Got a tip for us?
Let us know
Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
M4+ Chip Generation - Speculation Megathread [MERGED]
- Thread starter GoetzPhil
- Start date
- Sort by reaction score
How many inch does a MacBook need to fit the ULTRA chip?
How can it be that bigger chips from Intel, AMD... can fit in smaller devices?
Apple is using the old ARMv8.5
A lot of people are still waiting for the ARMv9.4 or at least ARMv9.2, which will be a very big improvement.
The M4 will have hopefully ARMv9.5 !!
Maybe you should actually look at the thing and it will become clear why you’re wrong.
Attachments
Screen size is relatively immaterial. the issue would primarily be how thick ('size') the luggable desktop replacement would be. But is really wouldn't be a "Macbook" anymore in any modern sense of the term.
Because it is not. Intel and AMD don't have bigger mobile chips that a Ultra. The Ultra is in similar class size in terms of die area coverage as a Nvidia 4090. Even Nvidia doesn't put the 4090 in laptops. Their mobile "4090" is a rebadged, underclocked 4080.
The intel chips that are in similar ballpart as a Ultra is a Xeon SP package. Again ... doesn't show up at all in laptops.
The two "Max" chips together bascially cover the same area as the reticle limit for fab production. That is about as BIG as you can make a single chip. It isn't in the 'small' or even 'medium sized' category at all.
The mobile chips that AMD and Intel sell are in the range of the plain Mn and Mn Pro dies , not the Ultra.
" ... , Apple could presumably fit as many as 415 M3 dies on a single 300-mm wafer, which indicates a die size of around 146 mm^2. ...
By contrast, AMD's Phoenix (which has a similar complexity) has a die size of 178 mm^2. We can only guess whether Apple's M3 is cheaper to make than AMD's Phoenix based on rumors about TSMC's quotes, but smaller chips are typically easier to yield and produce. ... "
Apple Spent $1 Billion to Tape Out New M3 Processors: Analyst
Very few companies can afford a platform like Apple's M3.
Mn Pro in in 200+ mm^2 range and Max is in approximateily 400mm^2 range. Mn Max class dies are as bigger as the top end GPU dies that AMD/Nvidia produce. It is nothing like the laptop 'CPU' dies.
In addition, because the all the RAM is mounted flat and on the top of the overall package the pragmatic footprint tends to be even larger.
So far, it looks like Apple has been concentrating on improving GPU performance, not as much on CPU. If you divide the GB6 ST score by the clock speed, the P-cores (upon which ST is based) do not appear to be gaining in IPC at all from M1 through M3. Perhaps they have a card up their sleeve in the near future, to make big IPC gains, but right now CPU progress is at a standstill on that front, at least for the P-cores. They may have hit some sort of wall that is extremely difficult to get over. Still, compared to x86 IPC, Apple has a pretty good lead, and they are not likely to lose it.
Well, the P cores at least. E cores seem to have been getting good uplifts each gen still.
Anyone knows what is Apple's P core IPC at? Is it close to retiring 1 per clock already? If so, the only way to improve performance would be increasing clock speed? I don't think it is realistically possible to increase IPC beyond 1?
Ya maybe my usage will change a bit, but most of my time is just spent with basic web browsing, office work, excel, etc. I don't do too much that requires multicore right now.
Eventually plan to get to more video work and coding again though where it can help. I'll probably look to upgrade sometime around 2027-2028 or so when this model is about 5-6 years old. By then hopefully it'll be about double - that's requires about 12% per year increases in single threaded performance.
We'll also see how AI plays into the picture. Feels like most important AI stuff will run in the cloud though. I'm not sure how much local AI processing will help.
Web browsing does, I think, benefit from MC performance, as a lot of page loading involves getting all the page resources, as much at once as possible. And good quality productivity apps are typically multithreaded, which can mean MC (spreadsheets especially, I would imagine). Single-threaded apps are hugely easier to write, but macOS is built to make threading less nightmarish.
Is the consensus after the M3 MBP launch with an M3 actually based on the newest A16 architecture that Apple will continue aligning M(x) Mac CPU updates with A(x) iPhone CPU updates going forward? So a 12 month MBP CPU refresh cycle with M4 in fall '24 (w/A18), M5 in fall '25 (w/A19) etc.?
It's been weird seeing the M(x) CPU architecture utilized in MacBook Pros lagging 1-2 generations behind the A(x) CPU architecture since the arm MacBook Pros launched and it's exciting to see Apple finally get the architecture they use for their Macbook Pro lineup synced up with the iPhone lineup's architecture.
It's been weird seeing the M(x) CPU architecture utilized in MacBook Pros lagging 1-2 generations behind the A(x) CPU architecture since the arm MacBook Pros launched and it's exciting to see Apple finally get the architecture they use for their Macbook Pro lineup synced up with the iPhone lineup's architecture.
The consensus the M3 chip is based on the A17. Comparison to the A16 keeps getting mentioned and then debunked.
MacBook Pro 16-inch M4 MAX
CPU: 24 cores (20 + 4)
GPU: 50 cores
Neural Engine: 32 cores!! 👍
RAM: 160 GB
SSD: 16 TB!! 👍
ARMv9.5!! 👍
Also:
pext
hyperthreading 👍
vnni512 👍
vnni256
avx512 👍
avxvnni
avx2
sse41
ssse3
sse2
sse
mmx
MacBook Pro 16-inch M4 MAX
CPU: 24 cores (20 + 4)
GPU: 50 cores
Neural Engine: 32 cores!! 👍
RAM: 160 GB
SSD: 16 TB!! 👍
ARMv9.5!! 👍
Also:
pext
hyperthreading 👍
vnni512 👍
vnni256
avx512 👍
avxvnni
avx2
sse41
ssse3
sse2
sse
mmx
Still no AltiVec. Steve would never have allowed this. Please resign Mr. Cook.
Overheard on the ATP podcast (and have seen other references to this on the web) that future versions of ASML NA EUV machines will halve the size of the reticle limit. Wouldn't that mean that whatever AS is using that TSMC/ASML combination needs to be smaller (in x-y dimensions) than any existing Ultra chip today?
Does that mean stacking Max chips in Ultras by 2nm?
Does that mean stacking Max chips in Ultras by 2nm?
This is no secret, has been known for years.
(a) Moving to high-NA EUV requires switching the reticle from 26mm x 33mm, so 858 mm2, to 26mm x 16.5mm, so 429 mm2. The first commercial high-NA EUV machines are scheduled to be delivered this year, and I assume that's on track. (A machine has already been delivered to IMEC.) The expectation is that volume manufacturing will begin in 2025 or 2026, so this is not an issue until then.
(b) Ultra's are created from two separate Max chiplets, so CONCEPTUALLY they are not affected by this.
(c) The size of an M1 Max is 432mm^2, about 22mmx20mm. I can't find the area of M2 or M3 Max, but I assume they are comparable. I assume going forward the Max will be designed to fit the reticle limit (including becoming more rectangular, less square, and slightly smaller) and will, for the near future anyway, stay a single-chip design.
BITCOIN?!? I thought you had to use the new improved APPLECOIN!
Well, M3 launched Q4 2023.
So that would place M4 in Q2 2025. Which would also mean the M2 was a quarter late due to Covid Supply chain issues - making it closer to an 18 month release cycle. I would also expect M5 in Q4 2026, but not 100% certain.
Also - I think M4 will be on N3E, and M5 will be N2. Especially with TSMC experiencing delays to N2.
Additionally - looking at the Apple line up currently - I'm thinking they're going to kill the Mx Pro line; likely with M5. It has no place in the current Mac Lineup.
Here is what I think:
M4 - 10 CPU Cores - 4 Perf, 6 efficiency, 14 GPU Cores (Cut down config with 12), starting out RAM, 12GB; max is 32GB (128 bit)
M4 Pro - 12 CPU Cores - 6 perf, 6 efficiency, 20 GPU Cores (Cut down config with 18), starting out RAM, 18GB; max is 48GB (192 bit)
M4 Max - 20 CPU Cores - 16 Perf, 4 efficiency, 52 GPU Cores (Cut down configs with 36 and 44 GPU Cores); max is 128GB (512 bit)
M4 Ultra - 2 x M4 Max - max ram is 256GB (2 x 512 bit)
M4 Extreme - 4 x M4 Max - with a IO Chip too for more PCIe lanes - max ram is 512GB (4 x 512 bit)
Apple Dedicated GPU based on the 4th Gen Desktop Apple Silicon:
224 GPU Cores: 4 x 64 modules (Cut down to 56 GPU Cores each), and a 256 bit bus (32/48GB) - $2499/$2999
256 GPU Cores: 4 x 64 GPU Modules + 4 x 64MB L3 Cache and a 256 bit bus (32/48GB) - $3299/$3799
336 GPU Cores: 6 x 64 GPU Modules (Cut down to 56 cores each) + 6 x 64MB L3 Cache and a 384 bit bus (48/72GB) - $4999/$5499
384 GPU Cores: 6 x 64 GPU Modules + 6 x 64MB L3 Cache and a 384 bit bus (48/72GB) - $6499/$6999
448 GPU Cores: 8 x 64 GPU Modules (Cut down to 56 cores each) + 8 x 64MB L3 Cache and a 512 bit bus (64/96GB) - $7999/$8499
512 GPU Cores: 8 x 64 GPU Modules + 8 x 64MB L3 Cache and a 512 bit bus (64GB/96GB/128GB/192GB) - $9499/$9999/$10499/$11999
M5 - 12 CPU Cores - 6 Perf, 6 efficiency, 24 GPU Cores (Cut down configs with 14 and 18), starting out RAM, 12GB; max is 64GB (128 bit)
M5 Max - 20 CPU cores - 16 Perf, 4 efficiency, 64 GPU Cores (Cut down configs with 48 and 56), starting out RAM, 48GB; max is 256GB (512 bit)
M5 Ultra - 2 x M5 Max - max ram is 512GB (2 x 512 bit)
M5 Ultra alt design 4 x (8 CPU Cores - 6 perf, 2 efficiency, 40 GPU Cores) - (Cut down configs with 112 GPU Cores and 136 GPU Cores) - max ram is 768GB (4 x 384 bit)
M5 Extreme: 8 x (8 CPU Cores - 6 perf, 2 efficiency, 40 GPU Cores) - (Cut down config with 272 GPU Cores) - max ram is 1.5TB (8 x 384 bit)
Apple Dedicated GPU based on the 5th Gen Desktop Apple Silicon:
432 GPU Cores: 6 x 80 GPU Modules (Cut down to 72 cores each) + 6 x 64MB L3 Cache and a 384 bit bus (48GB/96GB) - $2999/$3499
480 GPU Cores: 6 x 80 GPU Modules + 6 x 64MB L3 Cache and a 384 bit bus (48GB/96GB/192GB) - $3999/$4499/$5499
576 GPU Cores: 8 x 80 GPU Modules (Cut down to 72 cores each) + 8 x 64MB L3 Cache and a 512 bit bus (64GB/128GB/256GB) - $5299/$6099/$7499
640 GPU Cores: 8 x 80 GPU Modules + 8 x 64MB L3 Cache and a 512 bit bus (128GB/256GB) - $7799/$8499
864 GPU Cores: 12 x 80 (Cut down to 72 each) + 8 x 64MB L3 Cache + 4096 bit HBM4 (192/384GB) - $8999/$10499
960 GPU Cores: 12 x 80 + 12 x 64MB L3 Cache + 6144 bit HBM4 (288 or 576GB) - $11999/$13999
Here is what I think:
M4 - 10 CPU Cores - 4 Perf, 6 efficiency, 14 GPU Cores (Cut down config with 12), starting out RAM, 12GB; max is 32GB (128 bit)
M4 Pro - 12 CPU Cores - 6 perf, 6 efficiency, 20 GPU Cores (Cut down config with 18), starting out RAM, 18GB; max is 48GB (192 bit)
M4 Max - 20 CPU Cores - 16 Perf, 4 efficiency, 52 GPU Cores (Cut down configs with 36 and 44 GPU Cores); max is 128GB (512 bit)
M4 Ultra - 2 x M4 Max - max ram is 256GB (2 x 512 bit)
M4 Extreme - 4 x M4 Max - with a IO Chip too for more PCIe lanes - max ram is 512GB (4 x 512 bit)
Apple Dedicated GPU based on the 4th Gen Desktop Apple Silicon:
224 GPU Cores: 4 x 64 modules (Cut down to 56 GPU Cores each), and a 256 bit bus (32/48GB) - $2499/$2999
256 GPU Cores: 4 x 64 GPU Modules + 4 x 64MB L3 Cache and a 256 bit bus (32/48GB) - $3299/$3799
336 GPU Cores: 6 x 64 GPU Modules (Cut down to 56 cores each) + 6 x 64MB L3 Cache and a 384 bit bus (48/72GB) - $4999/$5499
384 GPU Cores: 6 x 64 GPU Modules + 6 x 64MB L3 Cache and a 384 bit bus (48/72GB) - $6499/$6999
448 GPU Cores: 8 x 64 GPU Modules (Cut down to 56 cores each) + 8 x 64MB L3 Cache and a 512 bit bus (64/96GB) - $7999/$8499
512 GPU Cores: 8 x 64 GPU Modules + 8 x 64MB L3 Cache and a 512 bit bus (64GB/96GB/128GB/192GB) - $9499/$9999/$10499/$11999
M5 - 12 CPU Cores - 6 Perf, 6 efficiency, 24 GPU Cores (Cut down configs with 14 and 18), starting out RAM, 12GB; max is 64GB (128 bit)
M5 Max - 20 CPU cores - 16 Perf, 4 efficiency, 64 GPU Cores (Cut down configs with 48 and 56), starting out RAM, 48GB; max is 256GB (512 bit)
M5 Ultra - 2 x M5 Max - max ram is 512GB (2 x 512 bit)
M5 Ultra alt design 4 x (8 CPU Cores - 6 perf, 2 efficiency, 40 GPU Cores) - (Cut down configs with 112 GPU Cores and 136 GPU Cores) - max ram is 768GB (4 x 384 bit)
M5 Extreme: 8 x (8 CPU Cores - 6 perf, 2 efficiency, 40 GPU Cores) - (Cut down config with 272 GPU Cores) - max ram is 1.5TB (8 x 384 bit)
Apple Dedicated GPU based on the 5th Gen Desktop Apple Silicon:
432 GPU Cores: 6 x 80 GPU Modules (Cut down to 72 cores each) + 6 x 64MB L3 Cache and a 384 bit bus (48GB/96GB) - $2999/$3499
480 GPU Cores: 6 x 80 GPU Modules + 6 x 64MB L3 Cache and a 384 bit bus (48GB/96GB/192GB) - $3999/$4499/$5499
576 GPU Cores: 8 x 80 GPU Modules (Cut down to 72 cores each) + 8 x 64MB L3 Cache and a 512 bit bus (64GB/128GB/256GB) - $5299/$6099/$7499
640 GPU Cores: 8 x 80 GPU Modules + 8 x 64MB L3 Cache and a 512 bit bus (128GB/256GB) - $7799/$8499
864 GPU Cores: 12 x 80 (Cut down to 72 each) + 8 x 64MB L3 Cache + 4096 bit HBM4 (192/384GB) - $8999/$10499
960 GPU Cores: 12 x 80 + 12 x 64MB L3 Cache + 6144 bit HBM4 (288 or 576GB) - $11999/$13999
What no room temperature qbits for quantum computing?
Qbits at room temperature hasn't been proven to work yet - likely looking impossible for at least the next 20 years.
If I got an M3 max now, which is plenty of power for me, will the future M4 series be able to do things that my M3 max cannot? Or will it just be that the M4 max has more neural engine, cores, etc. that improves AI functions?
I know we’re just speculating here, but curious for your thoughts.
I know we’re just speculating here, but curious for your thoughts.
I mean. The cpu is Turing complete so in a sense all new chip features are about doing it faster. You can do ray tracing on a Super Nintendo. You just can’t do real time ray tracing. New features in a chip may speed up tasks orders of magnitudes relative to brute forcing it with the general compute potential of the cpu
Now that I’m done being pedantic I do think there’s a decent chance there’ll be some exclusive chip features Apple will segment some future AI features around. I don’t think M4 will be required to get the most out of macOS 15. But I do think some foundational blocks will be in place that will be utilized for M4+ exclusives in macOS 16 or 17 onwards. Or third party applications making use of relevant APIs possibly before then.
But I don’t think it’ll necessarily be that impactful. You can already run massive LLM and whatnot with the GPU and even CPU power of M3 Max for example. Apple just would never ship that as part of the system running on device due to power consumption. Is my belief anyway.
Now that I’m done being pedantic I do think there’s a decent chance there’ll be some exclusive chip features Apple will segment some future AI features around. I don’t think M4 will be required to get the most out of macOS 15. But I do think some foundational blocks will be in place that will be utilized for M4+ exclusives in macOS 16 or 17 onwards. Or third party applications making use of relevant APIs possibly before then.
But I don’t think it’ll necessarily be that impactful. You can already run massive LLM and whatnot with the GPU and even CPU power of M3 Max for example. Apple just would never ship that as part of the system running on device due to power consumption. Is my belief anyway.