Intel Alder Lake vs. Apple M1

[leman]

45W in a 16" MBP chassis is easy and make it the default which is still faster than M1 Pro/Max 10CPU with an optional 55W performance mode.

That is a very dubious statement. Yes, ADL at 55W is faster at R23 and possibly some other multi core throughput tests than M1 at 30-40W. For stuff folks actually use computers though? Hardly.

So they haven’t measured the power themselves?

No idea what they did, probably just took package power numbers from Anandtech. I report what I see on my machine.

I’m not sure that’s actually true (unless you mean pumping additional power through the M1 Pro/Max CPU which Apple doesn’t allow). Given the differences in architecture, I would say that a higher core count M-series chip would scale far better on perf/W than Alder Lake on this graph.

Oh, no doubt about that. What I meant is simply that you can run ADL at 60 or 100W if you so want, abs will get your 20/40% boost. Can’t do the same with M1, you are locked into what Apple gives you.

 

[Andropov]

Also, some APIs (e.g. some GPU functionality) require memory allocations to be page-aligned.

Do you have any particular example in mind? I didn't know this, so I'm curious.

One picture is worth a thousand words.

View attachment 1954603

Love how the video is titled 'Apple & AMD can't keep up!'. Intel is the one who had to keep up. Apple and AMD offerings are already about a year an a half old (their core designs, anyway).

 

[Xiao_Xi]

No idea what they did, probably just took package power numbers from Anandtech.

Apple's M1 Pro, M1 Max SoCs Investigated: New Performance and Efficiency Heights

www.anandtech.com

 

[Gerdi]

Where Intel of course has an obvious edge is performance scaling with power. Which is again not surprising given their architecture and the focus on power-hungry desktop applications.

Crazy Dave already pointed out, that this is not true at all. The really big flaw in the drawing is, that architectures are not compared at iso voltage. The fundamental reason, that this happens is, that you compare CPUs with difference number of cores. CPUs with less cores are at an inherent disadvantage at iso performance or iso power points - unfortunately these are the points on the graph, people with no clue like the guys from TechPowerup or Mi7chy, tend to compare.
Let me illustrate this with following drawing. Here i took the AMD graph and scaled it up to 16 cores. Likewise i scaled the the single M1 point to 14 cores. Note, that i only changed the number of cores keeping the microarchitecture, voltage and hence frequency constant.

1. Lesson: Do not even try to compare 2 CPUs with different core counts at ISO power or ISO performance when you want to reason about architectural power efficiency.
Point in case, both AMD CPUs (8 core and 16 core) look to have vastly different efficiency in the above graph, while the architectural efficiency is of course equal.
2. Lesson: Intel Alderlake is far behind both AMD and Apple when it comes down to architectural power efficiency.


ps: Cinebench is not suited to compare with M1, because it is hand optimized for AVX and essentially emulated on M1. So take the above graph mostly for comparing Intel with AMD and do not try to include M1.

 

[leman]

Crazy Dave already pointed out, that this is not true at all. The really big flaw in the drawing is, that architectures are not compared at iso voltage. The fundamental reason, that this happens is, that you compare CPUs with difference number of cores.

As I already wrote in my reply to @crazy dave my statement referred to actual products, not architectural details. There is little doubt that say, a hypothetical 16-core M1 would run circles around a 16 core ADL. But we have what we have and these are the products customers can choose.

 

[JMacHack]

And, what Apple will do going forward. Wouldn't be surprised to see a refreshed 2019/2020 16" Macbook Pro with Alder Lake and something better than 5600m unless they want to halve their <10% marketshare.

45W in a 16" MBP chassis is easy and make it the default which is still faster than M1 Pro/Max 10CPU with an optional 55W performance mode. So, two lines of laptops, 16" Macbook Pro(fessional) with Alder Lake for people that aren't ready to give up native x64 compatibility and bootcamp. And, separate 16" Macbook Pro(sumer) with upsized iPad SoC.

L-L-Look a-at this single cherrypicked benchmark! I-I-Intel is still king! D-D-Don’t listen to anyone who praises the M Series, they’re all shills!

 

[JMacHack]

Crazy Dave already pointed out, that this is not true at all. The really big flaw in the drawing is, that architectures are not compared at iso voltage. The fundamental reason, that this happens is, that you compare CPUs with difference number of cores. CPUs with less cores are at an inherent disadvantage at iso performance or iso power points - unfortunately these are the points on the graph, people with no clue like the guys from TechPowerup or Mi7chy, tend to compare.
Let me illustrate this with following drawing. Here i took the AMD graph and scaled it up to 16 cores. Likewise i scaled the the single M1 point to 14 cores. Note, that i only changed the number of cores keeping the microarchitecture, voltage and hence frequency constant.

View attachment 1954743

1. Lesson: Do not even try to compare 2 CPUs with different core counts at ISO power or ISO performance when you want to reason about architectural power efficiency.
Point in case, both AMD CPUs (8 core and 16 core) look to have vastly different efficiency in the above graph, while the architectural efficiency is of course equal.
2. Lesson: Intel Alderlake is far behind both AMD and Apple when it comes down to architectural power efficiency.


ps: Cinebench is not suited to compare with M1, because it is hand optimized for AVX and essentially emulated on M1. So take the above graph mostly for comparing Intel with AMD and do not try to include M1.

Don’t explain this, he’ll just ignore it.

 

[Gerdi]

As I already wrote in my reply to @crazy dave my statement referred to actual products, not architectural details. There is little doubt that say, a hypothetical 16-core M1 would run circles around a 16 core ADL. But we have what we have and these are the products customers can choose.

Of course, if you strictly restrict your discussion to actual products, thats is a somewhat valid argument. And you can make an inefficient microarchitecture look good in actual products as long as you have sufficiently more cores compared to others. My drawing just illustrates how much the core-counts impacts the efficiency at iso-power comparison points.

But as you certainly acknowledge, people are fast to use these graphs to reason about architectural efficiency. In particular they take these graphs to compare the inherent efficiencies microarchitectures of Alder Lake, Zen 3 or M1 or even claim that the Intel architects have somehow beaten the other when it comes to efficiency - which is far from being the truth. The fact is Intel even with Alder Lake is far behind when it comes to efficiency - they only look somewhat competitive in these graphs because they have more cores.
There are even people like Mi7chy (lets just ignore his power efficiency for the moment, e.g. intelligent statements vs. used power), suggesting that Apple would go back to the worst available architecture like Alderlake based on these graphs.

Point in case. Assume i would not have labeled the graph above. You can be certain, that (some) people would conclude, that the architecture behind the grey graph is much more efficient than architecture behind the orange graph - while in fact both graphs come from the same architecture. (Coincidentally the particular CPU behind the grey graph even exists, it is called Ryzen 7 5950X).

 

[pshufd]

I saw Vadim (Max Tech) saying that the single-core on the M2 will bump over Alder Lake. Can anyone verify this? The earliest we'd see this would be with the M2 Air.

 

[mi7chy]

View attachment 1954743

Where does one buy that fake M1 14CPU?

Or, that fake AMD 16-core that does 20k+ on C23 at 70W? My 5950x power limited to ~70W does 11509 so about half of your fake results.

 

[leman]

But as you certainly acknowledge, people are fast to use these graphs to reason about architectural efficiency. In particular they take these graphs to compare the inherent efficiencies microarchitectures of Alder Lake, Zen 3 or M1 or even claim that the Intel architects have somehow beaten the other when it comes to efficiency - which is far from being the truth. The fact is Intel even with Alder Lake is far behind when it comes to efficiency - they only look somewhat competitive in these graphs because they have more cores.

Oh, sure. And of course, any claims of architectural parity between current Intel/AMD or Apple are simply laughable. It is a hard fact that Apple's cores can match the performance of best x86 cores at 1/3-1/4 of power consumption, and nobody can really argue about it. That's why all the e-peen comparisons have moved to multicore thought put where x86 can get ahead thanks for higher number of cores and higher power consumption.

But then again, there is certain validity to these arguments. Apple has a superior core technology, but the products they decided to build with that technology are not without limitations. E.g. they are excellent mobile cores that offer an unmatched combination of performance and efficiency, but Apple does not offer much for desktop enthusiasts or, say, mobile users who want a "desktop replacement". And of course, a 16-core M1 chip (if Apple ever makes one) is likely to be significantly more expensive than a 16-core Intel or AMD chip, simply because Apple uses more expensive baseline technology.

Personally, I think one has to make a strict line between architecture and product. Architecture-wise, Apple is the best, by a huge margin. In the domain of mobile performance, Apple is the best, by a very healthy margin. In the domain of single-core performance, Apple is top-tier (ADL is nominally faster but that hardly matters since the gap is minuscule). For multicore sustained performance, x86 can offer better performing products (with caveats).

There are even people like Mi7chy (lets just ignore his power efficiency for the moment, e.g. intelligent statements vs. used power), suggesting that Apple would go back to the worst available architecture like Alderlake based on these graphs.

Ah, well, mi7chy are trying their best. It's a pity they decided to fight a battle without purpose. I am not even sure what they are trying to prove anymore.

I saw Vadim (Max Tech) saying that the single-core on the M2 will bump over Alder Lake. Can anyone verify this? The earliest we'd see this would be with the M2 Air.

That is not unlikely. M2 will probably have a clock boost of at least 10% + more cache, it should put it on par or slightly above fastest desktop single-core ADL.

Or, that fake AMD 16-core that does 20k+ on C23 at 70W? My 5950x power limited to ~70W does 11509 so about half of your fake results.

Oh my, that's quite terrible. My M1 does the same score at 30W...

 

[cmaier]

All I know is that I just actively used MS Word for the last hour with two 400+ page documents open and 7 pdf documents loaded in preview (plus numerous other apps on other desktops, doing their thing int he background), and in that time my M1 Pro Max MBP 16” battery dropped from 80% to… 80%. My Intel MBP would have dropped 10% with that same workload.

 

[crazy dave]

Where does one buy that fake M1 14CPU?

Same place you would buy a 16" MBP with a 55W CPU or the totally awesome next gen Intel chip with node shrinks and the magical unicorn architecture. With actually somewhat less speculation than the latter.

Intel next year with node shrink will be killer.

Now to see what AMD has to offer this year with Ryzen 6000 mobile. They hinted at 1.3x multithreaded performance so it'll be close.

And, what Apple will do going forward. Wouldn't be surprised to see a refreshed 2019/2020 16" Macbook Pro with Alder Lake and something better than 5600m unless they want to halve their <10% marketshare.

45W in a 16" MBP chassis is easy and make it the default which is still faster than M1 Pro/Max 10CPU with an optional 55W performance mode. So, two lines of laptops, 16" Macbook Pro(fessional) with Alder Lake for people that aren't ready to give up native x64 compatibility and bootcamp. And, separate 16" Macbook Pro(sumer) with upsized iPad SoC.

===============

I saw Vadim (Max Tech) saying that the single-core on the M2 will bump over Alder Lake. Can anyone verify this? The earliest we'd see this would be with the M2 Air.

Aye we won't know for sure until it releases with the next Air but as @leman says Vadim's probably extrapolating from A15 vs A14 for M2 vs M1.

 

[thenewperson]

Ah, well, mi7chy are trying their best. It's a pity they decided to fight a battle without purpose. I am not even sure what they are trying to prove anymore.

Especially now that things are quiet on the Stockfish front

 

[crazy dave]

Oh, sure. And of course, any claims of architectural parity between current Intel/AMD or Apple are simply laughable. It is a hard fact that Apple's cores can match the performance of best x86 cores at 1/3-1/4 of power consumption, and nobody can really argue about it. That's why all the e-peen comparisons have moved to multicore thought put where x86 can get ahead thanks for higher number of cores and higher power consumption.

Personally, I think one has to make a strict line between architecture and product. Architecture-wise, Apple is the best, by a huge margin. In the domain of mobile performance, Apple is the best, by a very healthy margin. In the domain of single-core performance, Apple is top-tier (ADL is nominally faster but that hardly matters since the gap is minuscule). For multicore sustained performance, x86 can offer better performing products (with caveats).

Yeah I agree with all of this. When comparing actual products the products are the products. And when talking about how chips scale, it becomes an interesting theoretical exercise with two approaches: how does a current chip scale by pumping power through it and how does an architecture scale? Unfortunately many sites (and people) confuse the two and the original graph (and mi7chy) try make claims about architecture that then become unfounded. Apple and AMD can indeed scale and "keep up" as it were with their current core architectures, but product-wise they would need to offer a different configuration. Again, I'm pretty sure that you are in agreement with all of this, I'm just restating since these are similar questions with very subtle, but important distinctions that often cause people to talk past each other as we did initially here. I think @jeanlain and I had a similar discussion earlier in the thread about multicore throughput and how to properly compare results and it really comes down to, "are you focusing your comparison on architecture or product?" I think in that discussion I was focusing on product and he was focusing on architecture in a reversal of (my position) here. It really depends on how things are phrased and what the focus is.

But then again, there is certain validity to these arguments. Apple has a superior core technology, but the products they decided to build with that technology are not without limitations. E.g. they are excellent mobile cores that offer an unmatched combination of performance and efficiency, but Apple does not offer much for desktop enthusiasts or, say, mobile users who want a "desktop replacement". And of course, a 16-core M1 chip (if Apple ever makes one) is likely to be significantly more expensive than a 16-core Intel or AMD chip, simply because Apple uses more expensive baseline technology.

Not absolutely sure about this. Probably true since TSMC 5nm is probably more expensive than Intel's 7 and obviously more than TSMC 7nm (AMD). Also Apple has big caches which could mitigate their physical core size advantage. Someone adept at looking at die shots could probably estimate how big Apple's CPU complex is versus a similar AMD/Intel 8-10 core processor. Of course things like the system level cache for Apple is not being directly analogous to AMD/Intel 3rd level caches would complicate such a calculation. And we don't know exactly the costs (I think Sophie Wilson gave a talk about this where she said more than she probably should've, but I don't know the numbers off the top of my head and I think it might've stopped at TSMC 7nm). Having said that, since Apple is an OEM, no longer having to pay Intel or AMD's margins probably means the cost is still less for *them*. People seem convinced that Apple is probably paying less per chip than they were with Intel, but I'm not sure how concretely industry watchers know the relevant numbers especially the full SOC is hard to compare with what they were offering earlier.

 

[pshufd]

Aye we won't know for sure until it releases with the next Air but as @leman says Vadim's probably extrapolating from A15 vs A14 for M2 vs M1.

Oh that. I was hoping for something more concrete given the leaks about potential launches in March.

It doesn't really matter for me, though. I will be using the M1 Pro MacBook Pro for at least five years, and, ideally, ten years.

 

[crazy dave]

Oh that. I was hoping for something more concrete given the leaks about potential launches in March.

It doesn't really matter for me, though. I will be using the M1 Pro MacBook Pro for at least five years, and, ideally, ten years.

As far as I know there haven’t been any performance leaks and I think the latest leaks indicate that M2 will be mid year at the earliest. Heck I’m not even sure if the leaks confirm for certain that they will be avalanche/blizzard cores. We all expect them to be, but as I think @mr_roboto has pointed out previously to me, given the node timings Apple could theoretically launch the 4nm core designs on the M2 before the iPhone A16 gets them. Truthfully, I suspect not, but in theory they could. TSMC 4nm is on production *now*.

 

[Xiao_Xi]

Vadim's probably extrapolating from A15 vs A14 for M2 vs M1.

Could someone do the math?

 

[jdb8167]

Could someone do the math?

Hard to do. The problem is we don’t know what hardware they’re going to introduce. For example, the A15 in my iPad mini 6 is slower in single core than the M1 in my MacBook Air. The A15 in the mini is obviously downclocked but we don’t know if the M2 is going to be based on the A15 cores (probably) and if they’ll be clocked higher than the M1.

 

[JMacHack]

Ah, well, mi7chy are trying their best. It's a pity they decided to fight a battle without purpose. I am not even sure what they are trying to prove anymore.

Assuming he’s not a shill or very, very sad individual, if he’s trying to troll anyone he’s not doing a good job even. His arguments get looked by people far more knowledgeable every thread and ignored thus. Honestly the Stockfish thread was more entertaining and far more convincing than anything he’s ever posted.

I’m playing armchair psychologist of course, but if you have to repeatedly try and skew information to “own the fanbois” and prove the superiority of your PC then one needs a better hobby.

Who knows though, maybe he’s right and Intel somehow overcame 5 years of beatings to make a processor superior in performance per watt and pure performance than the M1 and variants.

Maybe Apple’s engineers are clowns who can’t beat the top minds of AMD, Intel, and NVidia and we on the forums, most of the tech press, many satisfied owners of the M series are all wrong and delusional and he’s the heroic knight in shining armor, who fights using his sword of ultimate truth to take down the evil tyrant of lies and corruption to make Apple’s inferior technology seem superior.

 

[crazy dave]

Could someone do the math?

@jdb8167 listed some of the caveats about doing so but just doing the simplest most naive extrapolation the M2 performance cores might be around 12-13% faster than the M1’s in ST if they follow the same relationship as the A15 cores do to the A14. This would put the M-series avalanche cores around desktop ADL in Spec ST (slightly faster in fp and slightly slower in int) and GB5 (slightly faster). This would be somewhat faster than a mobile 12900hk in ST (but less than 5% so not hugely important). Again, this is the simplest extrapolation based on unverified assumptions.

 

[BigPotatoLobbyist]

Crazy Dave already pointed out, that this is not true at all. The really big flaw in the drawing is, that architectures are not compared at iso voltage. The fundamental reason, that this happens is, that you compare CPUs with difference number of cores. CPUs with less cores are at an inherent disadvantage at iso performance or iso power points - unfortunately these are the points on the graph, people with no clue like the guys from TechPowerup or Mi7chy, tend to compare.
Let me illustrate this with following drawing. Here i took the AMD graph and scaled it up to 16 cores. Likewise i scaled the the single M1 point to 14 cores. Note, that i only changed the number of cores keeping the microarchitecture, voltage and hence frequency constant.

View attachment 1954743

1. Lesson: Do not even try to compare 2 CPUs with different core counts at ISO power or ISO performance when you want to reason about architectural power efficiency.
Point in case, both AMD CPUs (8 core and 16 core) look to have vastly different efficiency in the above graph, while the architectural efficiency is of course equal.
2. Lesson: Intel Alderlake is far behind both AMD and Apple when it comes down to architectural power efficiency.


ps: Cinebench is not suited to compare with M1, because it is hand optimized for AVX and essentially emulated on M1. So take the above graph mostly for comparing Intel with AMD and do not try to include M1.

I believe I said this too or maybe before Dave (?) if you scroll back... anyone can take more - even inferior - cores and scale them at their (relative) ideal curves for aggregate throughput that is competitive with systems that posess even more efficient cores that are being pushed to the upper ends of their performance limits in tandem.

If Apple wanted to throw four more P cores onto the M1 Pro instead of a fat GPU and the video encoding crap, it would blow Intel out of the water (or match it rather at remarkably less power, or offer them the ability to scale performance at less power) without an issue. Because

 

[JouniS]

I believe I said this too or maybe before Dave (?) if you scroll back... anyone can take more - even inferior - cores and scale them at their (relative) ideal curves for aggregate throughput that is competitive with systems that posess even more efficient cores that are being pushed to the upper ends of their performance limits in tandem.

That's a very hypothetical "can". Intel and AMD can make consumer chips with a high CPU core count in practice, because they make CPUs. Apple can't, because it makes SoCs. A SoC is always a compromise between various requirements, because the world is small and there is only limited demand for computing devices. In practice, it's always possible to scale a system with a separate CPU/GPU/etc further towards any particular extreme than a SoC.

 

[crazy dave]

That's a very hypothetical "can". Intel and AMD can make consumer chips with a high CPU core count in practice, because they make CPUs. Apple can't, because it makes SoCs. A SoC is always a compromise between various requirements, because the world is small and there is only limited demand for computing devices. In practice, it's always possible to scale a system with a separate CPU/GPU/etc further towards any particular extreme than a SoC.

I’m not sure about that. Apple makes SOCs but those SOCs replace both the CPU and dGPU (and the iGPU too since they often shipped laptops with iGPUs). So from their perspective as an integrator, a consumer OEM, it’s not actually different. In fact the SOC probably offers better economics. Now die sizes do become an issue eventually, but that’s where they start with multi-die chips. It’s why AMD already uses multi-die chips on its larger core count CPUs. Obviously that has downsides but it keeps the economics feasible and that should work just as well for the SOC. I guess we’ll find out since Apple is clearly prepping for it (so is Intel).

 

[JouniS]

I’m not sure about that. Apple makes SOCs but those SOCs replace both the CPU and dGPU (and the iGPU too since they often shipped laptops with iGPUs). So from their perspective as an integrator, a consumer OEM, it’s not actually different. In fact the SOC probably offers better economics. Now die sizes do become an issue eventually, but that’s where they start with multi-die chips. It’s why AMD already uses multi-die chips on its larger core count CPUs. Obviously that has downsides but it keeps the economics feasible and that should work just as well for the SOC. I guess we’ll find out since Apple is clearly prepping for it (so is Intel).

It's just the basic benefit of modularity. If there are 10 CPU types and 10 GPU types and the CPU and the GPU are manufactured separately, there are 20 different products. If you integrate them into the same chip, you may need up to 100 different products. Having too many product types makes everything more expensive, and predicting the demand for each product in advance becomes more difficult. A modular design makes it possible to serve smaller niches (such as consumers that need multithreaded performance) economically than an integrated design.