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

 

[FlameTail]

I doubt it.

I think maybe more than like Intel and AMD is quite possible, but I think they’ll go for a medium point on efficiency.

Indeed. They took the same approach with the GPU. It's more efficient than Intel/AMD but behind Apple.

 

[SpudLobby]

Indeed. They took the same approach with the GPU. It's more efficient than Intel/AMD but behind Apple.

Which I mean if we are being honest, this is exactly what we expected. I didn’t expect Qualcomm to literally match Apple in efficiency or battery life when we compare products with similar core counts and similar process nodes because Apple can spend more and still has architectural inertia.

That doesn’t mean the gap will be large however. The fundamentals of this platform are still assuredly going to be much, much closer to Apple’s products than they are AMD/Intel mobile stuff and the early demos and information suggest that. Also, they’ve got room to run too — we don’t actually have their full E cores yet for example.

We all know AMD and Intel leave a ton on the table from idle power to very low power operation to like, MT and ST efficiency throughout 5-30W (especially at the lower end of that obviously). And therefore battery life in “casual use” with bursty ST, and a few non-priority background tasks are where they look much worse than the legendary “Cinebench @ 15-30W” stuff.

 

[FlameTail]

Which I mean if we are being honest, this is exactly what we expected. I didn’t expect Qualcomm to literally match Apple in efficiency or battery life

Why should one not have expected X Elite's GPU to be more efficient than M2/M3's ?

It's a reasonable expectation, because Qualcomm's mobile GPUs are actually more efficient than Apple's (example: 8 Gen 3 vs A17 Pro; Geeekerwan review).

 

[SpudLobby]

Why should one not have expected X Elite's GPU to be more efficient than M2/M3's ?

It's a reasonable expectation, because Qualcomm's mobile GPUs are actually more efficient than Apple's (example: 8 Gen 3 vs A17 Pro; Geeekerwan review).

I realize that yes. But as for why, area. M3 is almost double (or 66% more) GPU cores than the iPhone 15 and M2 is double the 4/5c A15 stuff at 8/10. Apple does still raise clocks some ofc.

I don’t think Qualcomm added as many units, they added some I believe to the 740 but most of it seems like the result of basically doubling the frequency from the 660-750mhz range to 1.4GHz from the rumors, which brings us from that 2-2.2tflops range to the 4.6TFLOPs, albeit obviously perf/w will suffer a bit.

I think at 10-15W fwiw the Adreno in the X Elite is still going to blow out AMD and Intel or get just a bit ahead of Strix and Lunar Lake at worst. Supposedly Lunar GPU at 12W just matches the M1’s 2.6 TFLOPS — that I am confident Qulcomm will do better than, at least permitting quality drivers and synthetic tests.

 

[igor_kavinski]

The 8GB peasants get their SSD speed back: https://www.tomshardware.com/laptop...256gb-ssd-drastically-faster-than-predecessor

Yay?

 

[igor_kavinski]

Peasants can upgrade to 2TB SSD and avoid paying Apple. Most definitely a WIN!

Repair shop offers MacBook M-series storage upgrades: Up to 2TB at up to $550 (Updated)

Repair shop offers SSD upgrade service for M-series MacBooks

www.tomshardware.com

 

[poke01]

It's a reasonable expectation, because Qualcomm's mobile GPUs are actually more efficient than Apple's (example: 8 Gen 3 vs A17 Pro; Geeekerwan review).

Are they efficient because they focus solely on gaming? Qualcomms GPUs are not compute heavy.

It will be very interesting to see Qualcomms GPUs in a PC environment.

 

[FlameTail]

Are they efficient because they focus solely on gaming? Qualcomms GPUs are not compute heavy.

It will be very interesting to see Qualcomms GPUs in a PC environment.

See this:

Page 22 - Discussion - Qualcomm Snapdragon Thread

Page 22 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

forums.anandtech.com

 

[Eug]

The 8GB peasants get their SSD speed back: https://www.tomshardware.com/laptop...256gb-ssd-drastically-faster-than-predecessor

Yay?

Most of the entry level peasants never cared in the first place.

It was already plenty fast for 95% of that market. The complainers were mostly YouTubers, journalists, and forum geeks, despite the fact that most of them would buy the 512 GB model for more storage anyway.

That does make me wonder why Apple changed it back to two chips though.

 

[FlameTail]

This redditor claims Apple M1-M3 uses BSPDN

https://www.reddit.com/r/hardware/comments/1bb1onr/comment/kucoi1p

https://www.reddit.com/r/hardware/comments/1bb1onr/comment/kucn65p

Do you guys call bull on these comments?

 

[Hitman928]

This redditor claims Apple M1-M3 uses BSPDN

https://www.reddit.com/r/hardware/comments/1bb1onr/comment/kucoi1p

https://www.reddit.com/r/hardware/comments/1bb1onr/comment/kucn65p

Do you guys call bull on these comments?

He has no clue what he's talking about.

 

[FlameTail]

In his defense, he claims he works in the industry and also used to work on Qualcomm'a Oryon.

Hey...

Didn't somebody just post an Apple patent here a few days ago about using BSPDN on a FinFET process?

Is this it?

Connecting the dots... [Galaxybrain.gif]

 

[Hitman928]

In his defense, he claims he works in the industry and also used to work on Qualcomm'a Oryon.

Hey...

Didn't somebody just post an Apple patent here a few days ago about using BSPDN on a FinFET process?

Is this it?

Connecting the dots... [Galaxybrain.gif]

You mean people can lie about who they are on the internet? /s

I can tell you by the terms he is using and the way he is putting them together, he does not work in the industry, or at least not as a real designer and he doesn't really know what he is talking about.

I'll put it this way, if he was a part of the early Oryon design team, he would be working with the very people who designed the M1, so why would the team have to rely on a die tear down to find out about M1's power network and not just literally ask the guys on the team who helped design it?

 

[richardskrad]

Speedometer 3.0 has been released (by contributors from Apple, Google, Microsoft, and Mozilla). https://www.browserbench.org/Speedometer3.0/

I ran a test on my M1 MacBook AIr and got 23.0.

 

[Eug]

Speedometer 3.0 has been released (by contributors from Apple, Google, Microsoft, and Mozilla). https://www.browserbench.org/Speedometer3.0/

I ran a test on my M1 MacBook AIr and got 23.0.

My M1 results are consistent with yours: 22.76 (+/-0.18) - M1 Mac mini / Sonoma 14.3.1 / Safari 17.3.1

However, it is consistently faster in Chrome: 26.1 (+/-0.38) - M1 Mac mini / Sonoma 14.3.1 / Chrome 122

 

[FlameTail]

Apple vs Qualcomm

Who has the more area efficient GPU?



Snapdragon 8 Gen 3: 137 mm²
A17 Pro: 103 mm²

 

[FlameTail]

Sorry that's not an iso node comparison.

Here's A16 Bionic (113mm²) vs 8 Gen 2 (119mm²). Both are on TSMC N4.



Both 8 Gen 2 GPU and A16's GPU consume 8W of power in this test by Geekerwan. But 8 Gen 2 is more performant.

 

[okoroezenwa]

My M1 results are consistent with yours: 22.76 (+/-0.18) - M1 Mac mini / Sonoma 14.3.1 / Safari 17.3.1

However, it is consistently faster in Chrome: 26.1 (+/-0.38) - M1 Mac mini / Sonoma 14.3.1 / Chrome 122

Is it best to have a clear system (no other tabs in Safari + apps closed) to run this? I'm getting 19.0 (+/- 0.79) on my M1 Pro which seems pretty low compared to others.

 

[igor_kavinski]

Is it best to have a clear system (no other tabs in Safari + apps closed) to run this? I'm getting 19.0 (+/- 0.79) on my M1 Pro which seems pretty low compared to others.

Turn on High Power mode.

 

[Hitman928]

Is it best to have a clear system (no other tabs in Safari + apps closed) to run this? I'm getting 19.0 (+/- 0.79) on my M1 Pro which seems pretty low compared to others.

I suggest you run in a private window with all add-ons/extensions turned off and everything else closed (or at least minimized if you don't won't to close it).

 

[Doug S]

Is it best to have a clear system (no other tabs in Safari + apps closed) to run this? I'm getting 19.0 (+/- 0.79) on my M1 Pro which seems pretty low compared to others.

I'd plug it in (to eliminate any sort of power saving that might happen when battery is below a certain level) and reboot before testing and you should see similar results to what others are getting.

Could also be software related, if they are running the latest OS and yours is behind on OS version/patches.

 

[Eug]

Is it best to have a clear system (no other tabs in Safari + apps closed) to run this? I'm getting 19.0 (+/- 0.79) on my M1 Pro which seems pretty low compared to others.

Judging by those M1 Pro results, am I correct to assume this test is mainly just single-core?

But yeah, reboot and close everything you don't need.

I'd plug it in (to eliminate any sort of power saving that might happen when battery is below a certain level) and reboot before testing and you should see similar results to what others are getting.

Could also be software related, if they are running the latest OS and yours is behind on OS version/patches.

Ironically, my MacBook Core m3 did better after I unplugged it, perhaps because the heat from the battery charging affected performance in this fanless machine.

Performance was still poor though of course:

6.24 - 1.2 GHz Core m3-7Y32 / Ventura 13.6.5 / Safari 17.4
7.15 - 1.2 GHz Core m3-7Y32 / Ventura 13.6.5 / Chrome 122

 

[Thibsie]

M3 MacBook Air released, and it finally supports two external monitors (if the lid is closed).

Meanwhile, I’ve been using my m3 for 7 years already.

That's still a joke IMO.
Stupid market segmentation.

 

[Thibsie]

They still have a Mac to sell you for back to school whether it was released 12 months ago or 12 minutes ago, and most of those buyers just want to buy the Mac without caring too much about the specs. Even if it's a year old model at the time of purchase, it's still going to last all 4 years of school and probably several more after that, because it's a Mac and will have good build quality and support.

I dunno about current macs but good quality hasn't always been the case.
My MacBook late 2008 was plagued by defective gigabit NIC, one year warranty 'cos Apple didn't give a damn about legal European warranty (2 years) and no, they wouldn't replace your MacBook even with the defective NIC (thanks Nvidia bump gate). Also the screen was a piece of crap, not even trying to fake half sRGB colour reproduction.

Hope they learnt their lesson.

 

[Thibsie]

Let that un-upgradable hardware rot on the shelves, unless it's available for $499.

Buy a normal, more capable x86 laptop and put on it a MacOS lookalike Linux distro: https://itsfoss.com/macos-like-linux-distros/

No, just plain MacOS