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

 

[poke01]

I wonder if the R1 was bumped (InFO/CoWoS) or bumpless (SoIC)? Regardless, it was Apple's first crack at a truly disaggregated design with 11 tiles, although I'm not sure all of them are active.

https://twitter.com/x/status/1784016546500940244

here is a better shot

https://twitter.com/x/status/1790178060668641594

R1 is Apple's first tiled silicon

 

[repoman27]

Nah, people obsess too much about heat dissipation, especially in the context of Apple.
Highest power you can get out of an M3 Max is something like 80W, so two of them is 160W. That's way below what Intel or AMD generate in a smaller area.

I'd say a more important issues are:
- is the tech ready?
- is Apple ready?
- is it cheaper (or otherwise more appropriate) than the alternative?

All three seem false right now.
Most importantly, Apple doesn't jump on new tech for the sake of "FIRTS!!!". They didn't feel a need either to provide V-Cache first, or even second.
Right now vertical stacking of logic doesn't (as far as I can see) solve an important Apple problem.
Vertical stacking of memory? That's not quite so obvious, but I still don't see anything requiring a solution more drastic (this generation) than something like two ranks, one on each side of a board.

The entry level M2 Max Studio dissipated up to 145 W, while the fully equipped M2 Ultra version could pull 295 W. The highest power consumption for an M3 Max in a laptop isn't necessarily the highest power draw for an M3 if you were to let it off the leash. But you can't just stack two SoC dies vertically. If that were possible, everybody would already be doing it. You need to disaggregate, put the less power hungry stuff on the bottom, and figure out how to route everything. Case in point, Intel couldn't get Foveros to work in consumer products with the PCH as the base tile, and had to revert to using passive interposers (while still calling it Foveros).

Clamshelling won't work unless Apple moves the SDRAM off package, which they would be loath to do. They could go dual-rank by PoP stacking the memory packages, but that would likely impose a Z height penalty. However, if Apple did double the maximum RAM capacity, nobody would be able to afford it given the current pricing, so why bother?

 

[repoman27]

According to Kurnal's composite image, it looks like the R1 is 3 active tiles, 6 embedded silicon passives, and 8 dummy tiles. The smaller tiles flanking the main tile look like memory dies, and the d2d interface looks an awful lot like UltraFusion. So probably bumped InFO-LSI again, or possibly just InFO-RDL, since Kurnal didn't include any interposer chips in that image. Some fairly liberal cutting and pasting took place to create that image though, so who knows.

 

[Doug S]

M4 has new P-cores, but the E-cores, GPU and NPU are the same as M3/A17. They are just clocked higher.

So will M4 Pro/M4 Max/M4 Ultra/A18 also be that way?

We do know the A18 is getting a significantly beefed up NPU.

M4 feels like an M3.5

There's nothing that says every unit has to updated on the same schedule. I'm sure they have different teams working on the CPU versus GPU/NPU. I wouldn't be surprised if maybe P, E, GPU, and NPU all have their own teams. Back when A series was all there was it made sense to align everything to the iPhone's schedule, but with three different Apple Silicon dies added to the mix that's more of a challenge. Now they did manage to do that with A17P/M3/M3P/M3M where everything was ready (or at least released in products) within a couple months time, but maybe after they did it they decided "wow that was a lot of deadlines all hitting at once, let's never do that again!"

So perhaps with M4 we're seeing a new way of working where they spread those deadlines (getting ready for tapeout, post tapeout verification and potential steppings, etc.) around? If you do there's no longer any reason to have all the units adhering to the same rigid "gotta be ready to ship in products in September" schedule. So there's a tapeout deadline for chip 'X', and they take whatever the latest unit of each type there is. If the new NPU had been ready in time for M4 they would have included it, but it wasn't so they used the previous one.

So maybe A18 gets something that M4 didn't (i.e. that beefed up NPU) and if there ends up being enough of a gap between M4 and M4P/M4M maybe it gets something else like a new GPU.

 

[igor_kavinski]

There's nothing that says every unit has to updated on the same schedule. I'm sure they have different teams working on the CPU versus GPU/NPU. I wouldn't be surprised if maybe P, E, GPU, and NPU all have their own teams.

That's probable, though I would add complacency to that. They don't feel the need to push themselves hard. Just win some benchmark and hey, it's still gonna sell well. Not like their users can turn to any other competitor without turning their brains upside down. They may also be gearing up to launch something special by the end of the year, in time to face off against Lunar Lake.

 

[name99]

Where’s any indication that the M5 is coming in 2025? I would agree that it is unlikely that higher-end M4 and the base M5 would show up in the same timeframe. Gurman, though has been reasonably reliable about product releases. I think his claims about the Mac Studio releases in H2 2025 are more likely than any contradicting sources. and I think the reports of Apple’s testing of SoIC starting last year with production targeted for between 2025 and 2026 give Gurman claims even more weight.

To me that means either the M5 is not coming next year, or the next Ultra will be the M5 Ultra.



to the contrary, all of your issues seem true right now. SoIC is shipping now in AMD products. What is needed on TSMC’s part is boosting production capacity which they are doing.

Reports say Apple began testing SoIC at some point in 2023 with the target of going into production in 2025. It appears they will be ready. A key benefit to the technology Is bringing down the cost and complexity of the PCB, making it the cheaper way to go (after the higher upfront costs.)

2025...
If you agree with me that M4 high end's will appear in 2024, then we are saying the same thing!
If you think Apple will cruise till 2025H2 without an upgrade at the high end, no M4 Ultra, well, we have a fundamental disagreement, which time will resolve!

M4 Ultra SUCCESSOR, sure. (Except that, as I've argued, I suspect M5 will be a low-end only release, so it will be M6 Ultra that gets SoIC etc)

there are numerous other benefits as well. The space needed for the interconnect on the Max will be significantly smaller. And the capacity to present to the outside world as a single chip is enhanced by lower latency, and a significant increase in bandwidth capacity. The icing on the cake is the power savings from reducing the distance, the signals need to travel.

The more I read about it the more it seems it is absolutely worth it for Apple to wait on this despite it forcing Apple to try to sell Mac Studios and Mac Pros for yet another year with old cores. this could be mitigated by the new need for large amounts of M2 Ultras Internally.

And again, I’m hoping that timeframe also gives them space to tweak some of the weakest links in an otherwise compelling product for higher end AI workflows.

 

[name99]

The entry level M2 Max Studio dissipated up to 145 W, while the fully equipped M2 Ultra version could pull 295 W. The highest power consumption for an M3 Max in a laptop isn't necessarily the highest power draw for an M3 if you were to let it off the leash. But you can't just stack two SoC dies vertically. If that were possible, everybody would already be doing it. You need to disaggregate, put the less power hungry stuff on the bottom, and figure out how to route everything. Case in point, Intel couldn't get Foveros to work in consumer products with the PCH as the base tile, and had to revert to using passive interposers (while still calling it Foveros).

Clamshelling won't work unless Apple moves the SDRAM off package, which they would be loath to do. They could go dual-rank by PoP stacking the memory packages, but that would likely impose a Z height penalty. However, if Apple did double the maximum RAM capacity, nobody would be able to afford it given the current pricing, so why bother?

The factor that matters for this discussion is the SoC power - not IO power, not DRAM power.
The numbers I have seen for M3 Max SoC are as I gave.
But it's not important: Even if we accept those Studio Ultra numbers, that collaborates my point. Those match high end intel power – which clearly is feasible; that doesn't make the chip melt or misfunction or whatever.

Stacking two i9-1400KS may be a problem. But what we want to do is stack two Mn Max, NOT two i9-1400KS.

 

[Glo.]

Again, why HBM? What does it offer APPLE (not nVidia, not AMD, APPLE) that they can't get, and better, from LP-DDR5 or LP-DDR5X?


This is plausible, in different versions.
Many of the packaging patents suggest a basic manufacturing unit of an "Ultra" (ie two "Max's" manufactured side by side, with an RDL [think fancy version of Cerebras wiring between cheaps] between the two. This basic "two chip" unit is then diced as a single unit, and assembled into larger packages.

Such a scheme is certainly compatible with other ideas we have thrown out, like moving IO and displays to a separate chip.

HBM allows Apple to increase the bandwidth available to their SOCs.

What it means is that M4 Ultra can have BOTH, HBM and LPDDR memory controller, and use HBM as a "cache" RAM, while also shrinking the required bus width for Ultra chip.

Think about it this way. Scaling LPDDR5 will be more difficult, so M4 Ultra could come with HBM 2048 bit bus, and 512 bit bus on LPDDR5. It will still give 256 GB of RAM in the most extremere configuration, on 512 bit bus.

If Apple found a way to share the HBM bandwidth between two M4 Ultra chips through UltraFusion connection - Apple pretty much solved all of their problems with scaling those chips.

 

[The Hardcard]

2025...
If you agree with me that M4 high end's will appear in 2024, then we are saying the same thing!
If you think Apple will cruise till 2025H2 without an upgrade at the high end, no M4 Ultra, well, we have a fundamental disagreement, which time will resolve!

M4 Ultra SUCCESSOR, sure. (Except that, as I've argued, I suspect M5 will be a low-end only release, so it will be M6 Ultra that gets SoIC etc)

I would be happy to see new high end Macs this year, but I’m afraid we disagree. the buzz concerning 2025 high-end Macs continues to grow with no countervailing information that I have seen.

Macrumors just today reported on a Digitimes article corroborating the 2025 Apple Silicon plans. That would be to my knowledge the third source (obviously all unofficial hearsay information.) So multiple pieces of information about 2025, with as far as I know, no information about 2024. Not impossible, but if you’re gonna go on probability…

Another factor in the probability bank would be that given the indication that there will be a high-end Apple Silicon release in 2025, it would make a 2024 Mac Studio an abnormally short lived product for an upper level Mac.

Maybe MacBook Pros later this year. neither the Pro nor Max variants depend upon SoIC.

The most exciting thing for me about today’s report is that it looks like my hope is coming true. The report claims the new Ultra will be a dual use chip powering Apple datacenter hardware as well as going into Macs. So, it will be intentionally tweaked to maximize responsiveness in cloud AI requests and those tweaks will be available to the public for personal local use.

 

[poke01]

I feel like once the Pro and Max come out this year we will have a better understanding of Apples plans.

 

[FlameTail]

I feel like once the Pro and Max come out this year we will have a better understanding of Apples plans.

No you won't.

Apple functions in mysterious ways...

 

[Doug S]

M4 Ultra SUCCESSOR, sure. (Except that, as I've argued, I suspect M5 will be a low-end only release, so it will be M6 Ultra that gets SoIC etc)

What do you mean by "low end only release"? You mean like M3, where there's a Max but no Ultra?

I think M3 getting that treatment could easily be explained by N3B being such a problematic process, and Apple perhaps having incentive to stop using it as soon as possible. But now that they seem to have got themselves on the yearly schedule for Apple Silicon I always thought they would, there could be merit to the idea that they only do an Ultra every other year since that's such a small segment of the overall Mac market - though if they start building themselves a lot of servers out of Ultras then suddenly that NRE is being amortized across a lot more chips and it makes sense to roll a new one every year if only because that's what their "internal customer" will want.

 

[poke01]

What do you mean by "low end only release"? You mean like M3, where there's a Max but no Ultra?

I think M3 getting that treatment could easily be explained by N3B being such a problematic process, and Apple perhaps having incentive to stop using it as soon as possible. But now that they seem to have got themselves on the yearly schedule for Apple Silicon I always thought they would, there could be merit to the idea that they only do an Ultra every other year since that's such a small segment of the overall Mac market - though if they start building themselves a lot of servers out of Ultras then suddenly that NRE is being amortized across a lot more chips and it makes sense to roll a new one every year if only because that's what their "internal customer" will want.

Apple also scrapped the M3 iPad Pros. They really wanted to limit N3B node to high end iPhones and to MacBooks only.

 

[FlameTail]

M4 - N3E
M5 - N3P
M6 - N2
M7 - A16

(Speculation, assuming yearly release cadence)

 

[The Hardcard]

What do you mean by "low end only release"? You mean like M3, where there's a Max but no Ultra?

I think M3 getting that treatment could easily be explained by N3B being such a problematic process, and Apple perhaps having incentive to stop using it as soon as possible. But now that they seem to have got themselves on the yearly schedule for Apple Silicon I always thought they would, there could be merit to the idea that they only do an Ultra every other year since that's such a small segment of the overall Mac market - though if they start building themselves a lot of servers out of Ultras then suddenly that NRE is being amortized across a lot more chips and it makes sense to roll a new one every year if only because that's what their "internal customer" will want.

Does it seem like they have got themselves on the yearly schedule when it hasn’t been done once yet? I’m not as sure as you seem to be. I think it’s a mistake trying to use the previous releases to have any understanding to Apple’s plans for releasing its silicon.

The M4 release could also be easily explained by N3E getting better yields sooner, plus taking up more 5 nm supply for datacenter M2 Ultras.

As well, assuming the microbenching is close to accurate on revealing the core design, M1 through M4 could easily be explained by the chip engineers still figuring out the specifics of their design balance There has been a lot of bouncing around with the reorder and scheduling buffers and queues.

Is the semiconductor supply chain stable enough yet so you can rely on it to do yearly releases? Has TSMC solved the issues as they go into the angstrom era so that Apple can target a designs for N2 and A16 and not face delays? Are you sure there’s not going to be an N2B? No A16B?

Are they now satisfied with the balance of the core resources in the M4 design? Are they now ready for a straightforward, yearly progression? And not just the CPU. They struggled with the GPU design as well. Are those struggles now in the past?

Also, I think AI has shifted the plans for what every company plans about what they want to bring to market and when. I think the vision for the M5 and M6 are different than when they were having discussions about them 3 years ago, in ways that could easily affect when they are ready to release them.

So how does looking at post M1 history help anyone predict the M5 release better than just picking a random month out of a hat between now and March 2026?

 

[The Hardcard]

That's probable, though I would add complacency to that. They don't feel the need to push themselves hard. Just win some benchmark and hey, it's still gonna sell well. Not like their users can turn to any other competitor without turning their brains upside down. They may also be gearing up to launch something special by the end of the year, in time to face off against Lunar Lake.

There is no complacency at play with Apple Silicon. Every generation has had significantly modified performance cores. They have changed the cluster size and structure. The M3 Pro is a whole new SOC design.

The efficiency cores are a story unto themselves. With 1/3 of the performance for about 1/10 of the power they are far and away the most compute per watt in the industry, and have been reworked also throughout the generations. Each generation has significantly redesigned and new layouts for the GPU. The new display engine in the M4, The list goes on and on.

Apple is not thinking about benchmarks they are thinking about and delivering on performance. Performance is the target, benchmarks are just a reflection of that drive.

 

[name99]

HBM allows Apple to increase the bandwidth available to their SOCs.

What it means is that M4 Ultra can have BOTH, HBM and LPDDR memory controller, and use HBM as a "cache" RAM, while also shrinking the required bus width for Ultra chip.

Think about it this way. Scaling LPDDR5 will be more difficult, so M4 Ultra could come with HBM 2048 bit bus, and 512 bit bus on LPDDR5. It will still give 256 GB of RAM in the most extremere configuration, on 512 bit bus.

If Apple found a way to share the HBM bandwidth between two M4 Ultra chips through UltraFusion connection - Apple pretty much solved all of their problems with scaling those chips.

Look at the bandwidth available via DDR5/X right now.
It's enough for the amount of compute available, and scales upward in an obvious way as that compute scales.

Show me a realistic benchmark where an M3 Max is being throttled by not enough memory bandwidth.

 

[igor_kavinski]

With 1/3 of the performance for about 1/10 of the power they are far and away the most compute per watt industry

Apple being in possession of good stuff is just a waste for humanity. How much you wanna bet that there are engineers and other employees INSIDE Apple who want to use these efficient cores for something like 128-core, 256-core, 512-core and even 1024-core designs and completely blow away the cloud computing industry? But Apple executives won't have any of it. They would rather create their OWN data centers and provide this efficient computing power to paying customers with the magic of SaaS.

With Apple, it seems all new employees are taken through some top secret initiation ritual to wipe any semblance of humanity out of them so they work as humble and obedient servants for the purpose of achieving one and only one goal: Satisfying the unquenchable greed of Apple investors.

Generosity? It gets reported to HR and then you get a psych evaluation and if it becomes clear that you will never unlearn to be generous, they might decide to just let you go and be generous outside their campus.

 

[Doug S]

Is the semiconductor supply chain stable enough yet so you can rely on it to do yearly releases? Has TSMC solved the issues as they go into the angstrom era so that Apple can target a designs for N2 and A16 and not face delays?

They've been on a yearly cadence for iPhone since forever. Macs are a much lower volume product, and don't have a set release schedule in the same month every year. I don't know how you can argue that it is too hard to do a yearly release schedule for Apple Silicon when they've been proving they can do it year after year with the A series SoCs, despite covid and WFH, despite TSMC hiccups like N3B, despite key personnel leaving to form their own startups, despite simultaneously getting Apple Silicon off the ground.

In fact I would argue that it is BECAUSE of the requirement that they do yearly releases for iPhone at the same time every year that it becomes much easier to do the same for the Mac. They already have to tapeout an SoC that will ship in product by the tens of millions per month starting in September, so they've always aligned their core development roadmaps with that in mind. Now true Apple Silicon is three more tapeouts, but they are using the same cores so it isn't nearly as bad as that makes it sound.

Perhaps trying to do four tapeouts and verification cycles at the same time (i.e. for shipping Macs in October/November like they did with M3 last fall) stretches them a bit thin. If so that's easy to remedy, since Apple has never maintained a particular time of the year when a certain Mac model has to come out, but gives them perfect flexibility to choose the time of the year they do the other tapeout cycles to find what works best.

 

[Doug S]

Apple being in possession of good stuff is just a waste for humanity. How much you wanna bet that there are engineers and other employees INSIDE Apple who want to use these efficient cores for something like 128-core, 256-core, 512-core and even 1024-core designs and completely blow away the cloud computing industry? But Apple executives won't have any of it. They would rather create their OWN data centers and provide this efficient computing power to paying customers with the magic of SaaS.

With Apple, it seems all new employees are taken through some top secret initiation ritual to wipe any semblance of humanity out of them so they work as humble and obedient servants for the purpose of achieving one and only one goal: Satisfying the unquenchable greed of Apple investors.

Generosity? It gets reported to HR and then you get a psych evaluation and if it becomes clear that you will never unlearn to be generous, they might decide to just let you go and be generous outside their campus.

Ah yes, humanity would be in a much better place if they were like Nvidia and put that good stuff in a board they sold for $30,000 a pop to the cloud computing industry!

 

[igor_kavinski]

Ah yes, humanity would be in a much better place if they were like Nvidia and put that good stuff in a board they sold for $30,000 a pop to the cloud computing industry!

I know one thing about Nvidia and Apple: if I were transported into an alternate reality where they didn't exist, I wouldn't miss them one bit (despite whatever positive experiences I may have had using their products). Their arrogance seems to scale to new heights with every passing year.

 

[The Hardcard]

Apple being in possession of good stuff is just a waste for humanity. How much you wanna bet that there are engineers and other employees INSIDE Apple who want to use these efficient cores for something like 128-core, 256-core, 512-core and even 1024-core designs and completely blow away the cloud computing industry? But Apple executives won't have any of it. They would rather create their OWN data centers and provide this efficient computing power to paying customers with the magic of SaaS.

With Apple, it seems all new employees are taken through some top secret initiation ritual to wipe any semblance of humanity out of them so they work as humble and obedient servants for the purpose of achieving one and only one goal: Satisfying the unquenchable greed of Apple investors.

Generosity? It gets reported to HR and then you get a psych evaluation and if it becomes clear that you will never unlearn to be generous, they might decide to just let you go and be generous outside their campus.

Both of the following statements are equally accurate:

1 There is no major corporation more greedy than Apple.

2 There is no major corporation, less greedy than Apple.

Whatever chipmaker or other tech company that tricked you into believing that they are not scrambling for investor income as hard as Apple is has done you a disservice. Because they have gotten you hot and bothered over things you mistakenly think are unique to Apple when actually producing products for the maximum profits is what the world capital system is all about, and no chipmaker declines the money grab.

Ask anyone trying to buy H100s how sweet, kind, and benevolent Nvidia’s prices are. Ask someone who bought Xeons between 2010 and 2020 if they thought Intel made any financial sacrifices, because they just wanted to get good technology and peoples hands.

Nearly all Apple engineers want to stay at Apple. Investment dollars are coming down harder on chip design companies than the rain Noah faced. any chip designer can go work for any server company they want to for whatever price they asked for. in fact, this probably a number of engineers who sacrificing pay to stay at Apple. OpenAI is offering a king ransom to each and every chip engineer they can grab. if Sam Altman is able to produce a successful product, do you really see indications but he’ll be more benevolent than Apple?

 

[igor_kavinski]

I can see plenty of good, usable products featuring AMD/Intel CPUs and GPUs available at affordable prices. Reason? Competition created by the different OEMs vying for our dollar.

Apple has no competition and whatever they produce, even crappy stuff, has prices wayyyy exaggerated compared to the value they provide.

Nvidia charges the OEMs so much for their graphics card "kits" that they can't really compete on price if they want to stay in business.

Heck, even Snapdragon Elite X laptops are giving 32GB RAM (8448 MT/s!) for prices similar to Apple M2 8GB/512GB laptops.

Clearly, some companies are focused on producing stuff while others are focused on fleecing their customers.

 

[poke01]

Heck, even Snapdragon Elite X laptops are giving 32GB RAM (8448 MT/s!) for prices similar to Apple M2 8GB/512GB laptops.

RAM isn’t everything, you can’t do progressional work or even basic work on these X Elite models other than web browsing. Everything has perceived value the beauty of a free market is the ability to choose and complaining here won’t get you anywhere.

Many choose MacBooks because of macOS and its M chips. The lower end MacBooks always had bad value, the $2000 MacBook Pro with 18GB of RAM with M3 Pro is often on sale for $1600-1700 ands the one to get till Apple makes the Airs more valuable.

As for why AMD/Intel they have to produce affordable products in order to get sales, they don’t even make the same amount of revenue and profit like Apple or Nvidia do because they don’t have mindshare or IP that can make Apple-like revenue.

 

[Glo.]

Look at the bandwidth available via DDR5/X right now.
It's enough for the amount of compute available, and scales upward in an obvious way as that compute scales.

Show me a realistic benchmark where an M3 Max is being throttled by not enough memory bandwidth.

40 GPU cores may not be bandwidth limited, with 16 CPU cores.

But 24 Performance cores, with 80 GPU cores are not going to be bandwdith limited, considering how memory bandwdith intensive Apple architectures are?

And we are talking only about Ultra class chip.

M4 Extreme may combine both of those designs. Will LPDDR still be sufficient?