Do you think the M4 series Macs will come standard with 12 GB of memory for base configuration?

[Gudi]

Your post also got me thinking of how exactly Apple will decide when to cut off software support for M1 Macs. Will they just unilaterally decide that all M1 hardware (regardless of spec) stop getting macOS updates beyond a certain date, or make some sort of exception for devices with more ram?

Historically evidence points to no exceptions for better specced out machines. At best M1 Pro and M1 Max chips will last a year longer than M1 machines from the same year regardless of RAM. They will likely retire the entire chip family altogether.

But because we just had a chip transition and Apple wants to grow its install base to build a more appealing hardware platform for software developers, there's a good chance that support for the M1 will last much longer than the usual 7 years.

It could well be that the relentless progress of Moore's Law will make people want to retire their M1 Macs before software upgrades stop.

 

[WC7]

I wonder if Apple will try to do away with random memory? Could on chip caches replace enough slower RAM plus a redesign of software? Are there any apps that run without need to resort to 'off cache' memory? Can we just have the SSD for the storage and the cache (or a bigger cache)?

 

[WC7]

I mean von Neumann and those early 'architects' of the future computers didn't directly consider the memory as distinct chunks at different levels ... just memory, storage, processing. Maybe it is Apple's intent to get rid of RAM and build huge caches onto the chip! (The M2 & M3 L2 performance core caches have been increased a little from the M1 12 MB to 16 MB). So maybe there is little reason to go further ... out of sequence instruction optimized?

 

[kiranmk2]

Amazon, eBay and MacRumours are not basic web pages. Amazon is full of dynamically loading elements that bring in promotions on related products. Macrumors forums will slowly allocate more and more memory and does not release it. It can easily grow to a couple of GB if you don’t close the tab. A lot of web pages are surprisingly heavy and inefficient and no one is motivated to streamline them. In fact the motivations are to tack on yet another ad. The result is that they become RAM hogs. fortunately browser data is relatively easy for the OS to swap to storage in reasonable sized chunks.

That's exactly the point though. Web pages are memory hogs and Apple can do nothing about that. A basic MBA (or M3 MBP) owner will probably think that having a few simple pages open (simple as in not running video, music or obvious programs) should be handled by their new $1199-1599 laptop. I'm not saying there is a need today to bump the RAM as 8 GB would fulfill my needs, but there is a real, out of Apple's control, danger that in a couple of years today's 8 GB machines will start to beachball when doing what the basic user considers to be basic tasks. At that point, they may well question why the need to spend over $1k on a laptop that only lasted 2-3 years when they could just spend $500 on a Dell Inspiron that would have more than enough power for their needs.

Reputation takes years/decades to build and months to loose.

 

[Chuckeee]

At this point, the M1 chip is 3 years old. So in this regard, it's not impossible that a base model M3 iMac may end up getting 3 more years of software updates compared to an M1 iMac, even if the latter does have more ram and storage on paper.

I was under the impression that Apple’s 5/7 year support calendar countdown starts when the product is no longer offered for sale on the Apple Store (not counting the refurbished offerings). It is NOT based on when a product is introduced. Since a MBA is still offered with an M1, the software expiration countdown has even started for the M1 chip yet.

 

[Abazigal]

I was under the impression that Apple’s 5/7 year support calendar countdown starts when the product is no longer offered for sale on the Apple Store (not counting the refurbished offerings). It is NOT based on when a product is introduced. Since a MBA is still offered with an M1, the software expiration countdown has even started for the M1 chip yet.

Your guess is as good as mine, since Apple will never publicly commit to X years of support for any product. I purchased my M1 MBA in 2020 under the assumption that it would get around 6-7 years of software updates (my 2012 MBA got like 8?) but will likely upgrade before that becomes an issue.

 

[picpicmac]

I wonder if Apple will try to do away with random memory? Could on chip caches replace enough slower RAM plus a redesign of software?

If you look at the M3 chips, the cache sizes can be noted. They are considerably (like a couple of orders of magnitude) less is capacity than the off-chip LpDDR5 chips.

This is a known hurdle in design.

The Max chips are already about as large as can be made today, which is why the Ultra is made from fusing two Max chips.

Requiring even just 8GB of RAM on the chip will grow it to be too large to make.

 

[WC7]

That's exactly the point though. Web pages are memory hogs and Apple can do nothing about that. A basic MBA (or M3 MBP) owner will probably think that having a few simple pages open (simple as in not running video, music or obvious programs) should be handled by their new $1199-1599 laptop. I'm not saying there is a need today to bump the RAM as 8 GB would fulfill my needs, but there is a real, out of Apple's control, danger that in a couple of years today's 8 GB machines will start to beachball when doing what the basic user considers to be basic tasks. At that point, they may well question why the need to spend over $1k on a laptop that only lasted 2-3 years when they could just spend $500 on a Dell Inspiron that would have more than enough power for their needs.

Reputation takes years/decades to build and months to loose.

I don't know if it prevents the memory hogging but I try to use the Reader mode when possible.

 

[WC7]

If you look at the M3 chips, the cache sizes can be noted. They are considerably (like a couple of orders of magnitude) less is capacity than the off-chip LpDDR5 chips.

This is a known hurdle in design.

The Max chips are already about as large as can be made today, which is why the Ultra is made from fusing two Max chips.

Requiring even just 8GB of RAM on the chip will grow it to be too large to make.

I was hoping Apple could do something with the future chiplet type designs ... maybe some intermediary chiplet memory? I guess that's just more complexity in timing between everything and no difference between that and the DDR on SoC.

 

[jdb8167]

I wonder if Apple will try to do away with random memory? Could on chip caches replace enough slower RAM plus a redesign of software? Are there any apps that run without need to resort to 'off cache' memory? Can we just have the SSD for the storage and the cache (or a bigger cache)?

Can't be done with current SSD speeds. RAM read/write is in the nanosecond range while SSD is in the 10-100 microsecond range with a lot of complication when it comes to writing (need to write a page at a time.)

There are some experimental non-volatile persistent storage technologies that work at near RAM speeds. HP was experimenting with memristors but as far as I know that didn't go anywhere. There are a lot of experiments with spintronics too. But none of these are very close to actual products as far as I can tell.

 

[Tagbert]

I wonder if Apple will try to do away with random memory? Could on chip caches replace enough slower RAM plus a redesign of software? Are there any apps that run without need to resort to 'off cache' memory? Can we just have the SSD for the storage and the cache (or a bigger cache)?

Apple may go the other direction and break up the single chip into multiple chiplets. Apple has recently said that packaging is an area of interest for them. AMD has done extensive work on chiplets. The benefit would be that they could optimize output and not all components would need to be on one large die. This makes massive chips like the Ultra easier to produce.

 

[jdb8167]

I was under the impression that Apple’s 5/7 year support calendar countdown starts when the product is no longer offered for sale on the Apple Store (not counting the refurbished offerings). It is NOT based on when a product is introduced. Since a MBA is still offered with an M1, the software expiration countdown has even started for the M1 chip yet.

Not always. Apple was still selling the 2013 Mac Pro (MacPro6,1) up until December 2019. They discontinued OS support with Ventura less than three years after that model was discontinued. Obviously, the long duration of the 2013 Mac Pro's availability makes this not particularly unreasonable but it is still significantly less than 5 years.

 

[Pressure]

The Max chips are already about as large as can be made today, which is why the Ultra is made from fusing two Max chips.

That is incorrect.

The M3 Max is not anywhere near the reticle limit of EUV lithography (maximum field size being 26 mm by 33 mm or 858 mm²).

I don't have the die size of the M3 Max but here is the M1 Max as a basis for the comparison. That's around half the size (50.35%) of the reticle limit.

 

[staypuftforums]

Nobody defended the hard drives! They were something entirely different than 8 GB of unified memory. It’s because of fast SSDs and Swap memory that available RAM size is no longer a hard constraint for performance.

Unfortunately, plenty of people on this website defended the spinners. More space compared to SSD, you see. Not everyone needs fast SSD speeds, you see. SSD are too expensive, you see.

A lot of the same arguments people are making to defend 8 GB in a $1,600 laptop. "If you need a SSD, then pay for the upgrade" was one I read quite a lot.

 

[Tagbert]

That is incorrect.

The M3 Max is not anywhere near the reticle limit of EUV lithography (maximum field size being 26 mm by 33 mm or 858 mm²).

I don't have the die size of the M3 Max but here is the M1 Max as a basis for the comparison. That's around half the size (50.35%) of the reticle limit.

View attachment 2317450

I think the thinking is that yield rates for chips that large in 3nm are lower than what TSMC and Apple want and that raises costs. Anything larger would likely have an even lower yield.