Discussion Apple Silicon SoC thread

[Eug]

M1
5 nm
Unified memory architecture - LP-DDR4
16 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 12 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache
(Apple claims the 4 high-effiency cores alone perform like a dual-core Intel MacBook Air)

8-core iGPU (but there is a 7-core variant, likely with one inactive core)
128 execution units
Up to 24576 concurrent threads
2.6 Teraflops
82 Gigatexels/s
41 gigapixels/s

16-core neural engine
Secure Enclave
USB 4

Products:
$999 ($899 edu) 13" MacBook Air (fanless) - 18 hour video playback battery life
$699 Mac mini (with fan)
$1299 ($1199 edu) 13" MacBook Pro (with fan) - 20 hour video playback battery life

Memory options 8 GB and 16 GB. No 32 GB option (unless you go Intel).

It should be noted that the M1 chip in these three Macs is the same (aside from GPU core number). Basically, Apple is taking the same approach which these chips as they do the iPhones and iPads. Just one SKU (excluding the X variants), which is the same across all iDevices (aside from maybe slight clock speed differences occasionally).

EDIT:



M1 Pro 8-core CPU (6+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 14-core GPU
M1 Pro 10-core CPU (8+2), 16-core GPU
M1 Max 10-core CPU (8+2), 24-core GPU
M1 Max 10-core CPU (8+2), 32-core GPU

M1 Pro and M1 Max discussion here:

Page 78 - Discussion - Apple Silicon SoC thread

Page 78 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M1 Ultra discussion here:

Page 109 - Discussion - Apple Silicon SoC thread

Page 109 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M2 discussion here:

Page 127 - Discussion - Apple Silicon SoC thread

Page 127 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

Second Generation 5 nm
Unified memory architecture - LPDDR5, up to 24 GB and 100 GB/s
20 billion transistors

8-core CPU

4 high-performance cores
192 KB instruction cache
128 KB data cache
Shared 16 MB L2 cache

4 high-efficiency cores
128 KB instruction cache
64 KB data cache
Shared 4 MB L2 cache

10-core iGPU (but there is an 8-core variant)
3.6 Teraflops

16-core neural engine
Secure Enclave
USB 4

Hardware acceleration for 8K h.264, h.264, ProRes

M3 Family discussion here:

Page 215 - Discussion - Apple Silicon SoC thread

Page 215 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

M4 Family discussion here:

Page 263 - Discussion - Apple Silicon SoC thread

Page 263 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[coercitiv]

Not now, but I am willing to bet money in the next 5 years we are going to get $700 macbook airs, or the ipads will gain far more features with software where they are effectively mini macbook airs running iOS.

Well...

 

[amrnuke]

Well...

View attachment 33759

Look here, you little...

The iPad Pro is already at a base price of $799. You didn't have to come in here and smear the innocent iPad Air!

 

[coercitiv]

The iPad Pro is already at a base price of $799. You didn't have to come in here and smear the innocent iPad Air!

Are you telling me you'd rather have the A12Z @ $799 over the crazy speed of A14 @ 749 with superior battery ife?

 

[name99]

I'll note that on the SIMD floating point/integer side AMD and Intel still has a significant advantage due to vector size and cache bandwidth.

This is far from clear.
(a) Apple (as far as we KNOW) "only" provides NEON, which is 128-bits wide.
BUT they provide 4 NEON units, so 512-bits worth of "compute" capacity.
It's hard to directly compare this with AMD or Intel because the NEON units have a much richer instruction set. Even for just raw FP, the NEON units provide flat FMA whereas x86 has always had this weird split between fp multiply and add units. Then there is the fact that Intel ships some servers (but not clients) with substantially larger AVX512 resources.

What we can say is that on GB4's SGEMM benchmark (basically dense linear algebra) the A14 matches Ice Lake:

LG Electronics IceLake Platform vs iPhone13,3 - Geekbench

(b) Apple DEFINITELY have the AMX units which substantially boost the linear algebra capabilities. They have still not made public exactly what these can do. They are now available via Accelerate, so someone could write various linear algebra routines of different sizes and float vs double, integer vs fp, to start scoping out the capabilities, and I'm sure that will happen soon after people get their M1 macs.

(c) LLVM has been working for some time on better support for both variable length vectors (SVE/2) and matrix manipulation. The matrix work is led by Apple people, and they have contributed to the SVE/2 work.
It is reasonable, IMHO, to assume that AMX support will make it to XCode once this work is complete.
It is reasonable to assume that SVE/2 will be in future Apple cores.
It is POSSIBLE that SVE/2 is there already, just not visible till the compiler support is considered good enough by Apple.

(d) The value of SVE is not really the wider vectors; it is the much richer set of operations available, which allows for more complicated loops to be vectorized, and for much loop overhead to be removed.

(e) Cache bandwidth is also unclear. Apple keep boosting their load-store capabilities, and they can do more than you imagine with what look like weaker resources because they have such powerful instructions available (load/store paired and the SoA/AoS load/stores). I'd wait for the graphs before concluding that they have lower cache bandwidth, I'd wait for the benchmarks before concluding that this mattered.

 

[name99]

Why is the vector performance so low on the intel CPU?

I've never heard of Affinity Photo. Is that benchmark constrained to 128 bit vectors or something?

Why do you bother asking these questions? Every time someone answers you, you just ignore the answers and continue believing what you want to believe. It's really tiresome.

 

[name99]

tablets don't really count as "computers" to me. They are consumption devices. And don't come with your magic-keyboard non-sense. It's still an extremely limiting device in terms of connectivity and applications.



So vendor lock-in is this great vision to redefine personal computing I should totally buy-in to?

I mean don't get me wrong, I'm 100% sure these ARM Macs will beat x86 in terms of performance/watt by a huge margin for many use cases. But as said tablets are already pretty limiting devices and being vendor-locked is just a no go for me. Your bascially limited to apples app-store for software and anything non-mainstream will directly be a problem.

EDIT:

And there is also this story about how apple treats their users privacy. Only a matter of time till a fixed installed keylogger that raises flags at the FBI if you type "dangerous" thoughts and no option to opt-out because closed-platform.

- Apple sends data home for every app you open, unencrypted
- Apple apps on macOS Big Sur bypass firewall and VPN connections to phone home

Jesus, dude, look at yourself. You're behaving like a Foxhead, desperately latching onto today's talking points for why we hate Apple (walled garden! don't own your computer!), believing insane conspiracy theories coming from Facebook.

I'm here to discuss technology. If you want to rant about lunacy or follow the crowd political opinions, don't bother doing it in response to my posts.

 

[name99]

The Mac Mini is the least expensive Mac, there is such a thing as market segmentation you know.

Eug's analysis is better.
The basic point is that ALL these machines are transitional. That means that one thing changes at a time. The SoC changes (requiring a whole bunch of electrical and FW changes) but the CASE does not. Same with the MBA and its lousy camera.

Why not change the cases at the same time? Because
(a) One thing at at time. Otherwise you fall flat on your face ala Intel.

(b) Changing an Apple case is not as trivial as people seem to think. Small changes are just a waste of time. And consider big changes. If you want to change the MBA camera, you change it to the iDevice camera, otherwise what's the point? That means you want FaceID. That means you need to change the bezel to fit. That means you have to change the aluminum casing and the machine milling. etc etc.

Same with the mac mini. The case and fan are what they are right now. Making a small change isn't worth the effort; you go with the case that you have (which has the whole production line set up to handle it) until you are ready for BIG changes --- like maybe make it a third thinner and fanless?
It seems clear that a mac mini pro is coming; and personally for my use cases I'd be happy to leave the fan in the mac mini pro and buy a mac mini fanless for my HTPC...

Yesterday I thought the fans in the MBP and Mac mini were there to allow the machines to engage in marathons without throttling. Today I think that's maybe nice occasionally, but not the central point; the central point is reuse of the existing cases with as little change as possible. MBA loses the fan to make a particular STRATEGIC point, but that (minimal) retooling effort only needs to be done for one device.

My current guess is that the next models of MBA, lowest end MBP, and mac mini will all be fanless. Apple will present this as "we've reduced the power even more", but really it's the evolution I'm describing here.

(BTW we're all expecting an M2 end of next year. I'm not sure.
I think an M1X for the midrange some time next year is definite, along with whatever is done for Mac Pro and iMac Pro based on the A15. But iPad's don't get an annual update and I don't think Mac's will.
Its possible that the first round will be updated, just to establish whatever's in M2as the "eal" baseline going forward. But I'd expect an M2 update as maybe 50/50. and after that annual updates as about 10% chance.)

 

[name99]

Also x86 isn't dying. There's one company that's competing with AMD/Intel and they aren't selling their SoCs to anyone else. Apple doesn't care about getting 50% of the market if they have to slit their wrists to do so. They'll occupy the luxury brand area and sell to 10% of consumers for $200 in profit as opposed to 90% for $20 profit.

x86 won't die, that's not the claim. What will happen is that it becomes ever more relegated to biches, like IBM's z machines. Important to people in those niches, but not the ecosystem you build upon when you start something new.

The argument, by the way, is not based on Apple, it's based as much on servers. Graviton and the other ARM servers will eat away at data warehouse usage for anything but specialty cases. Then they will take over super computers (already shown to be viable with Fugaku). Apple will grow in the high end desktop market unless you really need Windows native. Tablets/Chromebooks will eat away at the low end.

And as x86 shrinks (especially in the highest profit segments like big data warehouse cores and Apple-type PCs) there's less and less money available to keep paying for new designs.

M1 and Apple are nice. But as far as x86 is concerned, the date that matters is Nov 21, the products that matter are Neoverse.

 

[name99]

The CPU comparison is an 6-core, 6-thread i5-9600KF, correct?

Look down the thread. There are links to many benchmarks against a variety of systems with different core counts, including some dGPUs (which, when paired with the iGPU, do better than the M1 -- but not by especially much).

 

[name99]

Not now, but I am willing to bet money in the next 5 years we are going to get $700 macbook airs, or the ipads will gain far more features with software where they are effectively mini macbook airs running iOS.

• Yes this is a large timeframe ...
• but the way I see it is 10 years+1 month ago was macbook air version 2, and the ...
• ipad 1 was 10 years+7 months ago,
• thus we are talking predictions of the 3rd era if we break up this 10 year old market into 5 year segments.

In some ways we are living the Steve Jobs dream of Mac Pro capabilities in something you can put into an envelope. (Thank You Apple Silicon!)

View attachment 33756

I agree with you completely. Now that Apple has control, their laptop segmentation will follow the same path as phones and iPads and now watches.

And if someone thinks Apple can't make a nice MBA at $700, let me suggest that the problem is your definition of nice: The iPad is a very nice machine, even if not's what you *personally* want, many people agree that it's nice, and that niceness is available at $300 (in theory $330, but you can find discounts). That's achieved by omitting some features, but the parts that matter most to Apple, like the screen and the battery, are not at all garbage.

 

[Carfax83]

Why do you bother asking these questions? Every time someone answers you, you just ignore the answers and continue believing what you want to believe. It's really tiresome.

Dude you need to calm down. You get way too excited and irritable during these conversations

Nobody answered the question, probably because no one is familiar that software. I just found it odd that the Intel CPU had such low vector performance according to that benchmark, when vector performance has typically been one of their strongest assets.

That's why I wondered whether it was properly optimized for AVX/AVX2. From the performance, I doubt that it has been.

Case in point, in this thread, several users have uploaded their results. Here's a result for an AMD 2700x:

And here's one for a 9900K overclocked to 5ghz:

Now to me that's very weird, given the raw SIMD performance of both of these CPUs. The 9900K should blow the doors off of that 2700x since the 2700x uses 128 bit paths if I'm not mistaken.

 

[Roland00Address]

Well...
[256 iPad Air for 749 price]

Agreed that Apple is selling hardware that is a good price and performance for that price.

My point is Apple will be wanting to change their software in order to capture more marketshare from Windows especially since the $700 and $1000 are the two major price points you make money at.

Thus software is close on iOS but it is not there yet, some changes need to occur to make it more multitasking friendly but also to encourage some desktop / laptop apps.

The other way to capture this price point is over the next 5 years drop the price of the macbook airs.

Sidenote one comparison I do not like iPad Air / Pro is how there is no good Apple official keyboard that is bundled with the device. Sure you can buy a great keyboard from Apple but then the price is several hundred more (increasing total price.)

While you can always buy a $20 Bluetooth keyboard and a $10 Bluetooth mouse people who buy $749 iPad Airs want something better even if the $250 iPad basic ($330 normal price, but value customers buy it on sale already) are fine with very cheap 3rd Party Accessories.

What I am saying is Apple or 3rd Party needs something that fits into the two extremes if the goal is a $700 final cost device.

 

[JoeRambo]

Now to me that's very weird, given the raw SIMD performance of both of these CPUs. The 9900K should blow the doors off of that 2700x since the 2700x uses 128 bit paths if I'm not mistaken.

ZEN2 uses 256 bit datapaths. But it is not only about widths. But also about what each pipe is capable off ( and what workload is ).

Skylake (client) is traditional Intel design -> 2 Ports ( pipes) for FP, 256 width, but said ports are shared in scheduler with Integer and if they are busy, Integer workloads have only one port left. So while width is 2x256, it is good only for "peak" matrix workloads like Linpack, that are handtuned and memory bw or bw/lat product bound.
Real world code is different, not as optimized and contains a mix of instructions and vectorization is rather exception, than rule. In these workloads, Intel Skylake can only do same 2 multiplications or adds per clock.
in practice this looks like:
864 SSE :ADDPS xmm, xmm L: 1.00ns= 4.0c T: 0.13ns= 0.50c
865 AVX :VADDPS xmm, xmm, xmm L: 1.00ns= 4.0c T: 0.13ns= 0.50c
( same 0.5c thr. for MUL)
2 per clock, but try executing
870 SSE :MULSS+ADDSS xmm, xmm L: 2.00ns= 8.0c T: 0.21ns= 0.85c
871 AVX :VMULSS+VADDSS xmm, xmm, xmm L: 2.00ns= 8.0c T: 0.22ns= 0.87c
872 SSE :MULPS+ADDPS xmm, xmm L: 2.00ns= 8.0c T: 0.23ns= 0.92c
873 AVX :VMULPS+VADDPS xmm, xmm, xmm L: 2.00ns= 8.0c T: 0.23ns= 0.92c


Can barely sustain above 1 IPC in such mixed codes.

Want to see what ZEN3 can do ?

864 SSE :ADDPS xmm, xmm L: 0.88ns= 3.0c T: 0.15ns= 0.50c
865 AVX :VADDPS xmm, xmm, xmm L: 0.88ns= 3.0c T: 0.15ns= 0.50c
( same 0.5c thr. for MUL)
so far so good, same 2 per clock, buuuuuuuuuuuut:

870 SSE :MULSS+ADDSS xmm, xmm L: 1.77ns= 6.0c T: 0.10ns= 0.32c
871 AVX :VMULSS+VADDSS xmm, xmm, xmm L: 1.77ns= 6.0c T: 0.12ns= 0.42c
872 SSE :MULPS+ADDPS xmm, xmm L: 1.77ns= 6.0c T: 0.11ns= 0.39c
873 AVX :VMULPS+VADDPS xmm, xmm, xmm L: 1.77ns= 6.0c T: 0.10ns= 0.34c

2.5 - 3 damn instructions per clock. In mixed FP workload (that is not FMA), -> is either "scalar" or vector 32bit and mixes instructions, AMD gets amazing throughput. it is usually not as bad for Intel due to other core resource limitations, but the point is that 2014 era hw can't compete with 2020 hw.

In things like Cinebench, where there are also ton of Integer ALU ops involved, things get even worse for Intel, cause when port is busy with FP during that clock it cannot serve INT. The fact that Rocket Lake can almost match ZEN3 score in CB20 is actually huge, means perf parity in other workloads.

And Apple A14 is even wider in execution ports and more relevant for GB5 type workloads and if native CB20 comes out for ARM, they will shine there as well.

 

[Hitman928]

ZEN2 uses 256 bit datapaths. But it is not only about widths. But also about what each pipe is capable off ( and what workload is ).

Skylake (client) is traditional Intel design -> 2 Ports ( pipes) for FP, 256 width, but said ports are shared in scheduler with Integer and if they are busy, Integer workloads have only one port left. So while width is 2x256, it is good only for "peak" matrix workloads like Linpack, that are handtuned and memory bw or bw/lat product bound.
Real world code is different, not as optimized and contains a mix of instructions and vectorization is rather exception, than rule. In these workloads, Intel Skylake can only do same 2 multiplications or adds per clock.
in practice this looks like:
864 SSE :ADDPS xmm, xmm L: 1.00ns= 4.0c T: 0.13ns= 0.50c
865 AVX :VADDPS xmm, xmm, xmm L: 1.00ns= 4.0c T: 0.13ns= 0.50c
( same 0.5c thr. for MUL)
2 per clock, but try executing
870 SSE :MULSS+ADDSS xmm, xmm L: 2.00ns= 8.0c T: 0.21ns= 0.85c
871 AVX :VMULSS+VADDSS xmm, xmm, xmm L: 2.00ns= 8.0c T: 0.22ns= 0.87c
872 SSE :MULPS+ADDPS xmm, xmm L: 2.00ns= 8.0c T: 0.23ns= 0.92c
873 AVX :VMULPS+VADDPS xmm, xmm, xmm L: 2.00ns= 8.0c T: 0.23ns= 0.92c


Can barely sustain above 1 IPC in such mixed codes.

Want to see what ZEN3 can do ?

864 SSE :ADDPS xmm, xmm L: 0.88ns= 3.0c T: 0.15ns= 0.50c
865 AVX :VADDPS xmm, xmm, xmm L: 0.88ns= 3.0c T: 0.15ns= 0.50c
( same 0.5c thr. for MUL)
so far so good, same 2 per clock, buuuuuuuuuuuut:

870 SSE :MULSS+ADDSS xmm, xmm L: 1.77ns= 6.0c T: 0.10ns= 0.32c
871 AVX :VMULSS+VADDSS xmm, xmm, xmm L: 1.77ns= 6.0c T: 0.12ns= 0.42c
872 SSE :MULPS+ADDPS xmm, xmm L: 1.77ns= 6.0c T: 0.11ns= 0.39c
873 AVX :VMULPS+VADDPS xmm, xmm, xmm L: 1.77ns= 6.0c T: 0.10ns= 0.34c

2.5 - 3 damn instructions per clock. In mixed FP workload (that is not FMA), -> is either "scalar" or vector 32bit and mixes instructions, AMD gets amazing throughput. it is usually not as bad for Intel due to other core resource limitations, but the point is that 2014 era hw can't compete with 2020 hw.

In things like Cinebench, where there are also ton of Integer ALU ops involved, things get even worse for Intel, cause when port is busy with FP during that clock it cannot serve INT. The fact that Rocket Lake can almost match ZEN3 score in CB20 is actually huge, means perf parity in other workloads.

And Apple A14 is even wider in execution ports and more relevant for GB5 type workloads and if native CB20 comes out for ARM, they will shine there as well.

Zen 2700x is Zen1+, not Zen2. Zen1+ has 128 bit width.

CB23 is already out with Apple Silicon support.

Edit: I also saw that the makers of DaVinci Resolve said that they'd have a version with Apple Silicon support ready for the new Mac launch.

 

[amrnuke]

Look down the thread. There are links to many benchmarks against a variety of systems with different core counts, including some dGPUs (which, when paired with the iGPU, do better than the M1 -- but not by especially much).

There are some very curious results in that thread for vector results.

M1
ST: 504
MT: 2032

3700X
ST: 284
MT: 1567

10600
ST: 294
MT: 1578

9900K
ST: 315
MT: 1837

3900X
ST: 295
MT: 2079

We know the 10900K is at most 15-25% off an A14 in single-threaded workloads (SPEC 2006, GB5). Yeah, the 9900K is slower than the 10900K but not by that much. For the ST vector score of a 9900K (overclocked!) to be off by 60% from the M1 makes little sense. It also makes little sense that a 3900X with 12 cores and 24 threads sees an MT score only 7 times its ST score.

I'll be waiting for other benchmarks. This one seems like quite the outlier, or at least too specialized/niche to extrapolate to anything meaningful for general comparison.

 

[JoeRambo]

Zen 2700x is Zen1+, not Zen2. Zen1+ has 128 bit width.

Yeah, You are right, still the same more and more flexible execution ports paradigm aplies to it as well:

870 SSE :MULSS+ADDSS xmm, xmm L: 1.62ns= 6.0c T: 0.13ns= 0.50c
871 AVX :VMULSS+VADDSS xmm, xmm, xmm L: 1.62ns= 6.0c T: 0.16ns= 0.58c
872 SSE :MULPS+ADDPS xmm, xmm L: 1.62ns= 6.0c T: 0.13ns= 0.50c
873 AVX :VMULPS+VADDPS xmm, xmm, xmm L: 1.62ns= 6.0c T: 0.16ns= 0.58c

Except of course this is best case and performance falls off in 256bit

1856 AVX :VMULPD+VADDPD ymm, ymm, ymm L: 1.90ns= 7.0c T: 0.27ns= 1.00c
Where ZEN3 advanced to:
1856 AVX :VMULPD+VADDPD ymm, ymm, ymm L: 1.77ns= 6.0c T: 0.10ns= 0.33c

 

[Doug S]

My point is Apple will be wanting to change their software in order to capture more marketshare from Windows especially since the $700 and $1000 are the two major price points you make money at.

If they do the same thing they do with mobile devices they'll be able to fill out their lineup with less expensive devices over time. For the last five years or so when Apple introduces new iPhones they sell last year's model at $100 off, and the low end of the two years ago model at $200 off.

I could easily see them doing the same thing with at least the Macbook Air / Mini. So in 2022 today's $699 Mini might be $499 (or you can get a better one with an A16 based CPU for the full $699) and the Air might be $799 for an M1 based model.

They might not follow exactly that plan, and maybe it would happen over a longer horizon, but I think they'll do something like that.

 

[Heartbreaker]

If they do the same thing they do with mobile devices they'll be able to fill out their lineup with less expensive devices over time. For the last five years or so when Apple introduces new iPhones they sell last year's model at $100 off, and the low end of the two years ago model at $200 off.

I could easily see them doing the same thing with at least the Macbook Air / Mini. So in 2022 today's $699 Mini might be $499 (or you can get a better one with an A16 based CPU for the full $699) and the Air might be $799 for an M1 based model.

They might not follow exactly that plan, and maybe it would happen over a longer horizon, but I think they'll do something like that.

Doubt it. At best they might drop $100.

IMO, iPads are intended to be the lower end computer.

 

[amrnuke]

If they do the same thing they do with mobile devices they'll be able to fill out their lineup with less expensive devices over time. For the last five years or so when Apple introduces new iPhones they sell last year's model at $100 off, and the low end of the two years ago model at $200 off.

I could easily see them doing the same thing with at least the Macbook Air / Mini. So in 2022 today's $699 Mini might be $499 (or you can get a better one with an A16 based CPU for the full $699) and the Air might be $799 for an M1 based model.

They might not follow exactly that plan, and maybe it would happen over a longer horizon, but I think they'll do something like that.

With their phones, with the iPhone 11 and iPhone SE last year, they both had the A13. And the iPhone 12 mini at $699 has the A14 just like the $1099 iPhone 12 Pro Max. I think these chips are just so cheap for them that it makes little sense to fragment things.

 

[kurosaki]

Go and ask for a low-end Ferrari in a respectable shop, see how fast the door finds you.


Nobody asked Apple to sell their MBA for $700, instead somebody argued the MBA is low-end, which is not true.

Ferrarri 400i goes crying in the corner

 

[amrnuke]

Ferrarri 400i goes crying in the corner

There are plenty of cheap Ferraris, cf:

 

[amrnuke]

Are you telling me you'd rather have the A12Z @ $799 over the crazy speed of A14 @ 749 with superior battery ife?

No, there is definitely no value to adding 2 high performance cores and doubling the number of GPU cores, all that matters is single-core speed!!!

I do wonder how different the M1 performance ends up being when compared to a theoretical A14Z.

 

[name99]

Doubt it. At best they might drop $100.

IMO, iPads are intended to be the lower end computer.

That is not the way Apple thinks. They have said this hundreds of time, but people gonna believe what they want to believe...

Apple do not see their products competing with each other, they complement each other. The ideal Apple customer (and most of them are ideal, as close as their budget will allow) does not ask "should I get a watch OR a phone", they get both, because they do slightly different jobs. They do not ask "should I get an iPad OR a macbook, again they get both and use each as best for the job at hand".

 

[Roland00Address]

I do wonder how different the M1 performance ends up being when compared to a theoretical A14Z.

What A14Z in theory, pretty much a M1 would map just onto A14Z almost 1 to 1.

A12Z was 4 big cores, 4 little cores, 8 gpu cores, and can go up to 16gb of ram with the mac mini developer edition for Big Sur.
M1 ___ is 4 big cores, 4 little cores, 8 gpu cores, and can go up to 16gb of ram with various macbook air, pro, and mac mini editions.

Pretty much we are seeing what a die shrink gets you but also the differences between Firestorm plus Icestorm cores instead of Tempest plus Vortex. Branding this new chip M1 instead of A14Z is merely "branding and marketing."

 

[coercitiv]

No, there is definitely no value to adding 2 high performance cores and doubling the number of GPU cores, all that matters is single-core speed!!!

For a light productivity device combined with media consumption? Yes, faster high performance cores, faster low performance cores and double the storage will be the better choice.