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

 

[Mopetar]

I don't understand why you believe this. Just see some of his impolite replies to me in the past. Do these sound like "polite disagreement"?

If you think that, then I definitely want "whatever you're smokin" as @Eug said.

The phrase "whatever you're smoking" is a colloquialism to indicate that a person finds a claim so wild or absurd as to believe you must be under the influence of drugs to make it. It's hardly unfriendly in and of itself and they're are considerably more rude ways of conveying that sentiment.

But the post you linked is over three and a half years old. What kind of grudge are you carrying that you would even dredge something like that up to try to prove a point in the first place?

Also, in reading that post and what was being replied to, it doesn't really make you look any better. We're about 18 months away from being able to actually evaluate your hypothetical (or whatever you would care to call it) of Apple hitting 50% PC market share. I don't think that's going to happen. They've only been above 20% in the US in a single quarter based on data from a quick Google search.

So I think it's quite fair for the other poster to have used that particular phrasing. Your prediction was wildly speculative and if you had bet money on it you would have lost. Even if that weren't the case, your own post was no less inflammatory in terms of language used, so I would not consider you to be some innocent victim receiving misdirected derision. I don't care to read back through three year old posts to see who started it, but it's my opinion that any curt replies you received were deserved.

 

[okoroezenwa]

9to5’s compilation of evidence for M4 on the iPad Pro makes sense and kind of has me convinced.

M4 iPad Pro: Will Apple put a brand new chip in its next iPad?

Bloomberg’s Mark Gurman broke the internet on Sunday by claiming that the next generation iPad Pro, expected to be announced...

9to5mac.com

If this does happen, maybe the speculation of [mention]Doug S [/mention] that the M4 Ultra shows up end of this year happens too!

 

[Eug]

9to5’s compilation of evidence for M4 on the iPad Pro makes sense and kind of has me convinced.

M4 iPad Pro: Will Apple put a brand new chip in its next iPad?

Bloomberg’s Mark Gurman broke the internet on Sunday by claiming that the next generation iPad Pro, expected to be announced...

9to5mac.com

If this does happen, maybe the speculation of [mention]Doug S [/mention] that the M4 Ultra shows up end of this year happens too!

None of that is evidence. As much as I’d like to have an M4 in my potential 2024 iPad Pro (basically just for bragging rights since I don’t need the performance), everything in that report (besides the statement that new iPad Pro models are coming) is just speculation, on top of Mark Gurman’s speculation. We shall see soon, but AFAIK there have been no actual M4 leaks. The main leak that seems to have multiple sources is that the next iPad Pro will have dual-stack OLED.

 

[Doug S]

None of that is evidence. As much as I’d like to have an M4 in my potential 2024 iPad Pro (basically just for bragging rights since I don’t need the performance), everything in that report (besides the statement that new iPad Pro models are coming) is just speculation, on top of Mark Gurman’s speculation. We shall see soon, but AFAIK there have been no actual M4 leaks. The main leak that seems to have multiple sources is that the next iPad Pro will have dual-stack OLED.

Nah I was skeptical but now I'm not, that is EXACTLY the kind of evidence I was looking for to believe that the new iPads will have M4. If they have M4 in iPad Pro now I don't think we will be waiting until the next iPhone launch to see M4 Macs either.

 

[okoroezenwa]

Nah I was skeptical but now I'm not, that is EXACTLY the kind of evidence I was looking for to believe that the new iPads will have M4. If they have M4 in iPad Pro now I don't think we will be waiting until the next iPhone launch to see M4 Macs either.

Exactly. Identifiers have long been used to determine what Apple will be releasing (to different degrees of course) and the logic in the 9to5 post makes sense so I don't see why it wouldn't count as evidence. I could see something like:

M4 - May 2024
M4 Pro/Max - October 2024
M4 Ultra - announced 2024?

Would certainly be nice to get a Mac Studio upgrade for the Max version at least. With only 1 data point it seems like they'd rather wait for the Max and Ultra to be ready before updating it so maybe announcing would count with the Max variant available immediately and the Ultra coming later. Or they could just leave it till next year anyway.

 

[Glo.]

Nah I was skeptical but now I'm not, that is EXACTLY the kind of evidence I was looking for to believe that the new iPads will have M4. If they have M4 in iPad Pro now I don't think we will be waiting until the next iPhone launch to see M4 Macs either.

Yep. Identifiers are very good source of information about what chip will be inside.

Also - the rumors of M4 being released much earlier explains why there is no UltraFusion in M3 Max. M3 series were effectively stop-gap solution before M4 series.

 

[smalM]

M3 series were effectively stop-gap solution before M4 series.

I only agree on the M3 Max as base for a M3 Ultra.
M3 maybe a short-lived generation but that could be based solely on the short-lived N3 node.
From the standpoint of functional units that form the SoC the M3 generation is not a stop-gap at all.
Also we saw the strategic shift of the Pro to become a separate solution instead of a choped Max.

 

[repoman27]

I highly doubt Apple will call whatever is in the new iPad Pros "M4", but if they do, it would be kinda stupid because it is just another Everest / Sawtooth SoC on TSMC N3. N3E didn't enter volume production until Q4'23, so products shipping before June is highly unlikely, and I doubt Apple is planning a paper launch here. Furthermore, the supposed SoC ID for the new iPad Pro chips is T8132, while the identifier for the A17 Pro is T8130, indicating that these chips are closely related. There is a small possibility that Apple has moved mountains and that the T8132 is the M3 ported to TSMC N3E and will be able to launch in May.

Apple has never introduced a new µarch in anything other than an A-series chip, and it would be odd for them to do so now on N3. The only thing that lends credence to this is that Apple hasn't introduced a new µarch since September 2022, so they are overdue, and maybe they had one more cued up for OG N3? It's a little odd that the A16 was T8120 and the M3 was T8122, despite the fact that the M3 pulled in the updated GPU cores from the A17 Pro. Then again, the M3 Max goes by both T6031 and T6034 depending on the number of memory channels enabled, despite being the same die in different packages, so this may also represent nothing more than a packaging change.

 

[Nothingness]

Or perhaps internal IDs don't mean much.

Public product IDs don't mean much anyway

 

[Eug]

Or perhaps internal IDs don't mean much.

They must have some meaning to Apple of course, but they have not always meant what we thought they meant, because we're just guessing. Educated guessing, but guessing nonetheless.

Anyhow, this ID thing has been discussed literally for months now on other forums and several ideas have been bandied about in the last few weeks:

1. T8132 is M4
2. T8132 is M3 variant
3. T8132 is M3 on N3E
4. T8132 isn't an accurate ID

However, it's all speculation.

In fact, that likely why Mark Gurman (and also 9to5mac) is now talking about it, because he's been reading those same forum posts. IOW, he has no new inside info.

 

[smalM]

because it is just another Everest / Sawtooth SoC on TSMC N3

nonsense

 

[Mopetar]

I'm looking to upgrade my iPad Pro and I'd love it if I could get an M4 even though it really wouldn't add anything for me and what I use it for.

However, unless Apple is just announcing a product that you can't yet buy, it would mean that the M4 would have come out a bit more than 7 months after the M3, which puts the release cadence all over the place.

If it weren't Gurman reporting this, I don't think I'd give it any stock at all given everything else. The more likely conclusion from this would be that what Apple would be calling the M4 would probably be something more akin to an M3.5 as opposed to a whole new M-series chip.

An M4 this soon raises a lot of interesting questions.

 

[Doug S]

Also - the rumors of M4 being released much earlier explains why there is no UltraFusion in M3 Max. M3 series were effectively stop-gap solution before M4 series.

Which is what I've been saying for months, ever since the die photos showed no I/O pads. There were still a lot of people saying there's no way the M4 would come out in October only a year after the M3. Well I guess I got that part wrong too, but less wrong than the people pushing '18 months' based on two data points influenced by covid and TSMC process issues

The question I have now is when do we see the M4's big brothers. I see no reason for Apple to wait until after the iPhone launch for new Macs, they may not be in any hurry to update the ones that got M3 but everything still on M2 like Mini, Studio and Pro are definite candidates for summer launch. Even the stuff like Macbook Pro might get in before iPhone, if their marketers still believe "back to school" launches will help their sales. Maybe in the future Mac launches are mostly in the summer, happening before iPhone. Something more Christmas oriented like Macbook Air might be late fall (having it updated later helps avoid cannibalizing MBP)

 

[poke01]

It's a little odd that the A16 was T8120 and the M3 was T8122, despite the fact that the M3 pulled in the updated GPU cores from the A17 Pro.

The M3 is very similar to the A17 Pro than A16. The CPU, GPU and media engines being the same back this up.

But why not give the public identity the same as the A17 Pro which is xxx16,x. Instead, Apple gave xxx15,x for all M3 Macs.

There was a rumour that A16 GPU had been delayed. Could it be that M3 is actually a mix of A16 and A17.

So this upcoming chip may be a mix of whatever was missed on A17 and some new IP from A18.

Here’s the kicker, A18s ID is T8140 and its Mx ID is T8142. So who so knows?

Again, it’s a mystery till next week

 

[Doug S]

M3 had the same CPU cores as A17P, not A16.

 

[repoman27]

M3 had the same CPU cores as A17P, not A16.

M3 had the same CPU cores as A17 Pro *and* A16—they all use Everest and Sawtooth. That's sort of the problem with inferring too much from the SoC identifier here. The A-series repeated CPU microarchitectures while the M-series appeared to skip over the A16 generation, so the SoC IDs aren't aligned anymore.

 

[poke01]

M3 had the same CPU cores as A17 Pro *and* A16—they all use Everest and Sawtooth. That's sort of the problem with inferring too much from the SoC identifier here. The A-series repeated CPU microarchitectures while the M-series appeared to skip over the A16 generation, so the SoC IDs aren't aligned anymore.

yep. One thing for sure. Whatever chip we get next week is not the same M3 we have now.

 

[repoman27]

If you go even further down the codename rabbit hole, the A17 Pro was "Coll" and the M4 is supposedly "Donan"—both Scottish islands. So A17 Pro = Coll = T8130 and M4 = Donan = T8132. Now the codenames don't necessarily indicate that both SoCs share all of their IP blocks, but it does suggest that development started around the same time.

This makes me think that Apple will call this the M4 and that it will be Everest and Sawtooth yet again. It's most likely the M3 ported to N3E with a significantly larger / upgraded NPU and maybe an LPDDR5X-8533 memory interface.

N3 is obviously going to be a short node, so Apple does need to transition to the main N3E-based process family as soon as possible. I figured they would skip N3E entirely and go right to N3S, but perhaps not. I still think the A18 will be on N3S rather than N3E. We all knew that Apple was the lead / only customer for N3, but I don't recall anyone ever saying that Apple would also be the lead customer on N3E. I don't see how this didn't leak, as it would have directly impacted other TSMC customers. All of which makes me wonder if the M4 will still be on N3.

Gurman originally suggested that the new iPad Pros would launch in March, then revised that to April and finally to May, of course claiming delays rather than admitting that he might have gotten it wrong (textbook leaker 101). But Apple didn't wait until Q1'24 to decide whether they were shipping the new iPad Pros with an M3 or an M4; that design would have been locked nearly 12 months ago. Apple also knew what process they were targeting for the M4 several years ago, and it would have been undergoing qualification 12 months ago. So given that N3E didn't enter volume manufacturing until Q4 2024, and it takes at least 6 months to go from volume manufacturing to shipping a product, the idea of Apple planning to ship an N3E based product in March would have required TSMC to ramp N3E earlier than they did. TSMC did say that they thought they could pull volume production for N3E in a few months, but that didn't materialize. Then again, maybe TSMC was ready with the process but Apple had been guaranteed the lead spot and had to do a respin, thus delaying volume production by several months. ¯\_(ツ)_/¯

 

[poke01]

If you go even further down the codename rabbit hole, the A17 Pro was "Coll" and the M4 is supposedly "Donan"—both Scottish islands. So A17 Pro = Coll = T8130 and M4 = Donan = T8132. Now the codenames don't necessarily indicate that both SoCs share all of their IP blocks, but it does suggest that development started around the same time.

The codenames names for M3 family are even more closely related. They are all Spanish islands.
While Coll is Scottish, Donan is Irish.
Both are Scottish, my bad.

Gurman originally suggested that the new iPad Pros would launch in March, then revised that to April and finally to May, of course claiming delays rather than admitting that he might have gotten it wrong (textbook leaker 101)

Gurman isn’t great at predicting events or releases months or even weeks from happening. The last event he didn’t even predict till Apple released an event invite.

My guess is Apple always intended to launch in May.

 

[Doug S]

Gurman originally suggested that the new iPad Pros would launch in March, then revised that to April and finally to May, of course claiming delays rather than admitting that he might have gotten it wrong (textbook leaker 101). But Apple didn't wait until Q1'24 to decide whether they were shipping the new iPad Pros with an M3 or an M4; that design would have been locked nearly 12 months ago. Apple also knew what process they were targeting for the M4 several years ago, and it would have been undergoing qualification 12 months ago. So given that N3E didn't enter volume manufacturing until Q4 2024, and it takes at least 6 months to go from volume manufacturing to shipping a product, the idea of Apple planning to ship an N3E based product in March would have required TSMC to ramp N3E earlier than they did. TSMC did say that they thought they could pull volume production for N3E in a few months, but that didn't materialize. Then again, maybe TSMC was ready with the process but Apple had been guaranteed the lead spot and had to do a respin, thus delaying volume production by several months. ¯\_(ツ)_/¯

Six months from start of mass production to first shipments isn't set in stone. It is probably three months at the outside from start of mass production for first wafers exiting the line. You need to do testing, packaging, etc. and then of course assemble the products they go into but if the first wafers entered Dec. 15 it wouldn't be unreasonable for the first shipments of iPad Pro to happen in mid/late May. Until N3 TSMC used to enter mass production in Q2 every year, and Apple shipped 10-20 million iPhones containing those chips before the end of Q3. Apple/Foxconn has pretty good manufacturing scale so they can take Apple Silicon packages from TSMC and ship them out in the door in products in no time.

TSMC is also doing a year's worth of risk production, and it isn't like they have a big switch labeled "mass production" that they pull and everything changes in a single second. The wafers coming out risk production shortly before mass production officially starts are basically identical, so there is some leeway for using risk production wafers run before the official start of mass production. Heck Intel used to ship chips from what TSMC would term "risk production" out of their development fab all the time when launching chips off a new process. That's why they were so scarce at first. Doing M4 for a lower volume product like iPad Pro would allow Apple to use some risk production wafers whereas for a high volume product like iPhone that would be impossible, though I don't think it would be necessary even if mass production didn't start until Dec. 31st.

 

[Eug]

Six months from start of mass production to first shipments isn't set in stone. It is probably three months at the outside from start of mass production for first wafers exiting the line. You need to do testing, packaging, etc. and then of course assemble the products they go into but if the first wafers entered Dec. 15 it wouldn't be unreasonable for the first shipments of iPad Pro to happen in mid/late May. Until N3 TSMC used to enter mass production in Q2 every year, and Apple shipped 10-20 million iPhones containing those chips before the end of Q3. Apple/Foxconn has pretty good manufacturing scale so they can take Apple Silicon packages from TSMC and ship them out in the door in products in no time.

TSMC is also doing a year's worth of risk production, and it isn't like they have a big switch labeled "mass production" that they pull and everything changes in a single second. The wafers coming out risk production shortly before mass production officially starts are basically identical, so there is some leeway for using risk production wafers run before the official start of mass production. Heck Intel used to ship chips from what TSMC would term "risk production" out of their development fab all the time when launching chips off a new process. That's why they were so scarce at first. Doing M4 for a lower volume product like iPad Pro would allow Apple to use some risk production wafers whereas for a high volume product like iPhone that would be impossible, though I don't think it would be necessary even if mass production didn't start until Dec. 31st.

I read an article on manufacturing somewhere saying a realistic time frame from receipt of chips to laptop units on store shelves in volume is roughly 3 months with large Chinese manufacturers. (This was for laptops not iPad Pros, but I assume it would be similar.)

What volume of chips are you talking about during risk production? Apple may ship about 20 million iPad units in a busy quarter. I'm not sure how many of those are iPad Pros, 5 million?

 

[repoman27]

Six months from start of mass production to first shipments isn't set in stone. It is probably three months at the outside from start of mass production for first wafers exiting the line. You need to do testing, packaging, etc. and then of course assemble the products they go into but if the first wafers entered Dec. 15 it wouldn't be unreasonable for the first shipments of iPad Pro to happen in mid/late May. Until N3 TSMC used to enter mass production in Q2 every year, and Apple shipped 10-20 million iPhones containing those chips before the end of Q3. Apple/Foxconn has pretty good manufacturing scale so they can take Apple Silicon packages from TSMC and ship them out in the door in products in no time.

TSMC is also doing a year's worth of risk production, and it isn't like they have a big switch labeled "mass production" that they pull and everything changes in a single second. The wafers coming out risk production shortly before mass production officially starts are basically identical, so there is some leeway for using risk production wafers run before the official start of mass production. Heck Intel used to ship chips from what TSMC would term "risk production" out of their development fab all the time when launching chips off a new process. That's why they were so scarce at first. Doing M4 for a lower volume product like iPad Pro would allow Apple to use some risk production wafers whereas for a high volume product like iPhone that would be impossible, though I don't think it would be necessary even if mass production didn't start until Dec. 31st.

You are correct, six months isn't set in stone. However, because EUV litho steps take significantly longer than DUV, the more EUV layers the process incorporates, the longer the cycle time. Cycle times for N3 are brutal. Some analysts were claiming 4 months. N3E is a bit better, but still probably at least 100 days and definitely in excess of 3 months. Apple probably needs around 2.5 million units of inventory for the launch quarter, which given the estimated die size, yield rates, and available capacity is probably at least two weeks worth of wafer starts.

After test, sort, and packaging, the SoCs need to be shipped to a contract manufacturer where they are integrated into the logic board assembly followed by more testing. Those assemblies are sent to other contract manufacturers for final device assembly. In an effort to reduce their reliance on any single contract manufacturer, and Chinese supply chains in general, device assembly is now taking place not just at Foxconn, but also Compal, Quanta, Pegatron, and Wistron facilities located in Vietnam, Thailand, India, and Brazil. After device assembly is completed, there is software loading and more testing, followed by packaging and then shipping to the various destination markets to fill the channel ahead of launch. And don't forget that Apple recently committed to shipping via ocean freight rather than air, which means that final leg of the journey can take 3 to 5 weeks all on its own.

So I think six months is a very reasonable estimate.

And TSMC does in fact have a big switch labeled "volume production" which exists due to their legal obligation to their shareholders. As soon as they start processing wafers for a customer order that will result in inventory that they can convert into revenue, they have an obligation to notify shareholders during the next quarterly earnings call. Apple's initial order for M4 chips for the iPad Pros is going to trigger nearly half a billion dollars worth of revenue. Making 2.5 million chips for your largest customer is the literal definition of volume production. Intel has PRQ which is similar and is even more clearly spelled out for shareholders. That is the point where risk production ends and qualified chips are converted on the balance sheet into inventory. Realize that Apple will ship more M4s in the launch quarter than Intel has shipped Meteor Lake SoCs to date. We are not talking about a low volume product. The iPhone is just in a category unto itself.

 

[SpudLobby]

Nah I was skeptical but now I'm not, that is EXACTLY the kind of evidence I was looking for to believe that the new iPads will have M4. If they have M4 in iPad Pro now I don't think we will be waiting until the next iPhone launch to see M4 Macs either.

Eh I dunno I think we could see a fall October or winter event. But it is interesting how fast they’re moving. As you know, I pointed out N3B was inferior and Apple would look to move to N3E ASAP (or certainly there wouldn’t be any delay in doing so), but man an M4 iPad in May after M3 MacBook Airs earlier this year is interesting.

 

[SpudLobby]

I read an article on manufacturing somewhere saying a realistic time frame from receipt of chips to laptop units on store shelves in volume is roughly 3 months with large Chinese manufacturers. (This was for laptops not iPad Pros, but I assume it would be similar.)

What volume of chips are you talking about during risk production? Apple may ship about 20 million iPad units in a busy quarter. I'm not sure how many of those are iPad Pros, 5 million?

I mean, TSMC’s risk production is very different from Intel and particularly the closer we get to mass production I could see this making sense.

But really look at it this way: 3 months for laptops with merchant chips, but this is iPads — and just one model — and Apple is very vertically integrated. If mass production started in Q4 and they’ve had 3-4 months of it for M4 and time to package and slot them in, it makes sense.

 

[SpudLobby]

Also this is an announcement event, so release could be late May early June. Wouldn’t be like July because it’s Apple but it’s not unheard of for them to have some couple weeks to a month of delays.