Why Apple does not list the CPU speed of Silicon Macs?

[MallardDuck]

That is a very good suggestion but...

1. I am considering a 16+ inch laptop that has no fan noise, no backlight bleed and no IPS glow. Too bad after waiting for several years, there seems to be no such Windows laptop.

2. In the Intel Mac era, in addition to Windows and Linux, I also developed things under MacOS. Right now those files are scattered across different drives. I don't feel comfortable with that and feel kind of missing something. I would like to consolidate all those files developed in the Intel Mac era in one machine.

3. Recently I fancy the idea of using a Mac with external storage as an ad hoc SMB file server and NAS. Is it better to do it under Mac OS rather than Windows?


Basically the following two posts summarize my dissatisfactions with Apple's direction under TC and Silicon Macs:

5 Things We Hate About Apple Silicon Macs

While the media is quick to praise Apple silicon Macs for their performance and efficiency, there are some things we really hate about them.

www.makeuseof.com

Under ‘It’s a closed Hardware platform’ of

Two reasons why I am skipping Apple Silicon Macs

And why you should think twice before buying Apple hardware

medium.com

I am also unhappy with Apple intentionally slowing down the SSD of base models of M2 MACs.

Well, if you're skipping Apple Silicon macs, you might as well give up. We're highly unlikely to ever see anything other than custom cores again.

The slowdown is an artifact of higher density chips, which help keep the cost down (that's relative to apple of course, which is overpriced). In practical terms for most users on the lower end machines, they're unlikely to see an impact in real-world usage. And you can avoid it by increasing storage so you have dual chips again.

As for file storage, just buy a NAS.

Apple has always been a walled garden - the intel/bootcamp era is the exception not the rule.

 

[leman]

Apple has always been a walled garden - the intel/bootcamp era is the exception not the rule.

Funnily enough Microsoft does officially support and endorse running Windows on Apple Silicon Macs (in a virtual environment). I don't think they ever offered such support in the Intel era. So I wouldn't even say that the Intel Macs was an exception — it was as much a walled garden as Mac ever was, it's just that it ran on hardware that was easier to use with another system. But that was always a poor idea anyway.

 

[theorist9]

Apple Silicon systems seem to have basically arrived at practical performance saturation, at least as far as the CPU goes. The really heavy work is handled by the GPU and other specialized hardware. There is just not much need for moar capability from the processor (except perhaps to serve you more ads more efficiently).

I've not yet made the transition to AS, and I'd be really curious to see if that's the case for routine office operations. But based on my experience with my 2019 i9 iMac, I'd expect it's not. That's a relatively fast machine, yet I experience repeated irritating delays of 0.5 s – 3 s for routine operations with Office apps—like clicking between cells in Excel—often accompanied by spinning beachbballs. AS would need to reduce these at least 10-fold for them to become imperceptible, yet I'd expect AS to be only about twice as fast as my i9 iMac for such operations.

Don't get me wrong—twice is a lot—but it's not enough to make such delays imperceptible. Of course, Excel really shouldn't take a couple of seconds when clicking between cells (it doesn't do this all the time, but it's frequent enough to be annoying). So better coding could also solve this. But Office is such complex software that this isn't going to change anytime soon.

Plus there's other heavier-duty office-type applications for which I know it's not the case that we've reached performance saturation. Suppose I have a flat PDF that I need to make searchable. You can do that using Adobe Acrobat Pro's Optical Character Reader—it uses text recognition to convert images to text, and works quite well. But it's single-threaded and takes a lot of CPU cycles: On my iMac, it takes ≈80 s to complete that for a 60-page PDF. I've had fellow posters try this on their AS machines, and it's ≈40 s.

Two-fold is an impressive speed-up, to be sure, but far from enough to reach performance saturation. For that, you'd need AS to be ≈20x faster than it is right now (which would reduce the time to ≈2 s) before you reach the point where user experience wouldn't be significantly improved. [Since this is not a repeated operation, ≈2 s would be fine.]

 

[hajime]

Funnily enough Microsoft does officially support and endorse running Windows on Apple Silicon Macs (in a virtual environment). I don't think they ever offered such support in the Intel era. So I wouldn't even say that the Intel Macs was an exception — it was as much a walled garden as Mac ever was, it's just that it ran on hardware that was easier to use with another system. But that was always a poor idea anyway.

Do MS, Apple and Parallels have some sort of under table deal?

Quite a lot people said that are satisfied with Parallels's performance but I found it slow and annoying.

 

[Sydde]

Suppose I have a flat PDF that I need to make searchable. You can do that using Adobe Acrobat Pro's Optical Character Reader—it uses text recognition to convert images to text, and works quite well. But it's single-threaded and takes a lot of CPU cycles: On my iMac, it takes ≈80 s to complete that for a 60-page PDF. I've had fellow posters try this on their AS machines, and it's ≈40 s.

Geekbench shows the M1 iMac having a SC score ~50% higher than the i9 iMac. This, I think, would not translate to getting the job done in half the time. Either Acrobat is using a multicore approach on M1, or it is making use of other SoC features (seems like the neural engine would be ideal for OCR, and iOS has built-in image character recognition, so M-series macOS probably does as well).

 

[theorist9]

Geekbench shows the M1 iMac having a SC score ~50% higher than the i9 iMac. This, I think, would not translate to getting the job done in half the time. Either Acrobat is using a multicore approach on M1, or it is making use of other SoC features (seems like the neural engine would be ideal for OCR, and iOS has built-in image character recognition, so M-series macOS probably does as well).

Those are certainly all possibilities. But it's actually not necessary to invoke any of them to explain the discrepancy between the 1.5x improvement reported by GB and the the 2x improvement seen for this task. Instead, such variation is expected even for CPU-only (no neural engine, etc.) SC tasks: Some CPU designs are simply better at certain tasks than others (an obvious example would be floating-point vs. integer calculations). Indeed, it's because of this variation that GB runs many different tasks to get an average CPU performance score.

Also note that, in this case, Intel and AS devices are running different ports of the same program. It could simply be those two factors (normal variation and different ports) that cause the discrepancy.

 

[xraydoc]

Quite a lot people said that are satisfied with Parallels's performance but I found it slow and annoying

On a fresh install of Windows 11 ARM, the performance for desktop apps (Office, etc.) is tolerable. I've not tried (nor require) anything more sophisticated on Windows.

 

[Sumo999]

I've not yet made the transition to AS, and I'd be really curious to see if that's the case for routine office operations. But based on my experience with my 2019 i9 iMac, I'd expect it's not. That's a relatively fast machine, yet I experience repeated irritating delays of 0.5 s – 3 s for routine operations with Office apps—like clicking between cells in Excel—often accompanied by spinning beachbballs. AS would need to reduce these at least 10-fold for them to become imperceptible, yet I'd expect AS to be only about twice as fast as my i9 iMac for such operations.

Don't get me wrong—twice is a lot—but it's not enough to make such delays imperceptible. Of course, Excel really shouldn't take a couple of seconds when clicking between cells (it doesn't do this all the time, but it's frequent enough to be annoying). So better coding could also solve this. But Office is such complex software that this isn't going to change anytime soon.

Plus there's other heavier-duty office-type applications for which I know it's not the case that we've reached performance saturation. Suppose I have a flat PDF that I need to make searchable. You can do that using Adobe Acrobat Pro's Optical Character Reader—it uses text recognition to convert images to text, and works quite well. But it's single-threaded and takes a lot of CPU cycles: On my iMac, it takes ≈80 s to complete that for a 60-page PDF. I've had fellow posters try this on their AS machines, and it's ≈40 s.

Two-fold is an impressive speed-up, to be sure, but far from enough to reach performance saturation. For that, you'd need AS to be ≈20x faster than it is right now (which would reduce the time to ≈2 s) before you reach the point where user experience wouldn't be significantly improved. [Since this is not a repeated operation, ≈2 s would be fine.]

I suggest you try out an Apple Silicon mac of any generation. Prepare to be amazed. You can say goodbye to beachballs. Hypothesizing what you "expect" to be 2x faster is no basis for judging actual real world performance. I work in Logic and my M1 handles everything I throw at it without a hiccup.

 

[theorist9]

I suggest you try out an Apple Silicon mac of any generation. Prepare to be amazed. You can say goodbye to beachballs. Hypothesizing "2x" is no basis for judging actual real world efficiency.

It's not enough to hear that someone doesn't experience spinning beachballs with AS Macs. What I'd like to hear is that someone whose workload is like mine, i.e., sufficient to generate routine spinning beachballs in Excel/Word/PowerPoint/Acrobat Pro with a 2019 or 2020 i9 iMac, found they went away with the M1 or M2.

If so, that would be good news. Though not quite enough to get me to switch yet, since I created my own benchmark with several representative Mathematica workloads, and asked posters with M1's to run it. I found they averaged only ≈ 15% faster than my i9 iMac—not enough of a difference to justify the nearly $7000 it would cost to replace my 128 GB iMac with a 128 GB Ultra Studio + 27" ASD. I'm waiting until they become faster, and until one can get larger RAM on lower-end machines (I don't need all the cores on an Ultra). See:
https://forums.macrumors.com/thread...es.2374458/page-3?post=31869605#post-31869605

 

[Sumo999]

It's not enough to hear that someone doesn't experience spinning beachballs with AS Macs. What I'd like to hear is that someone whose workload is like mine, i.e., sufficient to generate routine spinning beachballs in Excel/Word/PowerPoint/Acrobat Pro with a 2019 or 2020 i9 iMac, found they went away with the M1 or M2.

If so, that would be good news. Though not quite enough to get me to switch yet, since I created my own benchmark with several representative Mathematica workloads, and asked posters with M1's to run it. I found they averaged only ≈ 15% faster than my i9 iMac—not enough of a difference to justify the nearly $7000 it would cost to replace my 128 GB iMac with a 128 GB Ultra Studio + 27" ASD. I'm waiting until they become faster, and until one can get larger RAM on lower-end machines (I don't need all the cores on an Ultra). See:
https://forums.macrumors.com/thread...es.2374458/page-3?post=31869605#post-31869605

Your usage "workload" of routine data entry seems rather commonplace to me. Since you're simply repeating yourself, I'll do the same: try an M1/2 Mac for awhile and judge - rather than prejudge - for yourself. Logic sessions are exponentially more demanding of system resources than the apps you mention and the M1 never breaks a sweat.

 

[Gudi]

For some applications that don't take advantage of multi-core CPU, knowing the speed of the CPU is important for comparison on which one to buy.

That’s why Apple always gives you a delta versus their last generation chip and also the latest competitor. X percent faster is all you need to know and the only meaningful metric anyone can come up. A gigahertz number or even a Geekbench score say nothing on their own. Nobody knows how old the computer is you’re aiming to replace and which alternative options you’re looking at?

 

[theorist9]

Your usage "workload" of routine data entry seems rather commonplace to me. Since you're simply repeating yourself, I'll do the same: try an M1/2 Mac for awhile and judge - rather than prejudge - for yourself.

"Since you're simply repeating yourself, I'll do the same". Hey, I was trying to keep it friendly, so why the gratutious snark? And I wasn't repeating myself, I was adding new info....info. you chose to ignore or mischaracterize.

Anyways....

I don't have any way of trying it before purchasing one—which, as I said, is an ≈$7,000 investment to replace what I have now—so my only recourse is to do the best evaluation I can prior to purchase. That's why I did the work to create the Mathematica benchmarks, as that's the program that gives me the longest wait times (minutes to hours).

That's not prejudging, that's doing the work to collect actual data—which showed an average improvement of only about 15% for the kinds of tasks I run. "Prejudge" is inaccurate language on your part—it implies making a judgement without having a basis for it. I'm clearly trying to do the opposite, your mischaracterization notwithstanding.

The only one prejudging is you, such as when you said, without understanding what my workload actually is, that "Your usage 'workload' of routine data entry seems rather commonplace to me". You seem to be under the misimpression that all Excel/Word/Acrobat etc. tasks are the same. They're not, as evidenced by the fact that many using those programs even with Intel machines don't experience spinning beachballs. That means they're not using the programs the same way I am.

Thus the only way I can get a good prediction of what my experience would be like with these programs would be to find out what others who had the same issues as me with fast Intel iMacs found when they switched. Did you own a 2019 or 2020 i9 iMac and experience routine delays and spinning beachballs in Excel/Word/Acrobat?

 

[Sumo999]

I don't have any way of trying it before purchasing one—which, as I said, is nearly a $7,000 investment to replace what I have now—so my only recourse is to do the best evaluation I can prior to purchase. That's why I did the work to create the Mathematica benchmarks, as that's the program that gives me the longest wait times (minutes to hours).

That's not prejudging, that's doing the work to collect actual data—which showed an average improvement of only about 15% for the kinds of tasks I run. "Prejudge" is inaccurate language on your part--it implies making a judgement without having the data in hand. I'm clearly trying to do the opposite, your mischaracterization notwithstanding.

The only one prejudging is you, such as when you said, without understanding what my workload actually is, that "Your usage 'workload' of routine data entry seems rather commonplace to me". You seem to be under the misimpression that all Excel/Word/Acrobat etc. tasks are the same. They're not, as evidenced by the fact that many using those programs with Intel machines don't experience spinning beachballs. That means they're not using the programs the same way I am.

I get where you're coming from, I just imagined you are posting here for some alternative insight. Apple Silicon is a completely different beast, such that, as others have argued in this thread, older preconceptions are obsolete. In my opinion, given your skepticism, the only way to know for sure is hands-on, and prepare to be amazed.

For example, I have a Logic session of 140 tracks with dozens of plugins and CPU usage shows tons of overhead left unused when playing live (processing) in real time. This is on an M1 iMac (16GB ram, 2TB SSD)

Here's a screen shot of CPU usage while playing:



This would have been unthinkable on my Intel machine. I have a bunch of other apps open at the same time as well. GUI glitches are no longer a factor since they are completely nonexistent. There's no way in hell you need a maxed out Ultra for what you are doing. The end. I'll go have a more fruitful conversation with a brick wall.

 

[theorist9]

I just imagined you are posting here for some alternative insight. Apple Silicon is a completely different beast, such that, as others have argued in this thread, older preconceptions are obsolete. In my opinion, given your skepticism, the only way to know for sure is hands-on, and prepare to be amazed.

No, I was posting here to share my insight, and the data I've collected thus far on the performance difference between these two technologies. For instance, here's a screenshot from the link I posted earlier, showing the percent improvemement that an M1 Max gives over an i9 iMac on Mathematica on tasks representative of my workflow. I'm sure there are areas where the M1 trounces my i9, but this isn't one of them....clearly things are a lot more complicated than just "prepare to be amazed":

There's no way in hell you need a maxed out Ultra for what you are doing. The end.

Why is it that those who know the least tend to be the most vehement in their opinions? More specifically, why are there so many keyboard warriors on MR that presume to think they know our requirements better than we do ourselves? [And feel the need lecture us about it in very strong terms 😂!]

I was very specific on why I'd need a 128 GB Ultra to replace the current functionality of my iMac: It's because of the RAM, not the cores:

...the nearly $7000 it would cost to replace my 128 GB iMac with a 128 GB Ultra Studio + 27" ASD. I'm waiting until they become faster, and until one can get larger RAM on lower-end machines (I don't need all the cores on an Ultra).

Here's an example of how I use the 128 GB in my iMac. This is from a past Mathematica calculation:

And here's one I've actually been running since yesterday:

But please go ahead and continute to lecture me about how how you know my work better than do.

Need any help removing that egg from your face?

 

[badgerbadgerx2]

I have 128 Gb in 2020 iMac as well. My usage is different, Im all Adobe CC for print, not web, pushing some very large PSD files and 100 page magazines. ID< AI< PS< Acrobat < DW, multiple browsers all day -- and I hit swap often.

But *everything* I have read, watched, questioned, touched, etc, says M chips make 64 GBs look like 128. So would be OK with trying it. I feel ya, to replace my 2020 iMac is a $4000 spend. Studio display + 64/1TB Studio. (I only need a small internal) Jumping from a $3000 studio to a $5000 studo for 128 Gb ram doesnt seem like a good use of $.

I just think comparing Intel ram use to M ram use is not the same?

 

[Sumo999]

No, I was posting here to share my insight, and the data I've collected thus far on the performance difference between these two technologies. For instance, here's a screenshot from the link I posted earlier, showing the percent improvemement that an M1 Max gives over an i9 iMac on Mathematica on tasks representative of my workflow. I'm sure there are areas where the M1 trounces my i9, but this isn't one of them....clearly things are a lot more complicated than just "prepare to be amazed":

View attachment 2184043


Why is it that those who know the least tend to be the most vehement in their opinions? More specifically, why are there so many keyboard warriors on MR that presume to think they know our requirements better than we do ourselves? [And feel the need lecture us about it in very strong terms 😂!]

I was very specific on why I'd need a 128 GB Ultra to replace the current functionality of my iMac: It's because of the RAM, not the cores:


Here's an example of how I use the 128 GB in my iMac. This is from a past Mathematica calculation:

View attachment 2184041

And here's one I've actually been running since yesterday:
View attachment 2184042

But please go ahead and continute to lecture me about how how you know my work better than do.

Need any help removing that egg from your face?

So don't upgrade already, who cares? You have your mind made up about technology with which you have no experience so why bother us with retrograde trivia?

 

[theorist9]

So don't upgrade already, who cares? You have your mind made up about technology with which you have no experience so why bother us with retrograde trivia?

Who cares? Those who care about facts, that's who. You made some strikingly incorrect statements, and when I called you on them you refused to have the good grace to acknowledge you got things wrong.

Instead, you replied with a dishonest and pejorative mischaracterization of what I wrote ("you have your mind made up"..."retrograde trivia"), and then tried to have us believe that, since you can't win, the argument doesn't matter. Yeah, nice try. You're like the guy who's losing a chess match and angrily turns the board over rather than take the loss.

 

[theorist9]

I have 128 Gb in 2020 iMac as well. My usage is different, Im all Adobe CC for print, not web, pushing some very large PSD files and 100 page magazines. ID< AI< PS< Acrobat < DW, multiple browsers all day -- and I hit swap often.

But *everything* I have read, watched, questioned, touched, etc, says M chips make 64 GBs look like 128. So would be OK with trying it. I feel ya, to replace my 2020 iMac is a $4000 spend. Studio display + 64/1TB Studio. (I only need a small internal) Jumping from a $3000 studio to a $5000 studo for 128 Gb ram doesnt seem like a good use of $.

I just think comparing Intel ram use to M ram use is not the same?
View attachment 2184058

Mathematica needs a certain amount of RAM to do its calculations and that's not going to change under AS. And certainly if you need to create and operate on, say, an x GB array, AS isn't going to make that any smaller.

I expect that 128 GB will become available on the Max in the near future, obviating the need to buy an Ultra and thus making a 128 GB Mac a much less costly purchase. I expect that will be the case when I upgrade to AS, which will probably be in 2024 or so.

Specifically, while LPDDR5 is available in chip sizes up to 12 GB (which allows up to 96 GB in the M2 Max), LPDDR5x is available in 16 GB, which should allow up to 128 GB in future M# Max's.

And (speculating here), just as the max chip size of LPDDR5 went from 8 GB upon its initial release to its current 12 GB, we may also see a 50% increase in LPDDR5x chip sizes, to 24 GB. With that, you should be able to get a Pro with up to 96 GB and a Max with up to 192 GB (and 384 GB in the Ultra ).

 

[badgerbadgerx2]

I hope so. I plan on riding my 2020 for as long as I can. Running 2 extra 27" 4K screens, but rest of my needs are small, 1 TB internal only used for system, everything else goes on an external 4TB Thunderbolt drive, TMs to a 4 TB SSD, which CCCs to an 8TB GDD and with unlimited Backblaze Im set.

My other huge caveat is moving away from Catalina. New OSes look like candy crush to me. hate them. Cnat stand the wifes and kids M1s on Ventura. I have a dozen clients, and my workflow is a lot of moving around in Finder accessing files, Cat is far, far far cleaner. One click to access anything.

 

[okkibs]

But *everything* I have read, watched, questioned, touched, etc, says M chips make 64 GBs look like 128. So would be OK with trying it.

Do not fall for that, it's actually the opposite because at least on Intel Macs you had graphics cards with dedicated VRAM. On ASi Macs that unified memory means the GPU does not have dedicated VRAM and in VRAM heavy applications this will take away from the available RAM.

You have around 100GiB of application memory now, there is no magic to make that fit into 64GiB. It's as simple as that, if you need 100GiB on Intel, you'll probably be able to get by on 96GiB on M1/M2 with swap usage, but not with 64GiB.

I had a 16GiB Intel Macbook that I had to reboot every 48 hours or so when the swap would end up so big the device became unusable. Switched to a M1 Macbook when those came out with a 16GiB maximum, and the issue still persisted even though the Mac did run a couple days longer before stalling out on yellow/red memory pressure. Tried a 32GiB model as soon as they were available and on that the usage usually sat at around 20-25GiB without any intense workflows, which explains why 16GiB just wasn't enough. Now I am on a 64GiB model because unfortunately that was the limit for M1. (Can't justify spending that much money just for 32GiB extra for the rare occasions where my workflows could make use of it...)

Would not consider anything below a 96GiB M2 Max if I were to buy a new device in 2023 or later.

 

[leman]

Do not fall for that

You started well...

it's actually the opposite because at least on Intel Macs you had graphics cards with dedicated VRAM. On ASi Macs that unified memory means the GPU does not have dedicated VRAM and in VRAM heavy applications this will take away from the available RAM.

...but unfortunately ended with another urban myth... there is no evidence (that I know of) suggesting that unified memory architectures use more system RAM than systems with dedicated GPU RAM. The reason for that is quite simple: most of the time the driver or the application (or both) keep a copy of the data in the system RAM anyway. The only real issue with unified memory is that sometimes you end up with two copies — one for application and one for the driver.

 

[okkibs]

...but unfortunately ended with another urban myth... there is no evidence (that I know of) suggesting that unified memory architectures use more system RAM than systems with dedicated GPU RAM.

Okay, thank you for clearing that up - I was under the assumption that an application needs to explicitely support this and thus you shouldn't expect this to be default behaviour, but if that just works then there won't be an extra memory use. Then it will actually be more efficient for VRAM heavy workloads.

 

[Boil]

Specifically, while LPDDR5 is available in chip sizes up to 12 GB (which allows up to 96 GB in the M2 Max), LPDDR5x is available in 16 GB, which should allow up to 128 GB in future M# Max's.

And (speculating here), just as the max chip size of LPDDR5 went from 8 GB upon its initial release to its current 12 GB, we may also see a 50% increase in LPDDR5x chip sizes, to 24 GB. With that, you should be able to get a Pro with up to 96 GB and a Max with up to 192 GB (and 384 GB in the Ultra ).

LPDDR5X is supposed to scale up to 64GB chips, so theoretical maximums of:

• 64GB for Mn
• 128GB for Mn Pro
• 256GB for Mn Max
• 512GB for Mn Ultra
• 1TB for Mn Extreme

 

[leman]

Okay, thank you for clearing that up - I was under the assumption that an application needs to explicitely support this and thus you shouldn't expect this to be default behaviour, but if that just works then there won't be an extra memory use. Then it will actually be more efficient for VRAM heavy workloads.

At the end of the day, one has to measure. I can definitely imagine that for some applications that aggressively cache content in main memory you will end up with duplicate copies. But I don't think you will see this behaviour very often. Games usually keep a staging area buffer (which is not large) and stream assets from storage as needed. Of course, unified architectures give additional optimisation opportunities for that kind of stuff, but it's not mandatory.

And for VRAM heavy workloads it UMA be more efficient simply because you have more effective VRAM

 

[theorist9]

LPDDR5X is supposed to scale up to 64GB chips, so theoretical maximums of:

• 64GB for Mn
• 128GB for Mn Pro
• 256GB for Mn Max
• 512GB for Mn Ultra
• 1TB for Mn Extreme

I had read that, but didn't include it in my post because I wasn't sure if it would apply to Apple's custom memory packages, because LPDDR5x is currently limited to 16 GB chips, and the 64 GB figure would only be obtained by stacking four of those into a 64 GB package; I don't know if Apple's custom physical memory would allow for such stacking.

But hopefully that stacking could be adapted to Apple, as you've indicated in your post. It would be pretty nice to be able to get that kind of memory in Apple products.