Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.

Mintwurm

macrumors newbie
Original poster
Nov 20, 2020
7
4
Apple's new m1 chips are met with universal praise.
And rightly so: They completely outclass the intel chips in previous macs.
But I feel like there is some misrepresentation going on here.
Was this made possible by the move from x86 to ARM?
Or is it about engineering innovations and good chip design - that could also be had on the x86 platform?

If you look at cpu benchmarks, amd is already outperforming the m1.
Cinebench R23 shows a clear lead of the r4700u over the m1. This is comparing a macbook pro against a similar fan-cooled laptop.
Both the peak and sustained performance of the ryzen mobile chip are better than the m1.

Is this a fair comparison? Hard to say.
For a mobile chip, it's important to look at performance per watt.
And I think the fan-cooled m1 in the macbook pro is in a very similar power bracket as the amd ryzen chips.
So I would argue that yes, this is a fair comparison.

The ryzen 4xxxu chips are based on amds zen 2 architecture.
The newer version of this is called zen 3 and offers about 20% more performance per clock, plus slightly higher clocks, and less power draw.
So the new Ryzen 5000 mobile chips, based on zen 3 and soon to be released, will be a very serious competition to apple.
I would say that at least the current m1 will be very clearly outclassed by them.
Of course, by then I'd expect an m1x. Maybe with 12 cores in total, 8 performance cores?
That would be quite a battle then, between amd and apple :)

(tip: don't buy a macbook pro now. I would predict a refreshed macbook pro 16 and a new macbook pro 14, with an m1x chip)

Also very important: The m1 is manufactured in a 5nm process. Amd is using 7nm.
The manufacturing process has a significant influence on power efficiency.
This is very much connected to business politics and striking good deals with the manufacturing companies.
It is not directly related to x86 vs arm.

Long story short:
People say the move from intel x86 to arm is monumental and a huge technical breakthrough.
But I just don't see it.
Yes, these new Apple chips are awesome.
But amd is proving that you can do the same on x86.
Why did Apple ditch Intel then?
Because Intel was sleeping for years and just didn't provide innovation.
Why did Apple not move to amd?
Because back when this decision was made, amd didn't have anything good on the horizon.

Don't get me wrong. This is not about demeaning apple's engineering efforts.
It's great to see competition.
I think the new apple silicon is amazing for two reasons:

1) It scales down very well and delivers excellent performance in a fanless design.
This is something I haven't seen from the amd chips.
I have the impression that the sweet spot for operating an m1 is at lower wattage.
You still get very competitive performance at much lower power consumption.
But when pushed to higher clockrates and higher TDP, there are diminishing returns.

2) The gpu is great. It vastly outperforms any integrated gpu from either amd or intel.

Something noteworthy:
Two years ago, apple released the a12 for iphones and the a12x for ipads.
The a12 has 2 big and 4 small cpu cores, plus a 4 core gpu.
The a12x has 4 big and 4 small cpu cores, plus a 7 core gpu.

Now we have the a14.
It has 2 big and 4 small cpu cores, plus a 4 core gpu.
There is no a14x. The ipad pro this year was released with a very outdated a12z.
But there is the m1. With 4 big and 4 small cpu cores, plus a 7/8 core gpu.
Small improvements of the m1: Apple has given it a more generous l2 cache and has increased the max clock frequency from 3 to 3.2 Ghz.

Now look at tests of the ipad pro from the last two years (a12x and a12z).
The max power draw is about 15 to 20 watts.
Sustained load around 10 watts.
This is exactly what we're seeing from the macbook air now.
This new m1 chip is a mobile chip, put in a notebook.
Don't get me wrong, it makes perfect sense.
What we see here is convergence of apple's mobile and desktop chips.


EDIT:
Adding some sources:
 
Last edited:
  • Like
  • Sad
Reactions: m-a and calstanford
Apple's new m1 chips are met with universal praise.
And rightly so: They completely outclass the intel chips in previous macs.
But I feel like there is some misrepresentation going on here.
Was this made possible by the move from x86 to ARM?
Or is it about engineering innovations and good chip design - that could also be had on the x86 platform?

If you look at cpu benchmarks, amd is already outperforming the m1.
Cinebench R23 shows a clear lead of the r4700u over the m1. This is comparing a macbook pro against a similar fan-cooled laptop.
Both the peak and sustained performance of the ryzen mobile chip are better than the m1.

Is this a fair comparison? Hard to say.
For a mobile chip, it's important to look at performance per watt.
And I think the fan-cooled m1 in the macbook pro is in a very similar power bracket as the amd ryzen chips.
So I would argue that yes, this is a fair comparison.

The ryzen 4xxxu chips are based on amds zen 2 architecture.
The newer version of this is called zen 3 and offers about 20% more performance per clock, plus slightly higher clocks, and less power draw.
So the new Ryzen 5000 mobile chips, based on zen 3 and soon to be released, will be a very serious competition to apple.
I would say that at least the current m1 will be very clearly outclassed by them.
Of course, by then I'd expect an m1x. Maybe with 12 cores in total, 8 performance cores?
That would be quite a battle then, between amd and apple :)

(tip: don't buy a macbook pro now. I would predict a refreshed macbook pro 16 and a new macbook pro 14, with an m1x chip)

Also very important: The m1 is manufactured in a 5nm process. Amd is using 7nm.
The manufacturing process has a significant influence on power efficiency.
This is very much connected to business politics and striking good deals with the manufacturing companies.
It is not directly related to x86 vs arm.

Long story short:
People say the move from intel x86 to arm is monumental and a huge technical breakthrough.
But I just don't see it.
Yes, these new Apple chips are awesome.
But amd is proving that you can do the same on x86.
Why did Apple ditch Intel then?
Because Intel was sleeping for years and just didn't provide innovation.
Why did Apple not move to amd?
Because back when this decision was made, amd didn't have anything good on the horizon.

Don't get me wrong. This is not about demeaning apple's engineering efforts.
It's great to see competition.
I think the new apple silicon is amazing for two reasons:

1) It scales down very well and delivers excellent performance in a fanless design.
This is something I haven't seen from the amd chips.
I have the impression that the sweet spot for operating an m1 is at lower wattage.
You still get very competitive performance at much lower power consumption.
But when pushed to higher clockrates and higher TDP, there are diminishing returns.

2) The gpu is great. It vastly outperforms any integrated gpu from either amd or intel.

Something noteworthy:
Two years ago, apple released the a12 for iphones and the a12x for ipads.
The a12 has 2 big and 4 small cpu cores, plus a 4 core gpu.
The a12x has 4 big and 4 small cpu cores, plus a 7 core gpu.

Now we have the a14.
It has 2 big and 4 small cpu cores, plus a 4 core gpu.
There is no a14x. The ipad pro this year was released with a very outdated a12z.
But there is the m1. With 4 big and 4 small cpu cores, plus a 7/8 core gpu.
Small improvements of the m1: Apple has given it a more generous l2 cache and has increased the max clock frequency from 3 to 3.2 Ghz.

Now look at tests of the ipad pro from the last to years (a12x and a12z).
The max power draw is about 15 to 20 watts.
Sustained load around 10 watts.
This is exactly what we're seeing from the macbook air now.
This new m1 chip is a mobile chip, put in a notebook.
Don't get me wrong, it makes perfect sense.
What we see here is convergence of apple's mobile and desktop chips.

I don’t know what you mean by convergence of mobile and desktop. Apple used to use the same family of Motorola and PPC chips in desktop and laptops, just lower power versions in the laptops. So now the M1 will have lower power for mobile and higher power with more cores/ram for desktop.

It’s great to have the M1. It really feels like I have a REAL MAC again for the first time in a long time. I want to see the same fun ads again. PPC vs Intel ads were great fun.
 
Apple's new m1 chips are met with universal praise.
And rightly so: They completely outclass the intel chips in previous macs.
But I feel like there is some misrepresentation going on here.
Was this made possible by the move from x86 to ARM?
Or is it about engineering innovations and good chip design - that could also be had on the x86 platform?

Who knows. Maybe. But then again, Apple doesn't have a right to design x86 chips, so what are they supposed to do, wait another 5-6 years until Intel or AMD have a breakthrough? Not to mention that x86 is a mess ISA-wise. Aarch64 is a new, well-designed instruction set that leverages decades of experience and it has a lot of features that make design of high-performance CPUs easier.

If you look at cpu benchmarks, amd is already outperforming the m1.
Cinebench R23 shows a clear lead of the r4700u over the m1. This is comparing a macbook pro against a similar fan-cooled laptop.
Both the peak and sustained performance of the ryzen mobile chip are better than the m1.


Not really. In single-core, Apple outperforms any mobile Ryzen by a very healthy margin. In multi-core, Renoir chips can get ahead because they have double the amount of main CPU cores and they boost way beyond their sustained 15W spec. When you look at the R23 results, M1 runs at around 15W while the Renoir chips will be closer to 30-40 watts. Apple simply chose the M1 to say at a lower power bracket to keep cool and silent operation. An 8-core Apple chip with better cooling (two-fan 13"/14") will be at least 60% faster than the best Renoir chips while still retaining the excellent
battery like of M1.


And I think the fan-cooled m1 in the macbook pro is in a very similar power bracket as the amd ryzen chips.

As I explain above, not really. If you let them run at full power, the Ryzen will throttle — hard (check the continuous Cibenech results on notebookchek, where 25W 4800U throttles from 1600 to 1200 after couple of minutes). The M1 will happily chug along all day, because it never overboosts. It's dynamic frequency range is much lower and the performance more predictable.

One might say, well, for this kind of machine, nobody is going to do heavy-duty computation for hours and hours. And they will be right. If you need a light laptop for this kind of bursty work, a Renoir based one will be better suited. But it's still apples and oranges. Wait for a proper 8-core Mac laptop.


The newer version of this is called zen 3 and offers about 20% more performance per clock, plus slightly higher clocks, and less power draw.
So the new Ryzen 5000 mobile chips, based on zen 3 and soon to be released, will be a very serious competition to apple.
I would say that at least the current m1 will be very clearly outclassed by them.

I don't see it. Zen 3 at 20 watts has the same performance as M1 at 5 watts. Mobile Zen 3 will probably be clocked slightly lower than the desktop Zen 3, and M1 is a really more than a worthy match for a 5950X (a $900 CPU) core-per-core.

Now, an 8-core Zen3 laptop part will definitely have a decent advantage over the M1. But again.. apple and oranges. An 8 performance core part by Apple will handily outperform it, simply because Apple can run 8 cores at 3ghz within a 30W bracket and Zen 3 can't do the same with 8 cores at 4 ghz.

At any rate, early Cesanne benchmarks are out and While Zen3 is great it's not really closing the gap to M1


Also very important: The m1 is manufactured in a 5nm process. Amd is using 7nm.
The manufacturing process has a significant influence on power efficiency.
This is very much connected to business politics and striking good deals with the manufacturing companies.
It is not directly related to x86 vs arm.

Less than 10% when you look at improvements between the A13 and A14. Also, A14 adds some more execution units — a lot of performance improvements come from there rather from the process alone. A Zen3 on 5nm process will still have the double power consumption of M1 at the same performance level.

Long story short:
People say the move from intel x86 to arm is monumental and a huge technical breakthrough.
But I just don't see it.

Because you are looking at the wrong thing :) Look at the entire picture, not at isolated benchmark scores in narrow contexts. Also, Cinebench is not a good benchmark. It has poor CPU utilization. It's good for checking throttling behavior, but it's not a balanced representation of CPU capabilities.

It scales down very well and delivers excellent performance in a fanless design.
This is something I haven't seen from the amd chips.
I have the impression that the sweet spot for operating an m1 is at lower wattage.
You still get very competitive performance at much lower power consumption.
But when pushed to higher clockrates and higher TDP, there are diminishing returns.

Again, an M1 core consuming 5 watts has comparable performance to a Zen 3 core consuming 20 watts. We don't even know whether 3.2ghz is a practical ceiling for these cores. It is very possible that Apple is limiting the clocks to give themselves more breathing space for upgrades. The higher-tier 13" could run at 3.5ghz@10 watts for all we know.

Talking about good performance at low power consumption: that's precisely the selling point of Zen 2 mobile chips. They can't compete with Intel on single core, but they can scale well to multi-core precisely because AMD can clock them very low. Still, Apple is years ahead here. They could have an 8 core CPU, each core running at 3 ghz, with total TDP of under 30-40 watts.


What we see here is convergence of apple's mobile and desktop chips.

What we see here is Appel taking the chip design crown. Who cares whether you label a chip as "mobile" or as "desktop" if it gives you desktop-class performance?
 
Apple's new m1 chips are met with universal praise.
And rightly so: They completely outclass the intel chips in previous macs.
But I feel like there is some misrepresentation going on here.
Was this made possible by the move from x86 to ARM?
Or is it about engineering innovations and good chip design - that could also be had on the x86 platform?

If you look at cpu benchmarks, amd is already outperforming the m1.
Cinebench R23 shows a clear lead of the r4700u over the m1. This is comparing a macbook pro against a similar fan-cooled laptop.
Both the peak and sustained performance of the ryzen mobile chip are better than the m1.

Is this a fair comparison? Hard to say.
For a mobile chip, it's important to look at performance per watt.
And I think the fan-cooled m1 in the macbook pro is in a very similar power bracket as the amd ryzen chips.
So I would argue that yes, this is a fair comparison.

The ryzen 4xxxu chips are based on amds zen 2 architecture.
The newer version of this is called zen 3 and offers about 20% more performance per clock, plus slightly higher clocks, and less power draw.
So the new Ryzen 5000 mobile chips, based on zen 3 and soon to be released, will be a very serious competition to apple.
I would say that at least the current m1 will be very clearly outclassed by them.
Of course, by then I'd expect an m1x. Maybe with 12 cores in total, 8 performance cores?
That would be quite a battle then, between amd and apple :)

(tip: don't buy a macbook pro now. I would predict a refreshed macbook pro 16 and a new macbook pro 14, with an m1x chip)

Also very important: The m1 is manufactured in a 5nm process. Amd is using 7nm.
The manufacturing process has a significant influence on power efficiency.
This is very much connected to business politics and striking good deals with the manufacturing companies.
It is not directly related to x86 vs arm.

Long story short:
People say the move from intel x86 to arm is monumental and a huge technical breakthrough.
But I just don't see it.
Yes, these new Apple chips are awesome.
But amd is proving that you can do the same on x86.
Why did Apple ditch Intel then?
Because Intel was sleeping for years and just didn't provide innovation.
Why did Apple not move to amd?
Because back when this decision was made, amd didn't have anything good on the horizon.

Don't get me wrong. This is not about demeaning apple's engineering efforts.
It's great to see competition.
I think the new apple silicon is amazing for two reasons:

1) It scales down very well and delivers excellent performance in a fanless design.
This is something I haven't seen from the amd chips.
I have the impression that the sweet spot for operating an m1 is at lower wattage.
You still get very competitive performance at much lower power consumption.
But when pushed to higher clockrates and higher TDP, there are diminishing returns.

2) The gpu is great. It vastly outperforms any integrated gpu from either amd or intel.

Something noteworthy:
Two years ago, apple released the a12 for iphones and the a12x for ipads.
The a12 has 2 big and 4 small cpu cores, plus a 4 core gpu.
The a12x has 4 big and 4 small cpu cores, plus a 7 core gpu.

Now we have the a14.
It has 2 big and 4 small cpu cores, plus a 4 core gpu.
There is no a14x. The ipad pro this year was released with a very outdated a12z.
But there is the m1. With 4 big and 4 small cpu cores, plus a 7/8 core gpu.
Small improvements of the m1: Apple has given it a more generous l2 cache and has increased the max clock frequency from 3 to 3.2 Ghz.

Now look at tests of the ipad pro from the last two years (a12x and a12z).
The max power draw is about 15 to 20 watts.
Sustained load around 10 watts.
This is exactly what we're seeing from the macbook air now.
This new m1 chip is a mobile chip, put in a notebook.
Don't get me wrong, it makes perfect sense.
What we see here is convergence of apple's mobile and desktop chips.


EDIT:
Adding some sources:
Once upon a time, Apple made an SoC for the first generation iPad. The A4. It was great and the peasants rejoiced as Apple started to put it in the iPhone 4. Flash forward ten years and Apple has pretty much been advancing their SoCs, baking into them all sorts of crazy cool technologies into them (including their own GPUs and things like the Neural Engine, Secure Enclave, and other things that you never really had on Intel PCs or Macs, for that matter). Nevermind all of that for a second, though. What Apple has been doing is advancing their chips, year over year. The A7, their first 64-bit SoC, for instance, was light years ahead of their A6, which was pretty great compared to the A5, which made the A4 look like nothing. A9 was a crazy breakthrough (in the move from eMMC storage to PCIe storage alone!), A10 Fusion introduced a host of new things (asymmetric multi-processing) and A11 Bionic only took it further with the neural engine. By the time you get to A12, A12X and A12Z, you're at a night and day difference from where you were just four years earlier with A8. A8 was a 2014 SoC; A12 was a 2018 SoC. For Intel, that's the difference between Haswell and Coffee Lake or Kaby Lake Refresh (whichever flavor of 8th Gen you prefer). I'm not going to tell you that a top of the line Coffee Lake CPU isn't superior to a top of the line Haswell CPU, because that wouldn't be true. But the difference is nowhere near as substantial. 9th Gen and 10th Gen Intel was also pretty great overall, but were they game changers compared to 8th Gen? No. But you had A8 just six years ago with storage that wasn't even PCIe. Four years later, you're beating out every Intel MacBook Air, 13" MacBook Pro, and most 15" MacBook Pros that had ever existed with A12X. Hell, you're beating out the 2020 Intel MacBook Air in EMULATED performance on the A12Z, which is the same, but with an additional GPU core.

Unified memory and efficient SoC design with dedicated controllers is great and all. But you have a roadmap that is exponentially improving compared to what Intel is stuck with and pretty much has been stuck with since Haswell in 2014.
 
  • Like
Reactions: Mitthrawnuruodo
What we see here is Appel taking the chip design crown. Who cares whether you label a chip as "mobile" or as "desktop" if it gives you desktop-class performance?
Remember all of Intel's CORE / Conroe etc., was actually their Israeli Mobile team. I forget the name of their desktop arch before CORE, but man was it a dog. I remember the execs taking pride in their first dual core, which was literally two CPUs on two interconnected dies, but through a single socket.

Honestly, using a MacBook Air M1 this week has been remarkable. Snappy and silent is what computing should be in 2020. My poor old 16" Pro is a lovely piece of engineering hobbled by a space heater called i7.
 
Remember all of Intel's CORE / Conroe etc., was actually their Israeli Mobile team. I forget the name of their desktop arch before CORE, but man was it a dog. I remember the execs taking pride in their first dual core, which was literally two CPUs on two interconnected dies, but through a single socket.

Honestly, using a MacBook Air M1 this week has been remarkable. Snappy and silent is what computing should be in 2020. My poor old 16" Pro is a lovely piece of engineering hobbled by a space heater called i7.
It was ironically called NetBurst.
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.