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

 

[DavidC1]

Poor Intel, Apple just destroyed hard. Panther Lake should be a wonder if not, Apple would make a massive check mate for them.

If this isn't further indication that x86 vendors could do MUCH better, I don't know what it is.

80% faster than Lion Cove per clock that can fit in phones....

 

[Doug S]

Have you seen the nonsense that flows from this?

https://en.eeworld.com.cn/news/xfdz/eic711527.html

"
The media pointed out that the cost of Apple's A18 chip is estimated to be about US$45, while the cost of the A19 chip is estimated to be US$150, and the A20 chip may further increase to US$280, an increase of 86.67%, replacing the camera as the most expensive single component in the iPhone.
"

THIS is what I am complaining about. Utterly deranged claims that make absolutely no sense. This chatter is not useful to anyone, whether you're a consumer, an investor, or an engineer. It's just pure and total pollution of the infosphere.

Any individual site can make up whatever numbers they want, and they don't have to make sense. Maybe they have an agenda, maybe they are just stupid, maybe they are gullible and believe the wrong people (who either have an agenda or are stupid) If even the slightest critical thinking was applied it would be hard to figure the possible mechanism where going from N3E to N3P causes the SoC price to more than triple...

No one who matters will believe those numbers. Those who do can sh--post them in forums and amuse themselves, but it isn't going to cause Apple to raise iPhone 18 prices to by 50% to make up for those hugely expensive SoCs.

 

[jdubs03]

Some more M5 benchmarks:

 

[jdubs03]

I was checking this out and it got me curious.

I was under the impression that Speedometer doesn’t stress the CPU to the point of throttling.

But a slight frequency bump isn’t going to make nowhere near that difference.

I was checking out Computerbase’s review of the M5 iPad Pro, and they got that same 49 score. With the M4 getting 33. My 16 Pro Max (as noted recently got a 38 in Safari). Something is off there.

 

[Nothingness]

I was checking out Computerbase’s review of the M5 iPad Pro, and they got that same 49 score. With the M4 getting 33. My 16 Pro Max (as noted recently got a 38 in Safari). Something is off there.

Different security measures at OS/app level (among which MTE)?

 

[jdubs03]

Different security measures at OS/app level (among which MTE)?

But with the M4 vs A18 Pro comparison it’s still a win for my 16PM which to me didn’t make any sense.

I’m assuming the M5 has the same MTE as the A19P.

Maybe it’s a form factor thing? But still if not throttling, why so much lower on the iPad form factor.

Edit: In Max Techs video their M4 MacBook Pro scores 50. So it’s even more perplexing how an M4 in that form factor, could out score an M5 even in a slim form factor, assuming no throttling.. which it shouldn’t.

 

[Nothingness]

But with the M4 vs A18 Pro comparison it’s still a win for my 16PM which to me didn’t make any sense.

I’m assuming the M5 has the same MTE as the A19P.

Even if a feature such as MTE is available in HW it doesn't mean it's being used by the OS, there could be a difference between iOS and macOS. But that's just an hypothesis

 

[Doug S]

I was under the impression that Speedometer doesn’t stress the CPU to the point of throttling.

My 16 Pro Max (as noted recently got a 38 in Safari). Something is off there.

Maybe running Speedometer on your phone after it has spent an hour in the freezer while laying in a pile of ice cubes would bump that 38 by enough to show that Speedometer throttles after all?

 

[jdubs03]

Maybe running Speedometer on your phone after it has spent an hour in the freezer while laying in a pile of ice cubes would bump that 38 by enough to show that Speedometer throttles after all?

I have noticed that the score decreases from its max after a few runs. But for the record, no freezer was used lol, just a normal room temperature run.

 

[The Hardcard]

I was checking this out and it got me curious.
View attachment 132631
I was under the impression that Speedometer doesn’t stress the CPU to the point of throttling.

But a slight frequency bump isn’t going to make nowhere near that difference.

I was checking out Computerbase’s review of the M5 iPad Pro, and they got that same 49 score. With the M4 getting 33. My 16 Pro Max (as noted recently got a 38 in Safari). Something is off there.

The throttling could be because of the device rather than the SOC itself. It is possible that they are protecting the iPad skin temperatures. Apple’s ability to ship fanless M devices is even more impressive given how many don’t seem to have even basic passive cooling implemented. I haven’t watched any teardowns of the new iPads, but now I’m curious.

Because of the photo pipeline, A series gets 3 times the SLC as the base M. I wonder how that may affect Speedometer.

 

[Eug]

The more reviews I see, the more it seems @mvprod123 is correct. M5 has significantly improved hardware media encoding speed. M5 is faster than M4 Pro. (Both the M5 and M4 Pro have only a single media encoder.)



M4 Pro also runs hotter and uses much more battery than M5, not surprisingly.



Internal SSD speed is significantly faster on M5 than M4 Pro, but of course M4 Pro has Thunderbolt 5.

 

[johnsonwax]

I keep saying that asymmetrical compute is where it's going to be going forward. There's only so much you can squeeze out of general compute units, but you can get a TON of dedicated media compute, AI compute, etc. GPU was an example of that and now it's deep into diminishing returns as well (but gets its own asymmetrical benefits through ray tracing units, etc.) I keep wondering if we'll see a return of physics compute engines to the game consoles. The idea was sound when they were first introduced, but the applications weren't there - it was too early.

 

[poke01]

some memory bandwidth benchmarks on M4.
quite close to the theoretical peak of 120GB/s

 

[name99]

I keep saying that asymmetrical compute is where it's going to be going forward. There's only so much you can squeeze out of general compute units, but you can get a TON of dedicated media compute, AI compute, etc. GPU was an example of that and now it's deep into diminishing returns as well (but gets its own asymmetrical benefits through ray tracing units, etc.) I keep wondering if we'll see a return of physics compute engines to the game consoles. The idea was sound when they were first introduced, but the applications weren't there - it was too early.

There are also weird things you can imagine that seem plausible but no-one has done them (yet).

For example consider PiM, back in the headlines because of QC's announcement yesterday. Full scale PiM is tough because logic technology is so different from memory technology, but consider a partial solution -- what if you added some degree of logic to the SLC? At first this might just do things like move data and flood-fill data, but later maybe it could also handle simple ALU ops. With something like this you could prototype APIs and integration, until genuine PiM is available.

Vision processors may also make a comeback (not an ISP, which is camera -> good looking JPG; Vision processor is more camera -> primitives that are ready to feed into an AI process of some sort, so eg segmentation and face detection as simple examples) That may a large part of what the R1 is?

 

[StinkyPinky]

Would be interesting to see what an M5 Mini could do with more thermal headroom.

m5 Pro mini could potentially be a beast. Come on Apple, release it already.

 

[mvprod123]

https://twitter.com/x/status/1983276784947958038

 

[jdubs03]

iOS 26.1 RC gives a nice little bump to performance and battery life according to Brandon Butch.

 

[mvprod123]

 

[poke01]

This is funny

 

[Eug]

I just got an LG 6K 6144x3456 monitor. My M4 Mac mini has display options of 3072x1728 (2X scaled) and 6144x3456 (no scaling), but nothing in-between.



This is in contrast to the other higher end M4 series machines which apparently have other options like 3200x1800 and 3360x1890.



I don't know if this is a hardware limitation or a software limitation, and I wonder if M5 has the same limitation.

 

[poke01]

Get BetterDisplay from GitHub

 

[mvprod123]

View attachment 132899
This is funny

Intel's CPU design is shit

 

[Eug]

Get BetterDisplay from GitHub

I don’t actually need those resolutions which is why I haven’t installed BetterDisplay. However, I’ll note that this is what some Macs get with BetterDisplay on the Dell 6K.



Lots of HiDPI resolutions below 3072 but none above 3072.

 

[poke01]

I don’t actually need those resolutions which is why I haven’t installed BetterDisplay. However, I’ll note that this is what some Macs get with BetterDisplay on the Dell 6K.

View attachment 132918

Lots of HiDPI resolutions below 3072 but none above 3072.

Ugh Apple stuffed up their display support for M4

 

[StinkyPinky]

I wish they made a 42" version of that monitor.