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

 

[SpudLobby]

Something like $1299 for an M4 Air with 16/512 is what I’d consider an honest and fair price on Apple Standards.

No chance. if they increase specs I would expect a $200 price increase at minimum.

Yeah this is actually what I had drafted. But $100-200 with M4 would probably be fair.

I wouldn’t expect them to do more than $200 though. It’s just past time for an update.

 

[okoroezenwa]

Something like $1299 for an M4 Air with 16/512 is what I’d consider an honest and fair price on Apple Standards.

If they go with my speculation of 8/512 starting then I guess they'd have met that with their '$200 for 8GB more' scheme 🤔

They need to move on that.

 

[Ajay]

If they go with my speculation of 8/512 starting then I guess they'd have met that with their '$200 for 8GB more' scheme 🤔

They need to move on that.

Apple's been doing that for so, so long that it won't change. It's just part of their marketing and profit system.

 

[Eug]

Slightly off topic, but the gift card promo started yesterday. Combined with the school discount it’s quite a deal.

Yes, as mentioned above, I just ordered an M4 iPad Pro with this promotion. Strange that this is the only way to get an M4 right now.

True. I’m just not confident they won’t do what they did with the iPad Pro and keep 4GB inaccessible for whatever reason.

I suspect/hope this is for marketing reasons. IOW, they wouldn't release a 12 GB base iPad Pro when their base MacBook Pro is still only 8 GB RAM.

Once the Macs go to 12 GB base, then they will do the same for the iPad Pros.

So, the way I think it will play out is as follows:

2024-2025 - M4 MacBook Pro 12 GB or 16 GB* (or 18 GB?) base.
2025 - M4 MacBook Air 12 GB base
2026 - M5 iPad Pro 12 GB base

*Going by just marketing reasons, I'd favour the M4 MacBook Pro to be 12 GB. However, the fact that Xcode predictive code completion requires 16 GB makes me wonder if they FINALLY will spec the base MacBook Pro with 16 GB.

Anyhow, my hope is the M4 MacBook Air with 12 GB is available by June 2025. I'd buy one of those during the 2025 Back-To-School Promotion to replace our 2017 Broadwell dual-core MacBook Air. It's just a kitchen machine and for most kitchen usage it's fine, but the lack of hardware HEVC decode support has recently been becoming a significant real-world issue. Even an M1 machine would be totally fine for this usage of course, but I'd rather get a machine with 12 GB RAM.

 

[FlameTail]

M4 Labelled Die Shot is here!

 

[Hitman928]

M4 Labelled Die Shot is here!

What's with the die size?

Edit: Display engine should be 2x display engine. Not sure what the block at the bottom of the NPU is supposed to be / why is it marked as part of the NPU? E-core cache should be wider if he's going to use the same method as the P-core cache. If the die size is supposed to be in cm, then I think he's underestimating the size but a bit.

 

[igor_kavinski]

Apple's been doing that for so, so long that it won't change. It's just part of their marketing and profit system.

If people weren't dumb, they wouldn't be able to do that. But the first rule of getting wealthy seems to be, start throwing away money on things that make you look rich which in my books is one of the dumbest things possible and reeks of an inferiority complex. I mean, if you are rich, why do you want to advertise it? Only reason I can think of is greed so they can find more opportunities to get even richer. Which is again dumb coz greed and wisdom go together like oil and water.

 

[okoroezenwa]

However, the fact that Xcode predictive code completion requires 16 GB makes me wonder if they FINALLY will spec the base MacBook Pro with 16 GB

This is exactly why I feel that model *has to* get 16GB base. Would just be weird otherwise. Could see a $100 increase accompanying that though.

 

[igor_kavinski]

Would just be weird otherwise.

Filthy rich Apple executives won't be able to fathom why a developer can't max out his credit card to get the more expensive 16GB model coz obviously they are gonna make 10x their investment.

 

[repoman27]

I wouldn't call it "overclocked", but "not JEDEC standard".

I imagine with iterations in DRAM processes over the last few years that at least some DRAM OEMs are able to produce LPDDR5 that can clock higher than the highest JEDEC speed of 6400 MT/s, just like they are able to produce DDR5 that clocks above whatever the fastest JEDEC speed is there. That latter is well known, because gamers/enthusiasts are always in the market for faster RAM.

I guess Apple was too, because they must have wanted more memory bandwidth for M4, but whether they chose this route because they didn't want to compromise on latency as LPDDR5X apparently would require, or a more prosaic reason like not having LPDDR5X controllers ready in time to make the M4 tape out, who knows.

This also explains why it had that odd -7700 speed or whatever it was, instead of -8533 which seems to be a widely available JEDEC standard LPDDR5X speed. It didn't make a lot of sense why they'd underclock LPDDR5X like that.

I think something may have gotten lost in translation with the Geekerwan "overclocked" comment. LPDDR5X is just higher clocked LPDDR5. Micron makes no distinction between the two in their datasheets, parts catalogs, and part numbering schemes.

From teardowns, we've seen the part numbers for the Micron SDRAM packages Apple has paired with the M4 SoC thus far. We have Z8DMS, which decodes to MT62F768M64D4AS-026 XT:B, and D8DNV, which decodes to MT62F1G64D4AS-026 XT:C. Consulting Micron's datasheets, the "-026" in those part numbers indicates this is bog standard LPDDR5X SDRAM with a rated cycle time of 267 ps, which works out to 7500 MT/s. With an 128-bit memory interface, LPDDR5X-7500 can provide 120 GB/s of memory bandwidth, which is exactly what Apple advertises for the M4.

I have no idea why Anandtech initially reported 7700 MT/s, but the article and table have since been updated to LPDDR5X-7500. I also have no idea what Geekerwan meant or why people interpreted that statement the way they did. There is zero evidence that Apple is doing anything in the slightest bit unusual with their memory choice for the M4, except of course for shipping 12 GB instead of 8 GB in the lower end models.

The most obvious answer as to why that happened is that Apple contracted for memory ahead of the product ramp, but when it came time to deliver, Micron was supply constrained on 8 Gbit dies. They offered Apple 12 Gbit dies at the contract price to fill the order and Apple agreed. However, it is unlikely that Micron would guarantee that deal for the life of the iPads Pro (5th/7th Generation), and even less likely that SK Hynix or Samsung would follow suit, and Apple tends to multi-source memory for all of their products whenever possible. That means that Apple couldn't advertise and sell the new iPads as having 12 GB of SDRAM unless they raised their retail prices to hedge against inevitable future SDRAM price increases. The simplest solution was to take the 6 GB packages from Micron, make 1/3 of it unavailable to the system, and market the device as having 8 GB as originally planned. And if it weren't for those pesky kids and their part number decoders, the internet would have been none the wiser.

 

[The Hardcard]

Filthy rich Apple executives won't be able to fathom why a developer can't max out his credit card to get the more expensive 16GB model coz obviously they are gonna make 10x their investment.

Yeah, but if that develper can’t get the 16 GB, then Apple makes the base 16 GB and raises the price, then that developer can’t get any model. He’s just assed out.

This is why I don’t understand the complaints about the existence of 8 GB Macs. Apple sets the price for 16 GB at a certain amount regardless of whether there is an 8 GB model. Getting rid of the 8 GB model only hurts those who don’t need 16 GB. It’s does nothing to help those who do.

The $1499 for 16 GB/512 GB has nothing to do with the existence of lower specs. It is because that the price they chose for those specs. To me the logical push is to get Apple to cut the price of 16 GB/512 GB. Hopefully, any lower specs can drop accordingly.

It would be good if the AI push causes Apple to at least shift their RAM/storage strategy. Powerful models need more RAM and storing multiple powerful model needs more storage. The price of buying RAM and storage should be adjusted.

But, if they choose to keep selling a limited spec that can’t fully use the new models for the people who don’t want to use AI now, that doesn’t hurt the price pain for getting one with enough RAM.

 

[name99]

What's with the die size?

Not sure what the block at the bottom of the NPU is supposed to be / why is it marked as part of the NPU?

Consists of the ANE's Planar Engine, and "L2 Cache" (actually manually managed buffer with lots of fancy features including DMA, hardware loops, and hardware synchronization/flow control between producers and consumers)

 

[Eug]

I know this is late to the game, but just for fun I ran Geekbench 6.3 (CPU) exactly once on my new 11" iPad Pro (M4), and got 3726 / 13289.

This is fairly representative of the 11" iPad Pro in real world submitted benchmarks (for machines that aren't specifically configured for highest scores). Note that my machine hasn't even completed all its setup (like Photos syncing) after activation, but it was unplugged so I think some of these background processes were paused automatically.

For some reason the 13" model hasn't shown up yet on the Geekbench iOS leaderboard, so the 11" currently holds the iOS single-core and multi-core lead, by a very large margin.

At the bottom of the chart is my iPad Air 2, which is what I was temporarily using until today.

 

[Glo.]

M1 Max Mac Studio with 24 GPU cores, 32 GB of RAM, 512 GB SSD for equivalent to 1199 USD.

Opinion. Good deal, or bad deal?

 

[igor_kavinski]

M1 Max Mac Studio with 24 GPU cores, 32 GB of RAM, 512 GB SSD for equivalent to 1199 USD.

Opinion. Good deal, or bad deal?

Not a bad deal. If you have the option and want to use it for development, you could do worse. Just don't do heavy writes to the SSD and try to do such workloads on external SSD or create a RAMdrive if that is a possibility.

 

[Eug]

M1 Max Mac Studio with 24 GPU cores, 32 GB of RAM, 512 GB SSD for equivalent to 1199 USD.

Opinion. Good deal, or bad deal?

Decent price, esp. considering it's only 2 years old. That means it still has a decent length of time of OS support left.

However, for CPU it's actually slower than my iPad. I don't know if that matters to you or not.

https://browser.geekbench.com/v6/cpu/compare/6615413?baseline=6618549

Is that for a used or third party refurbished unit or something?

 

[FlameTail]

However, for CPU it's actually slower than my iPad.

Nice flexing of the world's fastest ST performance.

 

[Nothingness]

However, for CPU it's actually slower than my iPad. I don't know if that matters to you or not.

What is frightening, is that only looking at the Clang result (that one is interesting for developers), the M4 is 40% faster than the M1.

 

[igor_kavinski]

What is frightening

Aren't compilation workloads multithreaded? The Studio looks like a better deal if the alternative is an 8GB M4 laptop that he has to wait for to be released and then is generally unhappy with due to running into RAM capacity related limitations. But it's his preferences that will dictate his purchasing decisions.

 

[Nothingness]

Aren't compilation workloads multithreaded? The Studio looks like a better deal if the alternative is an 8GB M4 laptop that he has to wait for to be released and then is generally unhappy with due to running into RAM capacity related limitations. But it's his preferences that will dictate his purchasing decisions.

Certainly! I was only pointing out M4 screams

EDIT: Compilation is not multithreaded in general (except perhaps the linker); but each file can be compiled in a different process. This makes me wonder if any work has been done to multithread compilation itself, too lazy to search at the moment.

 

[igor_kavinski]

EDIT: Compilation is not multithreaded in general (except perhaps the linker); but each file can be compiled in a different process. This makes me wonder if any work has been done to multithread compilation itself, too lazy to search at the moment.

I built Libreoffice out of curiosity sometime back. Not sure about what it was doing exactly during the whole process but it used all eight threads and took 20 minutes to finish on i7-4770 on 500GB Samsung 860 EVO SSD.

Future curiosity adventure is to do the same on my 128 thread AMD Zen 2 server on a RAM drive and see how fast it goes

By the way, may also do that on M1 Air if I can find its damn charger. Don't even remember what the damn thing looks like.

 

[igor_kavinski]

Certainly! I was only pointing out M4 screams

I bet it's the pain of being clocked too high and near the limit

 

[The Hardcard]

Aren't compilation workloads multithreaded? The Studio looks like a better deal if the alternative is an 8GB M4 laptop that he has to wait for to be released and then is generally unhappy with due to running into RAM capacity related limitations. But it's his preferences that will dictate his purchasing decisions.

M4 laptops come in 16 GB and 24 GB alternatives as well with no indication they would be any cheaper if the 8 GB model was ended.

 

[igor_kavinski]

M4 laptops come in 16 GB and 24 GB alternatives

I doubt they would be priced anywhere close to 1199 USD which makes the Studio with 32GB RAM sound like a better buy.

 

[amosliu137]

this is the screenshot for JamesAslan review