browser windows. I have a 40GB Devonthink database which is open much of the time, and 60GB of emails in Outlook and Mac Mail. I run Fantastical, Messages, Notes, Apple Music, Acrobat, Stocks, Teams, Zoom and Webex. I have to use 3 different Edge Profiles simultaneously to manage 3 different Office 365 accounts for different work logins. I have 3 displays with at least 3-5 desktops open on each (sometimes more). At lunchtimes I sometime open up Logic or Final Cut Pro for a bit of fun stuff and use Pixelmator for photo editing (mostly iPhone images).

Nothing ever slows down or becomes unresponsive! Now, I do push memory pressure into the yellow or even red zones, but my measure of whether I have enough RAM or not, is whether or not my machine seems to be struggling or unresponsive. Other than when I had a memory leak with a beta version of Logitech Options (which was consuming over 20GB of ram!) I have never seen this machine be anything but buttery smooth to operate! Even then with a 20GB app in the background, I could only tell something was wrong because Mission Control got a little bit 'crunchy'. If I didn't know about memory pressure in activity monitor, I would never realize how much RAM was in use.

What really blows my mind is when I use my 8GB M2 Air. I don't have the VM, Logic or FCP on it, but I do have everything else and similarly it never gets choppy or unresponsive. This is the base M2 Air with the 'slow' single chip 256GB SSD...

You know, people worry about memory pressure and swapping leading to all sorts of bad things. Perhaps I have been lucky, but I have never worn out an SSD before the computer needed to be replaced.

I acknowledge that there are use cases for more RAM. Heavyweight multiple 4K/8K streams, huge Photoshop files, etc. but those people KNOW they need more RAM. General purpose computing is fine with 8GB and if you're doing a bit more then 16GB is perhaps a bit of extra peace of mind. But saying 16GB is tight and 8GB is unusable just isn't true for many categories of computer users.

 

[Digitalguy]

I'd be interested in knowing what your workflow is. It seems more demanding than most given your statements about 16GB being tight and 8GB being unusable.

I have a 16GB M1 Pro and there is nothing that I can do that slows this bad boy down. In a typical day I use MS Word, Excel and PowerPoint, a Parallels Windows 11 VM, and a million browser windows. I have a 40GB Devonthink database which is open much of the time, and 60GB of emails in Outlook and Mac Mail. I run Fantastical, Messages, Notes, Apple Music, Acrobat, Stocks, Teams, Zoom and Webex. I have to use 3 different Edge Profiles simultaneously to manage 3 different Office 365 accounts for different work logins. I have 3 displays with at least 3-5 desktops open on each (sometimes more). At lunchtimes I sometime open up Logic or Final Cut Pro for a bit of fun stuff and use Pixelmator for photo editing (mostly iPhone images).

Nothing ever slows down or becomes unresponsive! Now, I do push memory pressure into the yellow or even red zones, but my measure of whether I have enough RAM or not, is whether or not my machine seems to be struggling or unresponsive. Other than when I had a memory leak with a beta version of Logitech Options (which was consuming over 20GB of ram!) I have never seen this machine be anything but buttery smooth to operate! Even then with a 20GB app in the background, I could only tell something was wrong because Mission Control got a little bit 'crunchy'. If I didn't know about memory pressure in activity monitor, I would never realize how much RAM was in use.

What really blows my mind is when I use my 8GB M2 Air. I don't have the VM, Logic or FCP on it, but I do have everything else and similarly it never gets choppy or unresponsive. This is the base M2 Air with the 'slow' single chip 256GB SSD...

You know, people worry about memory pressure and swapping leading to all sorts of bad things. Perhaps I have been lucky, but I have never worn out an SSD before the computer needed to be replaced.

I acknowledge that there are use cases for more RAM. Heavyweight multiple 4K/8K streams, huge Photoshop files, etc. but those people KNOW they need more RAM. General purpose computing is fine with 8GB and if you're doing a bit more then 16GB is perhaps a bit of extra peace of mind. But saying 16GB is tight and 8GB is unusable just isn't true for many categories of computer users.

To be honest none of my 16GB machines has ever become unusable due to RAM, some minor slowdown when I push swap a lot, but nothing that makes the machine unusable. But precisely, I want to have peace of mind that I never need to get to that unresponsive state. Sometime the Parallels VM in my M1 Mini has become slow, but I don't know if it was only due to RAM or to other factors.
So yes, you are right you can push RAM usage into swap territorry and with currenty SSD speeds barely notice a difference, assuming you are not worried about wearing the SSD (I generally am not). But while I am regulary in swap terrotory I try to never push it too much. Maybe I should try and see how much I can push the machine with open stuff in the background (including Parallels which takes half the RAM) and see if the machine remains usable.
I will probably do a test with my M1 mini, having Chrome with plently of tabs + Safari + all my cloud and other apps + Windows with Edge and there again tons of tabs and see how it goes, especially after a couple of weeks of being in this state. It would be a very interesting experient.

 

[Analog Kid]

I wouldn't look too deep on timings, they weren't under stop watch. Basically, just observations of time it takes to paste 100gb of image files from one location to the other on the internal SSD.

If you do a copy on a APFS volume, then it doesn't require any data transfer, just some book keeping to point another file to the existing data. The speed benefit in that case is APFS, not the drive.

Your point was valid-- it would take a pretty large transfer to notice the difference in speeds.

Below is a graph showing Wired (Blue), Active (Red) and Compressed (Purple)

View attachment 2155114

I was thinking more of the Page Outs (at the bottom of the menu) like this:

This was just an example to test high paging load. I was having zero page outs until I launched a python script to allocate a 52500x52500 random matrix and invert it.

import numpy as np

print("Allocating...")
r=np.random.rand(52500,52500)
print("Inverting...")
i=np.linalg.inv(r)

The paging kicks in when the inversion starts.

In this case, the process consumes about twice my physical memory, so my swap file is approximately the size of my RAM. I'm running pretty close to the line here, because if I try to run a 60000 x 60000 matrix, the kernel kills the process for not having enough memory.

I spend a good fraction of the time with all cores at full load, so memory doesn't seem to be a bottle neck.

I'm able to multitask without much issue given that I'm running a background task with 100% CPU load. If I suspend the task, my memory use stays fixed but CPU load drops and everything is smooth as glass.

As you suggested, it's not the amount of memory (or swap) that matters it's the swap rate and CPU load.

 

[CalMin]

To be honest none of my 16GB machines has ever become unusable due to RAM, some minor slowdown when I push swap a lot, but nothing that makes the machine unusable. But precisely, I want to have peace of mind that I never need to get to that unresponsive state. Sometime the Parallels VM in my M1 Mini has become slow, but I don't know if it was only due to RAM or to other factors.
So yes, you are right you can push RAM usage into swap territorry and with currenty SSD speeds barely notice a difference, assuming you are not worried about wearing the SSD (I generally am not). But while I am regulary in swap terrotory I try to never push it too much. Maybe I should try and see how much I can push the machine with open stuff in the background (including Parallels which takes half the RAM) and see if the machine remains usable.
I will probably do a test with my M1 mini, having Chrome with plently of tabs + Safari + all my cloud and other apps + Windows with Edge and there again tons of tabs and see how it goes, especially after a couple of weeks of being in this state. It would be a very interesting experient.

Ok - then the statement about it being unusable was perhaps a bit strong.

Honestly though, I really feel like the memory pressure and swap stats in Activity Monitor get folks worried a lot without good reason. Especially those new to MacOs and Apple silicon. It's completely different than Windows or Intel Macs because the memory bandwidth with the SOC. It's not magic or anything, but the choke point is a lot higher before you experience (i.e. feel) the machine slowing down. Honestly, I think Activity Monitor should be renamed to Anxiety Monitor

By contrast, I bought prebuilt Windows gaming rig i7/16GB/3070 in December and it's my first Windows box in years. First thing I did was up the RAM to 32GB because moderately pushing that machine with 16GB made it FEEL slower. It got choppy real quick with what thought were moderate tasks and nothing like how hard I push my M1Pro!

I do wish these machines were upgradeable though. It's the cause of so much stress when you buy one because it's always a concern that may be you should have bought more RAM. I had horrible buyers remorse when I ordered my M1 Pro because I kept thinking I should have bought 32GB instead of 16GB. I pushed it so hard those two weeks trying to prove that I needed more RAM, and it was only once I had passed the return window that I calmed down about it!

 

[JouniS]

Parallels can slow down pretty much any Mac if you use it in a certain way. Basically, you have to make heavy use of data that resides outside the VM, and the size of the data should be comparable to RAM size or larger.

I'm not exactly sure what happens, but this is my best explanation. The VM does not read the files directly. Instead, Parallels memory-maps them and creates a virtual network interface. To the guest OS, the files appear to reside on a network drive. When you access the files, the guest OS caches them, which increases app memory usage on the macOS side. macOS also caches them, increasing RAM usage beyond total capacity. Because the RAM usage is mostly cached files, macOS thinks everything is fine and memory pressure remains green. Because the cached files have been used recently, macOS swaps out apps you haven't used in the past minute or so. When you switch back to those apps, you will notice the slowdown, assuming that their memory usage was high enough.

 

[Analog Kid]

Because people confuse horsepower (CPU/GPU) with things like RAM. For instance, for what I do the the power of M1 is plenty, as is an 11th gen Intel chip or later or a mobile Ryzen 5000 or 6000. However even 16GB is tight (and 8GB is unusable).
And I do have Macs too. And my M1 Mini with 16GB is swapping more often than not (again, it's on for weeks), maybe less that Windows devices, but still enough that 16GB feels tight.

What I mean by horsepower isn't CPU power, it's the confusion between "I use swap" and "I swap at high rates".

I keep coming back to metrics because "feels tight" is a subjective thing. Is it purely psychological or is there a physical manifestation that you can somehow measure. As I showed above, I can create a massive swap file and it has no impact on my active tasks. If I'm swapping constantly, and by constantly I mean tens of thousands or hundreds of thousands of swaps a second, then the process that is waiting for memory will slow (in my case the matrix inversion time will change drastically once I'm forced out of RAM).

Honestly though, I really feel like the memory pressure and swap stats in Activity Monitor get folks worried a lot without good reason. Especially those new to MacOs and Apple silicon. It's completely different than Windows or Intel Macs because the memory bandwidth with the SOC. It's not magic or anything, but the choke point is a lot higher before you experience (i.e. feel) the machine slowing down. Honestly, I think Activity Monitor should be renamed to Anxiety Monitor

Exactly. I'm not looking over people's shoulder to know, but I feel like people look and see that they're "out of memory" and therefore feel they need more physical RAM. The answer to that question isn't about how much swap you're using, it's about whether it's slowing a computation because an operand is on disk. It certainly isn't slowing the human down.

Someone mentioned 30-40 Chrome tabs. Ok, sure the system is going to push some of that out of active RAM because even if you looked at every tab in sequence for only a minute at a time, there's a ton of data you wouldn't access more than once every half hour. Access rates measured in hours or minutes or seconds aren't impacting performance.

I do wish these machines were upgradeable though. It's the cause of so much stress when you buy one because it's always a concern that may be you should have bought more RAM. I had horrible buyers remorse when I ordered my M1 Pro because I kept thinking I should have bought 32GB instead of 16GB. I pushed it so hard those two weeks trying to prove that I needed more RAM, and it was only once I had passed the return window that I calmed down about it!

The other way to look at it is that if it were upgradeable then you wouldn't have calmed down after the return window closed... Eventually you would buy the RAM because you wanted something to run faster and you'd convince yourself it worked, but I think what we're seeing in most of these systems is that it doesn't have much of an impact for most people and the people it does impact know that in advance.

I'm pretty sure @theorist9 wouldn't buy a 16GB machine thinking they'd upgrade it later. They do stuff that needs RAM and bought for that purpose.

 

[sam_dean]

The 2020 Developer Transition Kit that came out before the M1 had

- A12Z Bionic 8-core CPU SoC
- 16GB RAM
- 512GB SSD

 

[Toonartist]

If you do a copy on a APFS volume, then it doesn't require any data transfer, just some book keeping to point another file to the existing data. The speed benefit in that case is APFS, not the drive.

Your point was valid-- it would take a pretty large transfer to notice the difference in speeds.


I was thinking more of the Page Outs (at the bottom of the menu) like this:

View attachment 2155203
This was just an example to test high paging load. I was having zero page outs until I launched a python script to allocate a 52500x52500 random matrix and invert it.

import numpy as np

print("Allocating...")
r=np.random.rand(52500,52500)
print("Inverting...")
i=np.linalg.inv(r)

The paging kicks in when the inversion starts.

View attachment 2155198

In this case, the process consumes about twice my physical memory, so my swap file is approximately the size of my RAM. I'm running pretty close to the line here, because if I try to run a 60000 x 60000 matrix, the kernel kills the process for not having enough memory.

I spend a good fraction of the time with all cores at full load, so memory doesn't seem to be a bottle neck.

I'm able to multitask without much issue given that I'm running a background task with 100% CPU load. If I suspend the task, my memory use stays fixed but CPU load drops and everything is smooth as glass.

As you suggested, it's not the amount of memory (or swap) that matters it's the swap rate and CPU load.

That one is so small on the menu I'd completely overlooked it! Here you go. Different scale due to the larger page outs so looks a little more sparse compared to yours.

And here's the last 24hrs. You can see yesterday had few bigger spikes while working with 1gb DNG files in LR.

 

[dmccloud]

I don't understand their Surface line either -- I'd LOVE a Windows tablet like the latest Surface Pro, plus the keyboard, but the specs are just plain bad. Older processors, not enough RAM, and *expensive*. Give me 32G RAM minimum, a big fast SSD, and a modern processor!

I owned a Surface Book a few years ago and it worked well for me, if a bit heavy, but I just can't think of paying their prices with those specs.

To me, it's always felt like the Surface line in general was Microsoft's attempt to copy Apple with respect to designing hardware and software to work together. That being said, Microsoft's pricing structure for the Surface series is about the only thing that is even remotely close to Apple's approach.

 

[Digitalguy]

Ok - then the statement about it being unusable was perhaps a bit strong.

I used unusable for 8GB, not 16GB, you may want to check my posts (I said 16GB machines have never been unusable in my post, I did not mention 8GB ones). As for 8GB, I only have Intel machines as reference, and, what's more, dual core ones, and they can definitely become unresponsive if pushed (and it's hard to say how much is due to RAM and how much to the CPU). The only way to make my M1 Mac 8GB is to run Parallels (which takes 8GB). And in that context I have had the VM become unresponsive, but I haven't check how well MacOS was performance. That's why I said I should test it to see how much I can push the machine before it becomes noticeably slow (if ever).
Having said that, again, Parallels takes not only RAM but also half the cores, so it's still not apples to apples.

Honestly though, I really feel like the memory pressure and swap stats in Activity Monitor get folks worried a lot without good reason. Especially those new to MacOs and Apple silicon. It's completely different than Windows or Intel Macs because the memory bandwidth with the SOC. It's not magic or anything, but the choke point is a lot higher before you experience (i.e. feel) the machine slowing down. Honestly, I think Activity Monitor should be renamed to Anxiety Monitor

By contrast, I bought prebuilt Windows gaming rig i7/16GB/3070 in December and it's my first Windows box in years. First thing I did was up the RAM to 32GB because moderately pushing that machine with 16GB made it FEEL slower. It got choppy real quick with what thought were moderate tasks and nothing like how hard I push my M1Pro!

I do wish these machines were upgradeable though. It's the cause of so much stress when you buy one because it's always a concern that may be you should have bought more RAM. I had horrible buyers remorse when I ordered my M1 Pro because I kept thinking I should have bought 32GB instead of 16GB. I pushed it so hard those two weeks trying to prove that I needed more RAM, and it was only once I had passed the return window that I calmed down about it!

As far as RAM is concerned I have given up any hope of upgradable RAM on laptops. Not on miny PCs though, and that's why I have, instead, given up the idea of buying a M2 pro Mini or Mac Studio. So for desktop I will stay with Windows Mini PCs which are much more upgradable.
For laptops I just buy more RAM than what I currently need to have peace of mind. Again that makes me stay away from MacBooks because of the BTO model (expensive upgrades and no sales on upgraded versions) but at this point I am not buying Windows PCs either. I think I'll be on standby for a couple more years and see how things evolve.
When it comes to laptops I am particularly interested in ultraportables, so if Apples comes with an Apple Silicon device under 2 pounds and with cellular, I would buy an expensive BTO version of it.

 

[Silvestru Hosszu]

The 2020 Developer Transition Kit that came out before the M1 had

- A12Z Bionic 8-core CPU SoC
- 16GB RAM
- 512GB SSD

Exactly, it is a developer kit not a basic laptop for office warriors like me.

PS
Out of curiosity I yesterday started up an old base model 2013 MBA which collected dust in my office.
It came with 4gb's of ram and 128 ssd.
I loaded 3 word documents, opened safari with 5 relatively heavy pages and opened some pdf documents.
Was it perfectly smooth? No, it was not.
Was it unusable? Far from it.

This only proves that basic stuff does not need tons of RAM.

As I said somewhere else, I have 16 gb on my M2 PRO MBP, 32 on the M1 MAX and 64 on my windows desktop, but I bought these configs out of sheer spec snobbery.

 

[dgdosen]

- DRAM prices are about half of what they were in mid 2021.
- Apple execs were just talking about how cool it would be to have 96GB of ram
- Apple M2 chips have a minimum of 100GB/s throughput (up to 400)

I think it makes sense for Apple to push it and be a leader in this space. Why sell a WRX and slap Impreza tires on it?

MacroMicro

MacroMicro is committed to consolidating global economic data, while deploying technology to efficiently discover the clues to economic cycles. We believe that no investment decision should be taken without the consideration of economic fundamentals. MacroMicro is changing the way people invest...

en.macromicro.me

 

[leman]

- DRAM prices are about half of what they were in mid 2021.
- Apple execs were just talking about how cool it would be to have 96GB of ram
- Apple M2 chips have a minimum of 100GB/s throughput (up to 400)

I think it makes sense for Apple to push it and be a leader in this space. Why sell a WRX and slap Impreza tires on it?

MacroMicro

MacroMicro is committed to consolidating global economic data, while deploying technology to efficiently discover the clues to economic cycles. We believe that no investment decision should be taken without the consideration of economic fundamentals. MacroMicro is changing the way people invest...

en.macromicro.me

What’s the price of LPDDR5 6400 (plus custom packaging)?

 

[dgdosen]

What’s the price of LPDDR5 6400 (plus custom packaging)?

¯\_(ツ)_/¯

 

[Analog Kid]

The 2020 Developer Transition Kit that came out before the M1 had

- A12Z Bionic 8-core CPU SoC
- 16GB RAM
- 512GB SSD

MacroMicro

MacroMicro is committed to consolidating global economic data, while deploying technology to efficiently discover the clues to economic cycles. We believe that no investment decision should be taken without the consideration of economic fundamentals. MacroMicro is changing the way people invest...

en.macromicro.me

Pre-release dev kits, 4 and 8 gigabit DDR4 spot prices... All this random trivia is fun!

The Cray X-MP that came before the M1 had:

- 4 core processor
- 128MB RAM
- 38GB of disk storage
- $46M in 2023 dollars
- could not run Crysis

 

[sam_dean]

Pre-release dev kits, 4 and 8 gigabit DDR4 spot prices... All this random trivia is fun!

The Cray X-MP that came before the M1 had:

- 4 core processor
- 128MB RAM
- 38GB of disk storage
- $46M in 2023 dollars
- could not run Crysis

1982 mainframe 1500 nm process

 

[Analog Kid]

1982 mainframe 1500 nm process

With integrated sofa!

 

[leman]

¯\_(ツ)_/¯

Fair enough

It's just most of the time I see people talk about RAM getting cheaper they give data for ancient DDR4 2400 or something like that. DDR5 prices got much cheaper in 2022 (by almost 40%), but DDR5 is not LPDDR5 either. I mean, it doesn't change the fact that Apple RAM upgrade prices are arbitrary and not anchored to actual material costs, still might be interesting for the discussion to see how they relate to the underlaying prices.

 

[sam_dean]

Fair enough

It's just most of the time I see people talk about RAM getting cheaper they give data for ancient DDR4 2400 or something like that. DDR5 prices got much cheaper in 2022 (by almost 40%), but DDR5 is not LPDDR5 either. I mean, it doesn't change the fact that Apple RAM upgrade prices are arbitrary and not anchored to actual material costs, still might be interesting for the discussion to see how they relate to the underlaying prices.

Apple buys parts in bulk. The more units they buy the lower the per unit cost becomes.

Apple is transitioning all their devices to LPDDR5. When they suddenly make an exception just for their desktops, specifically the Mac Pro, then the price rises for all their devices.

As early as the 2006 iMac, Apple's first Intel iMac, used SODIMM RAM even when it had space for DIMMs. DIMMs were even cheaper.

That's why at 1st glance weird part decisions are made like putting an smartphone chip in a tablet/display/digital media player/microconsole or a laptop chip in a tablet.

(Preliminary results, shipments are in millions of units)

​

Devices​	iPhone​	Mac desktops (~20%)​	Mac laptops (~80%)​	iPad​	Total​	% of all Devices that are Desktops​
2021​	235.8​	5.58​	22.32​	57.8​	321.5​	1.74%​
2022​	226.4​	5.72​	22.88​	61.8​	316.8​	1.81%​
Total​	462.2​	11.3​	45.2​	119.6​	638.3​	1.77%​

Mac desktops covers

- Mac mini
- iMac
- Mac Studio
- Mac Pro

If I were to hazard a guess Mac Pros make up 1% of all Mac desktops sold? So they may not exceed 60,000/year? Guess that's why a Pro desktop with PCIe expansion slots was not refreshed from 2013-2018 and the past 3 years.

Why so low? Apple figured that >50% Mac Pro users do not want/need PCIe expansion slots.

 

[Analog Kid]

That's why at 1st glance weird part decisions are made like putting an smartphone chip in a tablet/display/digital media player/microconsole or a laptop chip in a tablet.

I‘d wager that it probably has more to do with maintainable software than component costs. People in this thread are obsessed with the cost of parts for some reason.

 

[sam_dean]

I‘d wager that it probably has more to do with maintainable software than component costs. People in this thread are obsessed with the cost of parts for some reason.

The cost of the Bill of Materials largely dictates MSRP.

Consumers prefer to pay the least and tend not ask how it was achieved.

That is why people look at the price of specific RAM & SSD chips Apple uses and their retail channel equivalents. I couldn't find them.

As for Apple chips Apple's challenge is where to put iPhone and Mac chips into when they have excess supply of said chips.

Like why no M1 Pro in a 2020 Mac mini or no M1 Ultra in a 2022 MBP 16" whose enclosure is based on a 2021 model.

M2 Pro in a 2023 Mac mini appears to be a solely parts supply issue as there is too much demand for the M1 Pro in a MBP 14"/16".

While no M1 Ultra in a MBP 16" appears to be lack of USB PD charging standard beyond 140W at the time of release and/or a parts supply issue as there is too much demand for it in a Mac Studio.

There are MR rumors and threads about a Mac Pro with one M2 Ultra or two M2 Ultra chips. Some think that those will be the SKUs of that 2023 model.

Currently the 2022 Mac Studio M1 Ultra is $4k while the 2019 Mac Pro starting price is $6k. A $2k difference just to have PCIe expansion slots, drive bays and RAM slots.

A Mac Studio with one M1 Max is $2k while a model with M1 Ultra (two M1 Max) is $4k. That's a $2k difference per M1 Max chip increase. This increase also includes more RAM, SSD, better I/O ports and a beefier HSF upgraded from cheaper aluminum to more conductive & heavier copper.

Using known Bill of Materials cost we can make estimates on what the 2023 Mac Pro will cost.

- $6k for Mac Pro with one M2 Ultra chip, 64GB RAM & 1TB SSD
- $10k for Mac Pro with two M2 Ultra chips, 128GB RAM & 2TB SSD