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

M5 Family discussion here:

Page 431 - Discussion - Apple Silicon SoC thread

Page 431 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[mavere]

That's interesting because it appears to be so much larger than the M4 - 18 to 35% improvement, depending on which random reddit post you believe; the most recent comparable value I can find is
(25 days ago)"
Not sure.
Running Speedometer 3.1 I get 44.1(-+1.7) on an M4 MacBook Pro Safari Version 26.0 on Tahoe.
"

This would suggest that my trace cache theory is correct (at least in the sense that it is present in the M5). Might be interesting to get some Speedometer 3.1 scores for A19 Pro vs A18 Pro...

Both SPEC and GB6 do very little to stress the front end of an Apple core, while browsers do a somewhat better job of this.

This review here also has a high ~62 score, but seems to indicate ~15% improvement in Speedometer 3.1 for M5 over M4. https://tbreak.com/apple-macbook-pro-m5-review-uae/

15% is higher than what I'd expect considering A19 over A18 though.

 

[Eug]

It's funny to read this thread and then get an m5 ipad and it's essentially indistinguishable from my m4 ipad (handing down to a family member, I'm not an insane person upgrading for no reason)

The wifi is much better, and the ram bump means apps will be swapped out of memory less often, but the cpu/gpu/npu improvements are basically meaningless, other than if the battery life is better from increased efficiency. Not sure yet if that'll be an appreciable difference

Yeah, the SoC performance difference on an M5 is completely meaningless to me for real world usage. Even a hypothetical M3 iPad Pro would have be fine performance-wise. I really just bought the M4 for the new keyboard, new tandem OLED display, and Face ID. The additional RAM would be a bonus, but it's sad we only finally get it in 2025, when it really should have been included on the M4 model (and it actually was, just with 4 GB deactivated).

In terms of battery life, they are rated the same, so while they may end up being a bit different, it shouldn't be by too much. The M5 does get much faster charging, but I never charge with more than 35 W anyway, so that feature would be lost on me. In fact, it would be lost on most people, since it comes with only a 20 W charger (or no charger at all in Europe).

For those who care though, the M5 version gets 4K 120 Hz external display support (along with its 6K 60 Hz), but that's still a pretty niche feature for an iPad Pro. The M4 supports only 60 Hz (both for 4K and 6K). I am looking forward to trying 6K 60 Hz on my M4 iPad Pro when I finally get the display, but I have zero intention of actually using the iPad Pro that way.

 

[jdubs03]

https://twitter.com/x/status/1981014969501331611

That’s pretty impressive. My M4 Mac Mini gets 47 (albeit with multiple windows open and a few Brave tabs). Maybe I’ll do a clean run to see how it goes.

Caveat, that score is in Safari.

 

[poke01]

the SoC performance difference on an M5 is completely meaningless to me for real world usage. Even a hypothetical M3 iPad Pro would have be fine performance-wise.

It’s just there are no large game libraries to take advantage of all the extra power on the M5 GPU and whatever AAA available games are capped at 30fps.

For those who care though, the M5 version gets 4K 120 Hz external display support (along with its 6K 60 Hz), but that's still a pretty niche feature for an iPad Pro. The M4 supports only 60 Hz (both for 4K and 6K). I am looking forward to trying 6K 60 Hz on my M4 iPad Pro when I finally get the display, but I have zero intention of actually using the iPad Pro that way

That’s a software limitation, the M4 does support 120hz on macOS. The fact one has to upgrade their whole iPad to get that feature is anti-consumer.


The iPad Pro lineup is pretty much meaningless SoC upgrades from on until they open up the operating system to allow for emulators that use JIT and allow installing outside the AppStore without apples blessing but that will never happen soon.

 

[jdubs03]

That's interesting because it appears to be so much larger than the M4 - 18 to 35% improvement, depending on which random reddit post you believe; the most recent comparable value I can find is
(25 days ago)"
Not sure.
Running Speedometer 3.1 I get 44.1(-+1.7) on an M4 MacBook Pro Safari Version 26.0 on Tahoe.
"

This would suggest that my trace cache theory is correct (at least in the sense that it is present in the M5). Might be interesting to get some Speedometer 3.1 scores for A19 Pro vs A18 Pro...

Both SPEC and GB6 do very little to stress the front end of an Apple core, while browsers do a somewhat better job of this.

I’ve seen as high as 38.1 on my 16 Pro Max. And I know one of our other members here got as high as 43.9 on a 17 Pro Max.

Btw clean run I got 48 on my M4 Mac Mini in Safari, MacOS 26.0.1.

 

[mvprod123]

PassMark

PassMark CPU Benchmarks - Single Thread Performance

Benchmarks of the single thread performance of CPUs. This chart comparing CPUs single thread performance is made using thousands of PerformanceTest benchmark results and is updated daily.

www.cpubenchmark.net

 

[Eug]

PassMark

PassMark CPU Benchmarks - Single Thread Performance

Benchmarks of the single thread performance of CPUs. This chart comparing CPUs single thread performance is made using thousands of PerformanceTest benchmark results and is updated daily.

www.cpubenchmark.net

View attachment 132491

It's hilarious that my phone is third highest in the world. And yes, it's Snappy™.

 

[sallymander]

I guess whatever weirdness happened with the M4 on passmark doesn't apply to A19/M5. The M4 scores were way below M3.

 

[mvprod123]

Still two NAND chips, but with twice the speed.

 

[digitaldreamer]

Reads are more than 3x the M4, according to that test!

 

[name99]

Hmmm... By most accounts, N3B was indeed quite expensive. Are you trying to deny this was the case?

I mean, a common refrain here was that a big part of M3's short life was due to N3B's cost.

I'm denying that the expense is of any significance to anyone outside Apple or TSMC.
OK, so it's more expensive? Why is that general news?
The claim is written not because it is news but because there is an agenda attached, and it is THAT which I am calling out. It's the same as the idiotic and always context-free claims about data center water use.

 

[Eug]

I'm denying that the expense is of any significance to anyone outside Apple or TSMC.
OK, so it's more expensive? Why is that general news?
The claim is written not because it is news but because there is an agenda attached, and it is THAT which I am calling out. It's the same as the idiotic and always context-free claims about data center water use.

That's a very strange take. If you're interested in these types of things, then it's news that may be appropriate for you. My wife couldn't care less what TSMC wafers and chips cost, but for example people who invest in tech companies might find it interesting for various reasons.

And no it doesn't apply to just Apple or TSMC, but to any company who might be considering using N2 relatively early on. However, even if it did just apply to Apple and TSMC, then if you're an investor in either Apple and TSMC, that info would be of interest of course.

 

[Eug]

Yes M5 GPU and AI performance are way, way better, but it also runs hotter, not surprisingly. @mvprod123, there are also some more video benches too here.

M5 on left, M4 on right

 

[mvprod123]

Yes M5 GPU and AI performance are way, way better, but it also runs hotter, not surprisingly. @mvprod123, there are also some more video benches too here.

M5 on left, M4 on right

View attachment 132549

View attachment 132548

View attachment 132550

View attachment 132551

View attachment 132552

View attachment 132553

Higher memory frequency and increased multi-core clock speed have resulted in increased power consumption. The GPU consumes approximately the same amount of power as the M4 GPU.

 

[The Hardcard]

Higher memory frequency and increased multi-core clock speed have resulted in increased power consumption. The GPU consumes approximately the same amount of power as the M4 GPU.

Luke Miani is the performance polar opposite to Geekerwan. He has for years gotten the lowest scores on nearly every benchmark for nearly every device, especially in GPU. I don’t know his testing conditions, but he seems to benchmark way closer to an oven than any freezer.

Case in point, the Geekbench browser shows numerous M4 AI GPU scores in the mid to high 9000s, some above10,000. He gets 6333. The M5 number is is very close to typical if he is using the quantized result, but that makes the difference much more dramatic than it actually is. It is another datapoint, but with a caveat.

 

[digitaldreamer]

Luke Miani is the performance polar opposite to Geekerwan. He has for years gotten the lowest scores on nearly every benchmark for nearly every device, especially in GPU. I don’t know his testing conditions, but he seems to benchmark way closer to an oven than any freezer.

Case in point, the Geekbench browser shows numerous M4 AI GPU scores in the mid to high 9000s, some above10,000. He gets 6333. The M5 number is is very close to typical if he is using the quantized result, but that makes the difference much more dramatic than it actually is. It is another datapoint, but with a caveat.

He probably does his tests immediately out of the box, without giving the system time to settle with the indexing and other tasks.

 

[name99]

That's a very strange take. If you're interested in these types of things, then it's news that may be appropriate for you. My wife couldn't care less what TSMC wafers and chips cost, but for example people who invest in tech companies might find it interesting for various reasons.

And no it doesn't apply to just Apple or TSMC, but to any company who might be considering using N2 relatively early on. However, even if it did just apply to Apple and TSMC, then if you're an investor in either Apple and TSMC, that info would be of interest of course.

I scoff because the claim is decoupled from anything that actually matters in reality.

Has the price of iPhones gone up since these horror stories about increasing TSMC costs started ten years ago? Well, I posted the graph, didn't I.

 

[name99]

I scoff because the claim is decoupled from anything that actually matters in reality.

Has the price of iPhones gone up since these horror stories about increasing TSMC costs started ten years ago? Well, I posted the graph, didn't I.

My bad, I posted the graphs on macRumors. Let me put them here.
From https://asymco.com/2025/09/10/hyper-tension/

 

[Eug]

I scoff because the claim is decoupled from anything that actually matters in reality.

Has the price of iPhones gone up since these horror stories about increasing TSMC costs started ten years ago? Well, I posted the graph, didn't I.

Hmmm… If that’s your main point, then you’ve totally missed my point.

It’s not necessarily just about prices that consumers see but also about BOM and gross margins, etc. This is interesting info for investors.

BTW, iPhone prices have indeed sometimes gone up over time, depending upon the model. They’ve hidden it for example by removing the 128 GB tier and selling the 256 GB as the new baseline but at the previous 256 GB price.

 

[Doug S]

Hmmm… If that’s your main point, then you’ve totally missed my point.

It’s not necessarily just about prices that consumers see but also about BOM and gross margins, etc. This is interesting info for investors.

BTW, iPhone prices have indeed sometimes gone up over time, depending upon the model. They’ve hidden it for example by removing the 128 GB tier and selling the 256 GB as the new baseline but at the previous 256 GB price.

What should be compared is the entry price for the lowest NAND. It has gone up over time, but I'd argue that the camera and display upgrades are a lot more responsible than their SoCs. Even at the current $20K per wafer with a 100 mm^2 die that's $33 each ignoring yield. If you assume yields have been roughly constant then a few generations ago when wafers were $10K that's an increase of only $16 in the BOM coming from the SoC.

Just look at the single tiny camera we had back in the day and now there are up to three large ones.

 

[Eug]

What should be compared is the entry price for the lowest NAND. It has gone up over time, but I'd argue that the camera and display upgrades are a lot more responsible than their SoCs. Even at the current $20K per wafer with a 100 mm^2 die that's $33 each ignoring yield. If you assume yields have been roughly constant then a few generations ago when wafers were $10K that's an increase of only $16 in the BOM coming from the SoC.

Just look at the single tiny camera we had back in the day and now there are up to three large ones.

My example of 128 GB to 256 GB was in iPhones that already had 3 cameras.

Anyhow a $16 increase is pretty significant. In recent years, just that increase alone would represent roughly 4% of total BOM. However, I agree it's just one of several factors.

BTW, one of the rumoured complaints of N3B was comparatively low yield for the first while, so I'm not sure it's appropriate to remove yield as a variable here.

 

[Doug S]

My example of 128 GB to 256 GB was in iPhones that already had 3 cameras.

Anyhow a $16 increase is pretty significant. In recent years, just that increase alone would represent roughly 4% of total BOM. However, I agree it's just one of several factors.

BTW, one of the rumoured complaints of N3B was comparatively low yield for the first while, so I'm not sure it's appropriate to remove yield as a variable here.

Yield didn't matter to Apple for N3B since they were paying for known good dies. All their other TSMC nodes have had good yields, so I think I'm on fairly solid ground talking about the number of candidates per wafer rather than trying to adjust the numbers by playing guess the yield. The "noise" from iPhone SoCs changing size (my "100 mm^2" was just sort of a round number / average of the range we've seen) is going to be a greater effect than any yield differences.

 

[Eug]

Xserve AI!

https://twitter.com/x/status/1981464918932279798