ARM Longevity

[TFK.]

Since Apple announced the transition to ARM I had a question on my mind, which is:

Q: Will a specific ARM processor generation will become slow and outdated comparing it to the current year generation?
E.g., lets say that I bought the current M1 (2020) and compare it to the MX (2025) , well it have that big of a difference like comparing the iPhone A14 Chip (2020) to the A9 (2015), I may sound crazy but should we update our Macs like we update our iPhones every year or two (I know some will hold for longer but am talking about the mentality of the targeted user) to be up to date with the latest features and technology?

Lastly I want add that am writing this thread on my Mid 2012 MacBook Pro Retina that I use as my daily work laptop and my semi-professional photography (Photoshop & Lightroom) and I never had the reason to change it despite its age (nearly 10 years), I may got the temptation to buy the 16" MacBook Pro when it came out but when I saw the thermal issues I just held tight to my favorite obsolete MacBook Pro.

 

[leman]

I think that Apple chips became mature enough so that huge generation jumps won’t happen anymore, but I’d still expect about 20% performance increase every year.

If you want to have latest and greatest, you’d obviously have to update frequently, but that is already the case with Intel Macs anyway (to a certain degree). Your old computer won’t get any slower though. One caveat: M1, being the first generation of desktop Silicon, does have some limitations that will likely be overcome with subsequent chips.

 

[bennyprofane]

I think that Apple chips became mature enough so that huge generation jumps won’t happen anymore, but I’d still expect about 20% performance increase every year.

If you want to have latest and greatest, you’d obviously have to update frequently, but that is already the case with Intel Macs anyway (to a certain degree). Your old computer won’t get any slower though. One caveat: M1, being the first generation of desktop Silicon, does have some limitations that will likely be overcome with subsequent chips.

What are those limitations of M1?

 

[leman]

What are those limitations of M1?

Multi-monitor support, hardware ray tracing, some ARM ISA extensions like SVE that we are likely to see in second-gen chips...

 

[jdb8167]

What are those limitations of M1?

Mostly IO related. Only 2 external monitors on the M1 mini and only 1 on the M1 MacBooks. There seems to be an issue with some USB SSDs not getting full speed of 10 Gbps. Many issues with waking monitors from sleep. Those are the ones that I know of.

 

[bennyprofane]

Multi-monitor support

Mostly IO related. Only 2 external monitors on the M1 mini and only 1 on the M1 MacBooks. There seems to be an issue with some USB SSDs not getting full speed of 10 Gbps. Many issues with waking monitors from sleep. Those are the ones that I know of.

I thought the multi monitor support has been solved:

M1 Mac external display workaround walkthrough - 9to5Mac

M1 Macs have suport for just 1 or 2 external displays. Follow along for a M1 Mac external display workaround for using up to 6.

9to5mac.com

The monitor waking issues should be software related, no? Since some Intel users are also having these issues with Big Sur.

The USB speed issues are the most worrying thing to me so far. These video claims that these issues can be solved with better cables or hubs:

 

[za9ra22]

I think one pertinent factor is that so far, what we have are entry-level Apple silicone powered Macs, so one would assume they are limited in capability to some degree by intent.

One can also see how, given the significant performance gain they represent, that it won't take long for software developers to begin swamping them with ever-more complex apps, which in turn will demand increasing power from the next, and subsequent CPUs.

Not much different really than the 68k processors, PPCs, and Intels before them.

 

[velocityg4]

Generational increases will likely be less impressive over time. As TSMC and well everyone else hits the wall of process size. A silicon atom is about 0.2nm. So, the question is what's the minimum number of atoms necessary for a transistor? Not to mention cleaning issues. Like how pure can they make water for washing between manufacturing steps?

14nm was thought to be the wall ten years ago. Then five years ago it was proclaimed that the die shrink would stop in 2021 due to costs. I've read about a 1nm carbon nanotube transistor made in a lab. The question will be if that may be scaled up for mass production. Also costs of R&D and building fabs has increased dramatically. Which will likely get worse with each shrink.

Anyway's TSMC is well on the way for 4, 3 and even 2nm. So, there should be decent increases for the next few years. Also there's room in clock rate and core count. Given the current low TDP of the M1. I'd expect there'll be a bigger speed difference between a 2020 M1 and 2025 Mx than a 2025 Mx and 2030 My.

Unless TSMC or someone else makes more breakthroughs in what was already thought impossible sizes. We'll probably hit the limit of size by 2030. It may not be TSMC either. Intel held the crown for a long time before they finally faltered, TSMC is just as fallable. Intel and Samsung both have massive R&D budgets and loads of engineering talent. So, there are at least three companies in the ring for that crown of hitting the smallest size possible first. Although TSMC and Samsung are in the lead.

I thought the multi monitor support has been solved:

M1 Mac external display workaround walkthrough - 9to5Mac

M1 Macs have suport for just 1 or 2 external displays. Follow along for a M1 Mac external display workaround for using up to 6.

9to5mac.com

The monitor waking issues should be software related, no? Since some Intel users are also having these issues with Big Sur.

The USB speed issues are the most worrying thing to me so far. These video claims that these issues can be solved with better cables or hubs:

That involves using DisplayLink. Which is a very low end USB GPU. It doesn't fix the issue for the integrated graphics not handling more display. It's just a workaround.

 

[TFK.]

Generational increases will likely be less impressive over time. As TSMC and well everyone else hits the wall of process size. A silicon atom is about 0.2nm. So, the question is what's the minimum number of atoms necessary for a transistor? Not to mention cleaning issues. Like how pure can they make water for washing between manufacturing steps?…

Thank you for your explanation I kinda see where this’s going, hopefully we could see major breakthroughs in the coming years.

 

[jdb8167]

I thought the multi monitor support has been solved:

M1 Mac external display workaround walkthrough - 9to5Mac

M1 Macs have suport for just 1 or 2 external displays. Follow along for a M1 Mac external display workaround for using up to 6.

9to5mac.com

The monitor waking issues should be software related, no? Since some Intel users are also having these issues with Big Sur.

The USB speed issues are the most worrying thing to me so far. These video claims that these issues can be solved with better cables or hubs:

I’m fairly certain that the monitor wake from sleep is purely software. My experience was that even when the monitor wouldn’t wake, I could get it to wake up consistently from another Mac and a remote connection to the M1 MacBook Air. Until it is resolved though, there could be a hardware problem that others are seeing.

As others have already pointed out, using additional monitors via DisplayLink is a workaround that can’t replace a real connection to the GPU.

Fixing the USB problems with a dock can work if the dock is Thunderbolt or USB4 then it is up to the dock to perform to the USB 3 speeds. Still a problem for the M1. It might also be a software issue. Luckily Thunderbolt/USB4 40 GBPs seems to work fine.

 

[cmaier]

Generational increases will likely be less impressive over time. As TSMC and well everyone else hits the wall of process size. A silicon atom is about 0.2nm. So, the question is what's the minimum number of atoms necessary for a transistor?

Transistor sizes and node size are not particularly well correlated. We are a very long way from having to worry about the number of atoms, and there are solutions even for when the gate size gets close to the atomic limit - for example you can get rid of gates, and switch to bipolar transistors.

 

[Yebubbleman]

Since Apple announced the transition to ARM I had a question on my mind, which is:

Q: Will a specific ARM processor generation will become slow and outdated comparing it to the current year generation?
E.g., lets say that I bought the current M1 (2020) and compare it to the MX (2025) , well it have that big of a difference like comparing the iPhone A14 Chip (2020) to the A9 (2015), I may sound crazy but should we update our Macs like we update our iPhones every year or two (I know some will hold for longer but am talking about the mentality of the targeted user) to be up to date with the latest features and technology?

Lastly I want add that am writing this thread on my Mid 2012 MacBook Pro Retina that I use as my daily work laptop and my semi-professional photography (Photoshop & Lightroom) and I never had the reason to change it despite its age (nearly 10 years), I may got the temptation to buy the 16" MacBook Pro when it came out but when I saw the thermal issues I just held tight to my favorite obsolete MacBook Pro.

Will you see massive speed gains year over year compared to that of the Intel Macs introduced post-Haswell (2015-2020)? Yes.

Will your computer be effectively unusable after 5-7 years of ownership in the way that iOS devices and iPads often have been? No. I'd imagine that Apple is going to do with Apple Silicon Macs what they've been doing with Intel Macs, which is to say that they'll likely be able to run current macOS releases for 7-8 years before being left out, and that Apple will continue to support each macOS version with security updates until it is older than two versions older than the current release. This would effectively make your Mac able to be viable for 9-10 years before it becomes left out of modern day software support (though some apps tend to support even older macOS releases, so your mileage may vary).

The only downside to this versus what you might get with an Intel Mac is that your Mid 2012 Retina MacBook Pro can't run the current macOS release (and is effectively capped on macOS Catalina, which is supported for another year and a half), but it CAN run the latest Windows 10 release (or, at the very least, run the current Long Term Servicing Channel Release of Windows 10 Enterprise, which is supported until 2029). The downside to there no longer being Intel, let alone a native boot solution that allows some form of Windows 10 to be installed on the Mac is that when your Mac is no longer supported by Apple, it won't be able to live a second life as a viable PC.

 

[guzhogi]

Difficult to say how quickly M# chips will become obsolete. I see two factors in this: actual speed increases (# GHz, giga/teraflop performance, etc.), and new features (like support for additional monitors). We might be able to extrapolate speed increases from Apple's much longer history of A# chips. Timeline on new features will be iffy. As there's a difference between iPhone/iPad chips and Mac chips, and we're still on the first generation of the M# line, I'm going to wait and see what Apple does. Apple could release new Macs every year like iPhones, maybe every other year, or some other regular period. Apple could release new computers at irregular intervals, whenever they want. Who knows? I don't see how we have enough information to properly make an accurate prediction

 

[Bodhitree]

Apple did say one of their main reasons for moving away from Intel was that Intel was no longer delivering on expected yearly performance increases. That said, I think the time of 50% year-on-year gains is now gone, and we are looking at more like 20%. So in 5 years your new M1 machine will be roughly half the speed of a new machine. Which matters to professionals who can never have enough, like in 3D rendering.

For the rest of us, I think that a machine which is fast enough to do everything we need is already here, and that’s the M1. The most demanding tasks are likely to be a bit of video editing or some large spreadsheets. Honestly I thought we reached this point some years ago, but Apple have proved with the M1 that there is a substantial appetite for more responsive systems.

The main limiting factor for the longevity of a new machine is the length of software support, not how fast the hardware is. If you get 6 years of macOS updates and then two or three of security updates, then afterwards you are taking chances on how well you are able to cope with a drive-by virus encounter, App Store compatibility and so on.

One thing to take note of is the new ARM v9 instruction set, which looks like it will arrive soon, but not in time for the first few generations of Apple Silicon. That may cause machines to become obsolete somewhat earlier than the 8-10 years that the Intel machines got support for.

 

[geejay9876]

Until Intel Sandy Bridge / 2011 appeared, I had a constant urge to upgrade my computers because the old machine seemed to be too slow.

Since then my interest in new machines had totally died because the speed improvements were negligible. Any Sandy Bridge computer is totally adequate for 90% of users today.

When the M1 appeared my "must-have" urge came back and I had to buy one. The M1 is a big leap in performance.

I do very much hope that ARM will start a new race between PC manufacturers for speed and I will gladly upgrade every 2-3 years again.

 

[velocityg4]

One thing to take note of is the new ARM v9 instruction set, which looks like it will arrive soon, but not in time for the first few generations of Apple Silicon. That may cause machines to become obsolete somewhat earlier than the 8-10 years that the Intel machines got support for.

Yea, that is a concern now. Sort of like the 32 to 64 bit transition for Core to Core 2 really killed the first gen Intel Mac support life. Then came the models with 32-bit EFI. But since v9 was just announced. I can't see that landing in any of the first gen lines coming out in 2021 or 2022. That should take a while to incorporate in new CPU designs. Usually any new industry standard takes manufacturers a few years to incorporate.

It's actually surprising Apple added USB 4 so fast. Probably because they were among those who helped develop the standard. They were working with prototype designs and just had to make minor changes for the finished product. Then again maybe that's why there's so many user reports of them being buggy.

 

[guzhogi]

Until Intel Sandy Bridge / 2011 appeared, I had a constant urge to upgrade my computers because the old machine seemed to be too slow.

Since then my interest in new machines had totally died because the speed improvements were negligible. Any Sandy Bridge computer is totally adequate for 90% of users today.

When the M1 appeared my "must-have" urge came back and I had to buy one. The M1 is a big leap in performance.

I do very much hope that ARM will start a new race between PC manufacturers for speed and I will gladly upgrade every 2-3 years again.

Same position. My first computer was a Macintosh LC, then a beige Power Mac G3, followed by a iBook G3, and then a MacBook Pro (I forget which year, but before 2013). I was always surprised by the speed improvements. Then the speed improvements slowed down. Right now, still running a late 2013 MBP. It's showing its age with some games, but still works pretty decently.

 

[Bodhitree]

Any Sandy Bridge computer is totally adequate for 90% of users today.

That’s precisely what motivated me to buy my 2011 MacBook Air. That machine has a Sandy Bridge processor and came with SSD storage, which means it still feels snappy today. I’m really surprised that Apple waited so long to make SSD standard across the lineup, it makes that big a difference to responsiveness of the system as a whole.

But the M1 did tempt me, severely. I’m due some upgrades, and it’s just a question whether I want a new laptop or a large screen on a desk. So I’m waiting to see what the new iMacs will be like before making a decision.

 

[trevpimp]

If you compare Apple Silicon to the majority of computers with intel then Apple is way ahead of the future. Arm processors are practically future proof for the moment and it will take a most amount of time for other computers to even come close to the speeds found in M1 computers. I think M1 will maintain leverage for quite some time taking years for other computers to catch up

 

[Significant1]

If you compare Apple Silicon to the majority of computers with intel then Apple is way ahead of the future. Arm processors are practically future proof for the moment and it will take a most amount of time for other computers to even come close to the speeds found in M1 computers. I think M1 will maintain leverage for quite some time taking years for other computers to catch up

I just worry that when Apple very soon adopts armv9 my m1 will be like my core duo when core 2 duo arrived with 64 bit support.

 

[robco74]

I just worry that when Apple very soon adopts armv9 my m1 will be like my core duo when core 2 duo arrived with 64 bit support.

There was still plenty of support moving from armv7 to armv8. The iPad 2 was supported from iOS 4 through iOS 9.

 

[rui no onna]

Until Intel Sandy Bridge / 2011 appeared, I had a constant urge to upgrade my computers because the old machine seemed to be too slow.

Since then my interest in new machines had totally died because the speed improvements were negligible. Any Sandy Bridge computer is totally adequate for 90% of users today.

I thought Windows 10 didn't support Sandy Brige? That said, I agree about its performance.

Personally, I was on Ivy Bridge and Haswell and I upgraded partly to downsize from an 11-liter Mini-ITX build to a teeny tiny NUC with 4K HEVC HW acceleration and USB-C/Thunderbolt.

 

[velocityg4]

I thought Windows 10 didn't support Sandy Brige? That said, I agree about its performance.

Personally, I was on Ivy Bridge and Haswell and I upgraded partly to downsize from an 11-liter Mini-ITX build to a teeny tiny NUC with 4K HEVC HW acceleration and USB-C/Thunderbolt.

Windows 10 should work way back to a Core Duo. I know there are some issues with the Pentium 4 but I think you can get those to work. Heck people probably got it working on the Pentium 2/3.

Windows is very forgiving on old hardware. MS would probably make it run on a 4004 if it was possible.

 

[rui no onna]

Windows 10 should work way back to a Core Duo. I know there are some issues with the Pentium 4 but I think you can get those to work. Heck people probably got it working on the Pentium 2/3.

Windows is very forgiving on old hardware. MS would probably make it run on a 4004 if it was possible.

Not performance-wise but driver support. Granted, this probably has more to do with the graphics drivers.

Does My Intel® Processor Support Microsoft Windows® 10?

Explains how to tell if your Intel® Processor is compatible with Microsoft Windows® 10.

www.intel.com

 

[Joelist]

ARM has little to do with M1 performance outside of the ISA having consistently sized ops - the big drivers are the die size and (moreso) the microarchitecture. The Firestorm cores in M1 and A14 are not only faster than Intel/AMD cores but also faster by a good deal than the rest of the ARM world. Having 8 decoders and being able to push 7+ instructions per core simultaneously is a big driver.

So Apple could sit on the same version of ARM or even fork off of it into their own ISA and still deliver the performance. Now forking could have issues in the world of software compatibility but in theory they could do it.