Prepare to have your minds blown in 2021 with Apple's new computers. Here's why...

[smoking monkey]

Remember this chart from WWDC? Well that very lowest line on Notebooks is the MBA M1. Think about that. The amazingly, stonkingly powerful MBA M1. We now have absolutely no reason to doubt the legitimacy of this chart. What we didn't know back then was the performance axis on this chart didn't start at ZERO, it started at 11! So with this chart being true then what we are going to see released next year and in 2022 will be absolutely mind blowing. We can barely comprehend the power packed into this M1 MBA, imagine what's coming...

 

[aleni]

I don’t need to wait until next year, I’m already blown away with the current base air.

 

[Piggie]

I have thought about this, and you may be right, however I feel it's easy to make a GIANT mistake here.

Apple started with their 1st ARM based chip, which as their knowledge of design, dia shrinking FAB improvements, and also of course improvements coming from ARM themselves have lead to massive improvements from the 1st ARM based Apple chip to todays A14 version.
And therefore, this A14 chip, with some additions becoming fast enough to finally power the new 3 M1 machines.

However, and here is where I feel we need to be careful about our expectations.
This M1 chip is at the very top of many many many years of technical evolution, in Chip design, Fab ability and Arm advances.
It's NOT Apple's 1st attempt at a chip, in the same way their actual 1st iPhone chip was.

I don't think you can look at the GIANT LEAPS over the past decade that happened over the lifetime, so far up till today, and think the M1 chip is the same as the 1st iPhone ARM chip and hence the M1 will mirror the same jumps the iPhone chips historically did.

That's not of course saying I'm not expecting jumps in performance as more CPU and GPU cores are added, and the FAB's can go from 5nm to 3nm and lower.

If you look back at almost any tech, it's always in the early days that the leaps happen.
Just like Intel used to have, when you went from 33mhz 486, then to 66Mhz, then Pentiums, with 90, then 133, and so on and on.
I'm sure Apple's Arm has been through this same path.

Please note: I'd love to be shocked, and every year the M1 chip jumps up and perhaps doubles in power.
I will admit however, I'm not expecting M1 to be an evolutionary mirror of the iphone A chip advancements.
I feel that is an unrealistic expectation.

You won't see, for example MacBook Air performance doubling every year as the M2, M3, M4 come out.
Whilst that's a lovely thought. As I said, the new M1 is in reality 14 versions in now, and not at version 1 really.

I can't find a A14 chart, so this chart below is rather out of date and would be much higher now, however, as said above I don't feel you can realistically put the M1 at the bottom left of a chart like this and expect the same increase.

The M1, M2, M3 etc are going to have an amazing future I'm certain of it.
And I'm so excited by this change, and REALLY hope Apple does not screw it up, by making amazing machines on one hand and also ruining some aspects of what they COULD potentially do amazingly well in other areas (Gaming?)

It will be super interesting to see what changes Apple makes to the A15, for example to make it even more suited to the desktop. The A14 is basically still just a very low power mobile chip.
We're a long way, it seems from a real desktop chip.

 

[Chilluminati91]

So wait, it didnt start at 0 unknown value, but at 11 unknown value?
If we take this graph literally we can see that the m1 notebooks have already reached peak performance, as the
lowest performing m1 desktop (first mac mini) is only marginally faster than the top notebook.

 

[smoking monkey]

So wait, it didnt start at 0 unknown value, but at 11 unknown value?
If we take this graph literally we can see that the m1 notebooks have already reached peak performance, as the
lowest performing m1 desktop (first mac mini) is only marginally faster than the top notebook.

Ha! I see what you did there. Very smart!
And yes, 11. One higher than 10 of course. The video will explain all...

(Starts at 1:18)

 

[Piggie]

So wait, it didnt start at 0 unknown value, but at 11 unknown value?
If we take this graph literally we can see that the m1 notebooks have already reached peak performance, as the
lowest performing m1 desktop (first mac mini) is only marginally faster than the top notebook.

Well it was the best chart I could find at the time to give a simple view of speed improvements over the years

I just feel we have 'some people' who are making the (I think) mistake of viewing the M1 chip, in the same was they may be remembering Apple's 1st ARM based chip, and expecting the same leaps year over year.
As opposed to perhaps understanding the A14 is the current pinnacle of Apple's chip design, and we all know, at version 14 to version 15 or 15 to 16, or 16 to 17, you can't expect the same leaps as you did from version 1 to 2 or 3 to 4.

Apple's main benefit now is they don't need to worry about power so much.
Who wants a high end iMac on their desk running a 10w mobile chip?
Or rather, who wants a chip constrained to being such low power?

Apple could throw 100w at a desktop chip (I'm sure they won't due to heat)
But now they can.
They can of course also add CPU and GPU cores.

It's certainly going to be interesting to see, over the next 5 to 10 years what improvements can be made to the current A14 / M1 chip we have today, to get it to grow, and to see is they can step over limitations that AMD and Intel have hit and can't seem to get past in the forthcoming years.

 

[David Hassholehoff]

So wait, it didnt start at 0 unknown value, but at 11 unknown value?
If we take this graph literally we can see that the m1 notebooks have already reached peak performance, as the
lowest performing m1 desktop (first mac mini) is only marginally faster than the top notebook.

Precisely this. Charts without units are borderline meaningless. They are barely okay when you are illustrating something on a napkin over lunch.

 

[Pressure]

Just look at this image from AnandTech. M1 reaches 69.4 in SPECint2006.

 

[smoking monkey]

Well it was the best chart I could find at the time to give a simple view of speed improvements over the years

Think the poster was referring to the OP Graph and my joke about it starting at 11!

Also, I was just referring to the 2 year release update schedule. Not further into the future. I'd imagine that after the initial release things might slow down somewhat because all we can reference at the moment is Intel. Whereas in a few years we'll be comparing 2G M chips with these first ones.

Precisely this. Charts without units are borderline meaningless. They are barely okay when you are illustrating something on a napkin over lunch.

Perhaps you've missed the point of the OP.

 

[Piggie]

Now I know, very sadly, Apple has never been THAT interested in gaming.
And I will say without question that 100% of the reason for this is Steve Jobs.
I'll state it as a fact that if Steve Loved Gaming, and had gaming in his DNA then Apple today would be far different.
But that was not what happened.

However, there is no reason for the past to enshrine the future in stone.
Apple could, now it's free to do as it wishes enhance greatly the abilities of future machines in this area.
And we don't have to call it gaming, as some people don't like that.

Simply just graphical ability, features and speed.

So if we look at the new PS5 and Xbox with their both AMD designed SOC's
Is there any reason to think Apple would be unable? or unwilling to be able to achieve what AMD has been able to achieve for both these devices as time moves on?

Xbox Series X GPU deep dive: 4K 120Hz, 8K, GDDR6, VRR, 12 TFLOPs beast

Microsoft's next-gen Xbox Series X APU: 8-core/16-thread Zen 2 CPU joined by 52 cores from RDNA 2 GPU, all on TSMC's 7nm node.

www.tweaktown.com

Apple's M1 chip (not really laid out as their promo images shows)

The amount of space dedicated to the GPU/s is the most obvious difference between these two designs.

I can't see Apple EVER in a million years dedicating so much area on an future M series chip for the GPU/s
However, perhaps with future designs/ideas they could achieve the same graphical performance in other ways?

Just guessing here

 

[djlythium]

Just look at this image from AnandTech. M1 reaches 69.4 in SPECint2006.

View attachment 1676487

So, it’s nearly off the chart. 😧

 

[Pressure]

Now I know, very sadly, Apple has never been THAT interested in gaming.
And I will say without question that 100% of the reason for this is Steve Jobs.
I'll state it as a fact that if Steve Loved Gaming, and had gaming in his DNA then Apple today would be far different.
But that was not what happened.

However, there is no reason for the past to enshrine the future in stone.
Apple could, now it's free to do as it wishes enhance greatly the abilities of future machines in this area.
And we don't have to call it gaming, as some people don't like that.

Simply just graphical ability, features and speed.

So if we look at the new PS5 and Xbox with their both AMD designed SOC's
Is there any reason to think Apple would be unable? or unwilling to be able to achieve what AMD has been able to achieve for both these devices as time moves on?

Xbox Series X GPU deep dive: 4K 120Hz, 8K, GDDR6, VRR, 12 TFLOPs beast

Microsoft's next-gen Xbox Series X APU: 8-core/16-thread Zen 2 CPU joined by 52 cores from RDNA 2 GPU, all on TSMC's 7nm node.

www.tweaktown.com

Apple's M1 chip (not really laid out as their promo images shows)

The amount of space dedicated to the GPU/s is the most obvious difference between these two designs.

I can't see Apple EVER in a million years dedicating so much area on an future M series chip for the GPU/s
However, perhaps with future designs/ideas they could achieve the same graphical performance in other ways?

Just guessing here

They don’t need to. A lot of things can be done with other resources and coprocessors, like machine learning using the 16-core neural engine instead of brute forcing it with the GPU. Encoding and decoding video has dedicated hardware exposed for NLEs (video-editing).

Are you primarily concerned about gaming?

 

[Strangedream]

Very nice chart, love the colors - but utterly useless. What is the y-axis supposed to be? Banana/seconds/alcohol unit?
And compared to what?

What would be really nice is to see the perf/consumption of say an intel iMac Pro vs an M1 MBA.

 

[JohnnyGo]

I have thought about this, and you may be right, however I feel it's easy to make a GIANT mistake here.

Apple started with their 1st ARM based chip, which as their knowledge of design, dia shrinking FAB improvements, and also of course improvements coming from ARM themselves have lead to massive improvements from the 1st ARM based Apple chip to todays A14 version.
And therefore, this A14 chip, with some additions becoming fast enough to finally power the new 3 M1 machines.

However, and here is where I feel we need to be careful about our expectations.
This M1 chip is at the very top of many many many years of technical evolution, in Chip design, Fab ability and Arm advances.
It's NOT Apple's 1st attempt at a chip, in the same way their actual 1st iPhone chip was.

I don't think you can look at the GIANT LEAPS over the past decade that happened over the lifetime, so far up till today, and think the M1 chip is the same as the 1st iPhone ARM chip and hence the M1 will mirror the same jumps the iPhone chips historically did.

That's not of course saying I'm not expecting jumps in performance as more CPU and GPU cores are added, and the FAB's can go from 5nm to 3nm and lower.

If you look back at almost any tech, it's always in the early days that the leaps happen.
Just like Intel used to have, when you went from 33mhz 486, then to 66Mhz, then Pentiums, with 90, then 133, and so on and on.
I'm sure Apple's Arm has been through this same path.

Please note: I'd love to be shocked, and every year the M1 chip jumps up and perhaps doubles in power.
I will admit however, I'm not expecting M1 to be an evolutionary mirror of the iphone A chip advancements.
I feel that is an unrealistic expectation.

You won't see, for example MacBook Air performance doubling every year as the M2, M3, M4 come out.
Whilst that's a lovely thought. As I said, the new M1 is in reality 14 versions in now, and not at version 1 really.

I can't find a A14 chart, so this chart below is rather out of date and would be much higher now, however, as said above I don't feel you can realistically put the M1 at the bottom left of a chart like this and expect the same increase.

The M1, M2, M3 etc are going to have an amazing future I'm certain of it.
And I'm so excited by this change, and REALLY hope Apple does not screw it up, by making amazing machines on one hand and also ruining some aspects of what they COULD potentially do amazingly well in other areas (Gaming?)

It will be super interesting to see what changes Apple makes to the A15, for example to make it even more suited to the desktop. The A14 is basically still just a very low power mobile chip.
We're a long way, it seems from a real desktop chip.

Thanks for a thorough post.

I agree 100% with your premises and your conclusion of future performance gains for the MacBook Air. The 10W M1 soc as it evolves will not be on an “revolution” cycle every year or every other year but more like an evolution.

Notwithstanding, as it pertains to desktop SOCs, I disagree. No one knows which name to use. For simplicity let’s call it the M1X or M2X (if it arrives after the M2).

My thinking is that such a desktop SOC will be based on the same technology (5nm, firestorm cores) that is in the M1 SOC. however, the M1X can be set at 45-65 watts meaning it can be double the size and have at least twice as many cores. If it has the same POP packaging, it will carry up to 2 RAM chips with 16/32 or 32/64Gb configs.

In summary, desktop M SOCs will be able to achieve close to twice the same performance (multicore) without new technologies/revolution that you correctly explained.

 

[Piggie]

Thanks for a thorough post.

I agree 100% with your premises and your conclusion of future performance gains for the MacBook Air. The 10W M1 soc as it evolves will not be on an “revolution” cycle every year or every other year but more like an evolution.

Notwithstanding, as it pertains to desktop SOCs, I disagree. No one knows which name to use. For simplicity let’s call it the M1X or M2X (if it arrives after the M2).

My thinking is that such a desktop SOC will be based on the same technology (5nm, firestorm cores) that is in the M1 SOC. however, the M1X can be set at 45-65 watts meaning it can be double the size and have at least twice as many cores. If it has the same POP packaging, it will carry up to 2 RAM chips with 16/32 or 32/64Gb configs.

In summary, desktop M SOCs will be able to achieve close to twice the same performance (multicore) without new technologies/revolution that you correctly explained.

Interesting reply thanks.
Do you think Apple will be able to greatly increase the physical size, and power needs of their chips, whilst at the same time avoiding the problems with heat/limitations that have plagued Intel and AMD for literally decades.

I'll admit I've been disappointed in the lack of imagination chip designers have had when it comes to cooling.
You could literally run cooling solutions right through the middle of a chip when a different design, but no one seems to have even attempted such a thing.

if Apple can get this performance from a 10w M1, then, a 45-65w in an iMac would scream.

I do have just one tiny concern and that is, will Apple deliberately hold back power in the future for lower machines to keep them low? I hope not. I hope the whole range will be enhanced each year or two.
Generally you are going to hit hard limitations at the top end much quicker than the lower end.
And you don't want the lower end starting to grow in power to a point where it threatens your higher end.

 

[JohnnyGo]

Apple is not holding back with M1. The 10W design is appropriate for their vision for laptops. Lots of power with minimal battery consumption.

The current M1 has to throttle back in the fanless MBA. That does not happen with active cooling in the MBP. That’s why long running benchmarks perform close to 50% better in the MBP.

Why they chose the same M1 for the Mac mini? Cost savings and a lower entry price. I’m positive that in 12 months we will see another model Mini with a more powerful SOC keeping the current M1 or similar M2 in the lineup (almost like a Mini vs Mini Pro).

 

[Piggie]

Apple is not holding back with M1. The 10W design is appropriate for their vision for laptops. Lots of power with minimal battery consumption.

The current M1 has to throttle back in the fanless MBA. That does not happen with active cooling in the MBP. That’s why long running benchmarks perform close to 50% better in the MBP.

Why they chose the same M1 for the Mac mini? Cost savings and a lower entry price. I’m positive that in 12 months we will see another model Mini with a more powerful SOC keeping the current M1 or similar M2 in the lineup (almost like a Mini vs Mini Pro).

Oh no no..... I did not mean to suggest Apple is/was holding anything back with the M1. I was thinking into the future.
AS the SOC's improve, and larger more power hungry ones get developed also, there is going to become a point, like with AMD, Intel, Nvidia you are going to start hitting hard limits at the upper end, which are going to get harder to overcome.
Like it's easy to get a 5w chip to become a 10w chip and double what you get, but you are not going to get a 60w chip into a 120w chip without major problems. So logically the higher end chips are going to hit certain limits sooner than the lower end chips, and it's reasonable to guess the gap between the worst and best from Apple will shrink.
Therefore logically you need to not make the low end chips too good to keep the price/value steps in the range or products.
Not a problem for a few years I'm sure though

I genuinely hope that Apple does not neglect the GPU side of things, simply as it's not of much interest to them.
There is no reason for it to be of no interest.
I mean, would it not be amazing for Apple to not only say they can produce a machine with a CPU as fast or faster than the best consumer products from AMD or Intel, but to also say they also have better GPU's than AMD/Nvidia.

I've literally been disappointed in Apple for the past couple of Decades that they seemed happy to fit such weak GPU's in their products.
Now, they have it all under their control, so I hope this is going to change.
After all, the statement has always been, we make the very best machines we can, which has never really been true when you fit a mobile chipset, low power GPU and 5400rpm hard drive whilst saying it.

I hope, esp with Jony Ive out of the way with his form over function mentality that Apple will shake this off, and, given they can design everything themselves, will product something which is the best in all areas.

 

[JohnnyGo]

Apple is certainly investing a lot on a dedicated GPU chip to better utilize the power envelope of a desktop.

It’s easier to have two separate chips drawing each 45-65watts than a monster 100-150W chip. Be it for cooling be it for manufacturing/chip design.

Agree with you. GPU is an important part of the evolution of M1 / Apple Silicon

 

[dmccloud]

Looking at that chart from Anandtech, it appears that the A-series of chips have increased performance linearly from the A9 to the A14 - you can literally draw a straight line connecting each of those results for the Apple chips. This would seemingly contradict Piggie's statement that:

"Please note: I'd love to be shocked, and every year the M1 chip jumps up and perhaps doubles in power.
I will admit however, I'm not expecting M1 to be an evolutionary mirror of the iphone A chip advancements.
I feel that is an unrealistic expectation.

You won't see, for example MacBook Air performance doubling every year as the M2, M3, M4 come out.
Whilst that's a lovely thought. As I said, the new M1 is in reality 14 versions in now, and not at version 1 really."

For any CPU manufacturer to be able to sustain performance gains on that level over several generations is notable, especially given how Intel's performance over the same timeframe fluctuates and has a noticeably smaller overall increase (41.4 to 58.6 between 2013 and today) compared to Apple's total gains in the same timeframe (21.2 to 63.3 based on what that chart actually shows). I think that 11th gen result for Intel should be tossed out though, because Intel has only made 11th gen CPUs to replace their U and Y series (notebook chips), whereas the rest of those results are for their K series (desktop unlocked) SKUs with the lone exception of the 5775C (5th gen) Intel chip. You can draw a straight line between the Intel results if you omit both the 5th and 11th gen scores, but even then you see a switch from the i7 to the i9 processor (9900K and 10900K) in order to maintain that relative performance increase.

The other consideration is that while Apple has booked out all of TSMC's 5nm capacity for the foreseeable future, TSMC is already at work on both 3nm and 2nm processes, either of which Apple could move to when the processes and fabrication capacity have been finalized. Those process shrinks will also factor into future performance boosts as Apple continues to develop new versions of their processors and SoCs.

 

[JohnnyGo]

Part of Apple’s success certainly was due to chip manufacturing improvements.

After starting off with Samsung, Apple moved to TSMC and never looked back. Intel tried to step in as a contract manufacturer but Apple bet on TSMC. The bet paid off !

Of course, the shrinking process size helped but Apple did a LOT of work internally. You can see that most tech pundits have no idea about UMA and how it impacts performance or memory use. Apple designs its own GPU cores.

It’s an incredible long play (10+ years) that Steve Job envisioned and Tim Cook executed to perfection

 

[Macalway]

This is true, but the real mind blow happened a few days ago. You don't blow minds like that often.



My 2 cnts