Early M2 Max benchmarks may have just leaked online

AppleInsider · November 30, 2022 11:43AM

New benchmark results for what may be a new Mac using a M2 Max processor have surfaced, but do not show a dramatic improvement over its predecessor.

The figures on Geekbench are for a device that identifies itself as "Mac14,6." References to this Mac were first spotted back in July 2022, though it remains unclear whether it's a MacBook Pro or another device such as a new Mac Studio.

New Mac's Geekbench 5 scores!
12 cores, 96GB of RAMhttps://t.co/74dOcmgTod pic.twitter.com/PXl8Ul9eVp
-- ShrimpApplePro (@VNchocoTaco)
November 30, 2022

As first spotted by leaker ShrimpApplePro, the Geekbench figures includes details of the device's configuration. As reported, the device features 96GB RAM, which is more than a current MacBook Pro can offer, but less than the Mac Studio.

The CPU is listed as being "Apple M2 Max," and the data includes that it's one 12-core processor running at 3.54 GHz. The single-core score is 1853, and the multi-core score is 13855.

This appears to be the sole M2 Max record on Geekbench. There are obviously many for its year-old predecessor, the M1 Max, and a typical result for that is 1787 single-core, 12826 multi-core.

So it appears from this one example -- assuming it is not fabricated -- that the M2 Max may not offer as significant performance improvement as expected.

In case we haven't been clear enough, the provenance on the benchmark is unclear.

It had been predicted that a new MacBook Pro with an M2 or M2 Max processor would be released before the end of 2022. However, Tim Cook's comments at the last earnings call have made it sound unlikely, and Apple does not often release new hardware in December that it has not already previously launched.

Read on AppleInsider

blastdoor · November 30, 2022 12:52PM

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

polymnia · November 30, 2022 3:13PM

I’ll be quite satisfied with 96 GB of RAM. I can’t replace my 2017 iMac and it’s 64 GB RAM with a MBP until it offers a higher RAM ceiling. I do advanced Photoshop editing in my work that taxes any RAM configuration. I had guessed at this capacity after seeing the M2 MBA 1.5x it’s max RAM. Good to see that guess pan out (?).

Faster processors are fine, but the RAM ceiling of the MBPs prior to Apple Silicon was the real pain point for people like me. Glad to see they continue to move the ball forward.

mpantone · November 30, 2022 4:48PM

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

9secondkox2 · November 30, 2022 5:36PM

M2 is a good setup, just not mind blowing as the m1 max was at launch. M3 has been the one to wait for. Must be why Apple wanted to move m2 to 3nm. Automatic gains across the board, likely with more overhead for higher clocks as well.

96 GB in a notebook would be a barrier breaker. No reason to not get one. Personally, I’ve been holding for a notebook or iMac M series with 128 GB or more, but 96 may be a sweet spot.

The studio is a rip off at its price points (should have just called it Mac mini pro-since that’s what it is-and sold it reasonably) and not really looking to Jump into Mac pro pricing.

The m2 currently doesn’t seem like a confidence point for Apple to jump into a Mac Pro with. So they may be accelerating a run with M3 - which may be where the M2 3nm rumors originated. Even an m2 at 3nm “extreme” setup would perhaps only match nvidia graphics from last year and match the most radical volcanic intel setups on the cpu side. With the Mac Pro taking so long, more is expected of it and Apple wants to make a statement.

Won’t be surprised to see an M3 Extreme Mac Pro announced this summer and available by the fall. M2 based MacBook pros probably launch in February or march.

Marvin · November 30, 2022 5:41PM

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

The M2 is very similar to the M1 in Geekbench:

https://browser.geekbench.com/mac-benchmarks

It is however much faster for GPU (40%) and some video encoding.

These tests would suggest M2 Pro/Max might be the worst outcome expected for an upgrade - delayed to 2023 and using the same N5P process as M2 so the better N3 upgrade might not come until 2024.

There's a possibility they could make the GPU cores on N3 and the CPU cores on N5P but it's more likely they will do N5P.

While it's disappointing not having significant CPU improvements, a 40% GPU boost is a pretty good upgrade. It would have been nice to have it released this year though.

nubus · November 30, 2022 6:32PM

Just as expected. We have seen that iPhone 12/13/14/14 Pro have comparable performance. They all run on 5nm TSMC in different variants. Even with updates there is a limit to 5nm. With M1 Apple made the jump from 14nm and DRR 4 RAM. M2 Pro and Max should offer more memory just like M2 and 8K ProRes.

3nm is where things get interesting again. TSMC N3E will ship next summer but likely only for iPhone 15 Pro. The iPhone 15 will stay on the same tech as 12/13/14. MacBook Pro M3... can only ship when supply is available. That could push M3 into 2024.

mpantone · November 30, 2022 6:43PM

Marvin said:

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

The M2 is very similar to the M1 in Geekbench:

https://browser.geekbench.com/mac-benchmarks

It is however much faster for GPU (40%) and some video encoding.

These tests would suggest M2 Pro/Max might be the worst outcome expected for an upgrade - delayed to 2023 and using the same N5P process as M2 so the better N3 upgrade might not come until 2024.

There's a possibility they could make the GPU cores on N3 and the CPU cores on N5P but it's more likely they will do N5P.

It's highly unlikely that Apple will do this. After all the CPU and GPU cores are on the same die, right? And using split process nodes might end up as a cost increase. How much is Apple willing to pass onto the consumer? There are already plenty of other factors driving price up right now.

Again, this is all ignoring the fact that the purported M2 Max is running the same as the vanilla M2, with no accommodation for the M2 Pro.

It is pretty far fetched to think that Apple will release three M2 SoCs with identical single core performance.

I'm not saying that it can't be done. I'm just saying that the likelihood of this happening is extremely low. And doing so would also put them at risk of being the laughingstock of the semiconductor industry for years to come. I'm thinking that Johny Sroudji wouldn't care for that very much.

Apple has done some bizarre stuff before but I just don't see the business case in Apple releasing an M2 Max SoC that's supposed to be two tiers above the vanilla M2 with an identical Geekbench score, despite the fact that Geekbench is a seriously flawed measurement.

danox · November 30, 2022 7:34PM

Marvin said:

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

The M2 is very similar to the M1 in Geekbench:

https://browser.geekbench.com/mac-benchmarks

It is however much faster for GPU (40%) and some video encoding.

These tests would suggest M2 Pro/Max might be the worst outcome expected for an upgrade - delayed to 2023 and using the same N5P process as M2 so the better N3 upgrade might not come until 2024.

There's a possibility they could make the GPU cores on N3 and the CPU cores on N5P but it's more likely they will do N5P.

While it's disappointing not having significant CPU improvements, a 40% GPU boost is a pretty good upgrade. It would have been nice to have it released this year though.

The form factor is more important at this point not absolute speed ie, a bigger screen iMac and or a Mac Pro with some expansion capability with decent i/o ports.

The strength of the Mac in the past was and is the total package of the OS and hardware working together as one. Computing trucks are needed too.

Marvin · November 30, 2022 9:33PM

mpantone said:

Marvin said:

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

The M2 is very similar to the M1 in Geekbench:

https://browser.geekbench.com/mac-benchmarks

It is however much faster for GPU (40%) and some video encoding.

These tests would suggest M2 Pro/Max might be the worst outcome expected for an upgrade - delayed to 2023 and using the same N5P process as M2 so the better N3 upgrade might not come until 2024.

There's a possibility they could make the GPU cores on N3 and the CPU cores on N5P but it's more likely they will do N5P.

It's highly unlikely that Apple will do this. After all the CPU and GPU cores are on the same die, right?

There's a report saying Intel plans to do this:

https://www.tomshardware.com/news/tiled-design-and-tsmc-n3-will-enable-intel-to-offer-monstrous-igpus

They use tiles. Apple's chip design might prevent them doing the same.

mpantone said:

Marvin said:

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

The purported benchmark is for an M2 Max. If the score and name of the processor are to be believed then there is no performance improvement with the M2 Max. And Apple most certainly will not release a new SoC that has the identical performance as a predecessor.

Remember: M_ < M_ Pro < M_ Max < M_ Ultra

This Geekbench score is likely fake or maybe the SoC's name was incorrectly reported.

My belief is that an M2 Max will need a 15-30% performance uplift over an M2 Pro to make it marketable.

Assuming Apple does not jump process nodes with the M2 Max, I'm guessing that the performance boost on standard integer and floating point tests will be more modest with greater improvements for machine learning tasks.

The M2 is very similar to the M1 in Geekbench:

https://browser.geekbench.com/mac-benchmarks

It is however much faster for GPU (40%) and some video encoding.

These tests would suggest M2 Pro/Max might be the worst outcome expected for an upgrade - delayed to 2023 and using the same N5P process as M2 so the better N3 upgrade might not come until 2024.

There's a possibility they could make the GPU cores on N3 and the CPU cores on N5P but it's more likely they will do N5P.
It is pretty far fetched to think that Apple will release three M2 SoCs with identical single core performance.

I'm not saying that it can't be done. I'm just saying that the likelihood of this happening is extremely low. And doing so would also put them at risk of being the laughingstock of the semiconductor industry for years to come. I'm thinking that Johny Sroudji wouldn't care for that very much.

Apple has done some bizarre stuff before but I just don't see the business case in Apple releasing an M2 Max SoC that's supposed to be two tiers above the vanilla M2 with an identical Geekbench score, despite the fact that Geekbench is a seriously flawed measurement.

They already did this with M1. M1/M1 Pro/M1Max all have the same single core performance. It's the same with the Intel chips, i5/i7/i9 in the same generation have very similar single core.

All that matters for the higher-end chips is better multi-core performance.

M2 Max multi-core (12-core) = 13855
M2 multi-core (8-core) = 8737

12/8 = 1.5x, 13855/8737 = 1.58x

The GPU offers the most value these days so hopefully that will get a significant boost. If M2 managed 40% increase (3.6TFLOPs vs 2.6TFLOPs), M2 Pro and Max can get a similar improvement from increasing GPU core count. This means M2 Pro is 7.2TFLOPs and M2 Max is 14.5TFLOPs. N3 would mean the M3 Max would be close to an Ultra in a laptop.

tenthousandthings · November 30, 2022 9:56PM

I've been waiting for a thread to post this in, I guess it can go here. The tl;dr version is there has been one other unusual, long time interval in the past history of Apple Silicon, between the A10 and the A10X, and that is the only case where Apple switched process nodes in between. This fact could indicate that the reason for the similarly long interval between the M2 and the M2Pro/Max (at least) is because of a process node change.

Below is a history of the A-Series and M-series to date. This is selective, but everything here is accurate and known, not speculative. I don't track every device. For Macs, I only note those that are not transitional. For example, for the M1 I only list the iMac 24" -- every other M1 Mac was transitional. I also track the TSMC process nodes.

A4 (March 2010) iPad 1 :: iPhone 4

A5 (March 2011) iPad 2 :: iPhone 4S

A5X (March 2012) iPad 3

A6 (September 2012) iPhone 5

A6X (October 2012) iPad 4

A7 (September 2013) iPhone 5S :: iPad Air 1

A8 (September 2014) iPhone 6 [TSMC 20nm]

A8X (October 2014) iPad Air 2 [TSMC 20nm]

A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm]

A9X (November 2015) iPad Pro 1 [TSMC 16nm]

A10 (September 2016) iPhone 7 :: iPad 6 :: iPad 7 [TSMC 16nm gen3]

A10X (June 2017) iPad Pro 2 [TSMC 10nm]

A11 (September 2017) iPhone 8, iPhone X [TSMC 10nm]

A12 (September 2018) iPhone XS :: iPad Air 3 :: iPad 8 [TSMC 7nm gen1 "N7"]

A12X (October 2018) iPad Pro 3 [TSMC 7nm gen1 "N7"]

A12Z (March 2020) iPad Pro 4 :: Developer Transition Kit [TSMC 7nm gen1 "N7"]

A13 (September 2019) iPhone 11 :: iPad 9 [TSMC 7nm gen2 "N7P" (P = Plus)]

A14 (October 2020) iPhone 12 :: iPad Air 4 :: iPad 10 [TSMC 5nm gen1 "N5"]

M1 (November 2020) iMac 24" :: iPad Pro 5 [TSMC 5nm gen1 "N5"]

M1 Pro/Max (October 2021) MacBook Pro 14" 16" [TSMC 5nm gen1 "N5"]

M1 Ultra (March 2022) Mac Studio [TSMC 5nm gen1 "N5"]

A15 (September 2021) iPhone 13 :: iPhone 14 [TSMC 5nm gen2 "N5P"]

M2 (June 2022) MacBook Air 4 :: iPad Pro 6 [TSMC 5nm gen2 "N5P"]

A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen3 "N4"]

Notice the long time interval between the A10 and the A10X. In every other case (the A6/A6X, A8/A8X, A9/A9X, and A12/A12X), there is one month between the release of the flagship iPhone SoC and the flagship iPad SoC. Only the A10 and A10X have a nine-month gap between them. How is that different from the others? It's the only one with a process node change. Moreover, it's a big jump, from the third generation of TSMC's 16nm node to its new 10nm node.

The second thing to point out is that all of the M1 SoCs share the same process node. That's currently our only M-series data point, so we can't draw any conclusions, but I've arranged the releases in a way that highlights the fact the M-series has replaced the X variants.

Next, a lot of speculation has been made about future 3nm TSMC process nodes. But TSMC has two still-upcoming 5nm nodes: gen4 "N4P" (P = Plus) and gen5 "N4X" (X = Extreme). The N4X node is especially intriguing with respect to a new Mac Pro: https://pr.tsmc.com/english/news/2895 -- I don't know about you, but this press release, with its emphasis on HPC, advanced packaging, and "the common design rules of the N5 process," reads like a recipe for M2 Ultra/Extreme SoCs.

In conclusion: if the past history of the A-series is any indication, the long, likely eight- or nine-month gap between the M2 and the M2 Pro+ probably indicates a process node change. It still seems too early for TSMC's 3nm tech, so that leaves the N4P and N4X nodes. Here's a wild guess for what the end of my list above will look like prior to the start of a new cycle with the M3.

M2 Pro/Max (March 2023) iMac 24" :: Mac mini 6 :: MacBook Pro 14" 16" [TSMC 5nm gen4 "N4P"]

M2 Ultra/Extreme (June 2023) Mac Studio :: Mac Pro [TSMC 5nm gen5 "N4X"]

blastdoor · November 30, 2022 10:49PM

tenthousandthings said:

I've been waiting for a thread to post this in, I guess it can go here. The tl;dr version is there has been one other unusual, long time interval in the past history of Apple Silicon, between the A10 and the A10X, and that is the only case where Apple switched process nodes in between. This fact could indicate that the reason for the similarly long interval between the M2 and the M2Pro/Max (at least) is because of a process node change.

Below is a history of the A-Series and M-series to date. This is selective, but everything here is accurate and known, not speculative. I don't track every device. For Macs, I only note those that are not transitional. For example, for the M1 I only list the iMac 24" -- every other M1 Mac was transitional. I also track the TSMC process nodes.

A4 (March 2010) iPad 1 :: iPhone 4
A5 (March 2011) iPad 2 :: iPhone 4S
A5X (March 2012) iPad 3

A6 (September 2012) iPhone 5
A6X (October 2012) iPad 4

A7 (September 2013) iPhone 5S :: iPad Air 1

A8 (September 2014) iPhone 6 [TSMC 20nm]
A8X (October 2014) iPad Air 2 [TSMC 20nm]

A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm]
A9X (November 2015) iPad Pro 1 [TSMC 16nm]

A10 (September 2016) iPhone 7 :: iPad 6 :: iPad 7 [TSMC 16nm gen3]
A10X (June 2017) iPad Pro 2 [TSMC 10nm]

A11 (September 2017) iPhone 8, iPhone X [TSMC 10nm]

A12 (September 2018) iPhone XS :: iPad Air 3 :: iPad 8 [TSMC 7nm gen1 "N7"]
A12X (October 2018) iPad Pro 3 [TSMC 7nm gen1 "N7"]
A12Z (March 2020) iPad Pro 4 :: Developer Transition Kit [TSMC 7nm gen1 "N7"]

A13 (September 2019) iPhone 11 :: iPad 9 [TSMC 7nm gen2 "N7P" (P = Plus)]

A14 (October 2020) iPhone 12 :: iPad Air 4 :: iPad 10 [TSMC 5nm gen1 "N5"]
M1 (November 2020) iMac 24" :: iPad Pro 5 [TSMC 5nm gen1 "N5"]
M1 Pro/Max (October 2021) MacBook Pro 14" 16" [TSMC 5nm gen1 "N5"]
M1 Ultra (March 2022) Mac Studio [TSMC 5nm gen1 "N5"]

A15 (September 2021) iPhone 13 :: iPhone 14 [TSMC 5nm gen2 "N5P"]
M2 (June 2022) MacBook Air 4 :: iPad Pro 6 [TSMC 5nm gen2 "N5P"]

A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen3 "N4"]

Notice the long time interval between the A10 and the A10X. In every other case (the A6/A6X, A8/A8X, A9/A9X, and A12/A12X), there is one month between the release of the flagship iPhone SoC and the flagship iPad SoC. Only the A10 and A10X have a nine-month gap between them. How is that different from the others? It's the only one with a process node change. Moreover, it's a big jump, from the third generation of TSMC's 16nm node to its new 10nm node.

The second thing to point out is that all of the M1 SoCs share the same process node. That's currently our only M-series data point, so we can't draw any conclusions, but I've arranged the releases in a way that highlights the fact the M-series has replaced the X variants.

Next, a lot of speculation has been made about future 3nm TSMC process nodes. But TSMC has two still-upcoming 5nm nodes: gen4 "N4P" (P = Plus) and gen5 "N4X" (X = Extreme). The N4X node is especially intriguing with respect to a new Mac Pro: https://pr.tsmc.com/english/news/2895 -- I don't know about you, but this press release, with its emphasis on HPC, advanced packaging, and "the common design rules of the N5 process," reads like a recipe for M2 Ultra/Extreme SoCs.

In conclusion: if the past history of the A-series is any indication, the long, likely eight- or nine-month gap between the M2 and the M2 Pro+ probably indicates a process node change. It still seems too early for TSMC's 3nm tech, so that leaves the N4P and N4X nodes. Here's a wild guess for what the end of my list above will look like prior to the start of a new cycle with the M3.

M2 Pro/Max (March 2023) iMac 24" :: Mac mini 6 :: MacBook Pro 14" 16" [TSMC 5nm gen4 "N4P"]
M2 Ultra/Extreme (June 2023) Mac Studio :: Mac Pro [TSMC 5nm gen5 "N4X"]

I had totally lost sight of that process difference between the A10 and A10X. I had pretty much given up hope that the M2 Pro etc chips would be on anything other than 5nm, but you point to a really interesting historical data point!

I'm still not very optimistic, though. There was a pretty long gap between the M1 and the M1 Pro/Max, too.

bulk001 · December 1, 2022 12:24AM

So what’s it been? A year or two and Apple is already basically at the same place Intel is with Incremental updates spread out over a period of years unable to deliver on a predictable timetable. Yes there are some battery life advantages and the initial jump of the v1 chip but it is not very promising moving forward if this is accurate!

fastasleep · December 1, 2022 1:35AM

bulk001 said:

So what’s it been? A year or two and Apple is already basically at the same place Intel is with Incremental updates spread out over a period of years unable to deliver on a predictable timetable. Yes there are some battery life advantages and the initial jump of the v1 chip but it is not very promising moving forward if this is accurate!

Horseshit. It's been two years, and you can see the exact timeline in the post above yours. The M1 and M1 Pro/Max were 11 months apart. If the M2 Pro/Max comes out in Spring, say March for example, it'll have been 8 months since the M2. There's hardly a timetable much less a predictable or unpredictable one at this point!

FileMakerFeller · December 1, 2022 8:03AM

bulk001 said:

So what’s it been? A year or two and Apple is already basically at the same place Intel is with Incremental updates spread out over a period of years unable to deliver on a predictable timetable. Yes there are some battery life advantages and the initial jump of the v1 chip but it is not very promising moving forward if this is accurate!

... while dealing with the massive disruption brought about by a global pandemic. I don't share your pessimism. Remember that Apple famously ships a great v1.0 and then iterates consistently; some people moan about the lack of regular "OMG" updates but over time the consistency of improvement leads to massive gains.

Also, tenthousandthings pointed out that in 2014 (a mere 8 years ago!) TSMC was using a 20nm process node. I mentally used a swear word when I read that. Astonishing progress to be shipping at 5nm and imminently 3nm in that timeframe. Well done to everyone at TSMC, that is spectacular!

commentzilla · December 1, 2022 10:39AM

mpantone said:

blastdoor said:

The M2 in the MBA gets a single core score of 1899, so these results seem ballpark reasonable to me -- within the range of testing reliability.

This makes zero sense. The two scores are nearly identical.

You’re assuming that’s the only improvement. The ability to host more RAM, energy efficiency and increased graphics power are three possible improvements that will have a significant impact. What if the graphics power is dramatically improved? This is where I expect to see a significant jump.

tenthousandthings · December 1, 2022 12:07PM

blastdoor said:

tenthousandthings said:

I've been waiting for a thread to post this in, I guess it can go here. The tl;dr version is there has been one other unusual, long time interval in the past history of Apple Silicon, between the A10 and the A10X, and that is the only case where Apple switched process nodes in between. This fact could indicate that the reason for the similarly long interval between the M2 and the M2Pro/Max (at least) is because of a process node change.

Below is a history of the A-Series and M-series to date. This is selective, but everything here is accurate and known, not speculative. I don't track every device. For Macs, I only note those that are not transitional. For example, for the M1 I only list the iMac 24" -- every other M1 Mac was transitional. I also track the TSMC process nodes.

A4 (March 2010) iPad 1 :: iPhone 4
A5 (March 2011) iPad 2 :: iPhone 4S
A5X (March 2012) iPad 3

A6 (September 2012) iPhone 5
A6X (October 2012) iPad 4

A7 (September 2013) iPhone 5S :: iPad Air 1

A8 (September 2014) iPhone 6 [TSMC 20nm]
A8X (October 2014) iPad Air 2 [TSMC 20nm]

A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm]
A9X (November 2015) iPad Pro 1 [TSMC 16nm]

A10 (September 2016) iPhone 7 :: iPad 6 :: iPad 7 [TSMC 16nm gen3]
A10X (June 2017) iPad Pro 2 [TSMC 10nm]

A11 (September 2017) iPhone 8, iPhone X [TSMC 10nm]

A12 (September 2018) iPhone XS :: iPad Air 3 :: iPad 8 [TSMC 7nm gen1 "N7"]
A12X (October 2018) iPad Pro 3 [TSMC 7nm gen1 "N7"]
A12Z (March 2020) iPad Pro 4 :: Developer Transition Kit [TSMC 7nm gen1 "N7"]

A13 (September 2019) iPhone 11 :: iPad 9 [TSMC 7nm gen2 "N7P" (P = Plus)]

A14 (October 2020) iPhone 12 :: iPad Air 4 :: iPad 10 [TSMC 5nm gen1 "N5"]
M1 (November 2020) iMac 24" :: iPad Pro 5 [TSMC 5nm gen1 "N5"]
M1 Pro/Max (October 2021) MacBook Pro 14" 16" [TSMC 5nm gen1 "N5"]
M1 Ultra (March 2022) Mac Studio [TSMC 5nm gen1 "N5"]

A15 (September 2021) iPhone 13 :: iPhone 14 [TSMC 5nm gen2 "N5P"]
M2 (June 2022) MacBook Air 4 :: iPad Pro 6 [TSMC 5nm gen2 "N5P"]

A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen3 "N4"]

Notice the long time interval between the A10 and the A10X. In every other case (the A6/A6X, A8/A8X, A9/A9X, and A12/A12X), there is one month between the release of the flagship iPhone SoC and the flagship iPad SoC. Only the A10 and A10X have a nine-month gap between them. How is that different from the others? It's the only one with a process node change. Moreover, it's a big jump, from the third generation of TSMC's 16nm node to its new 10nm node.

The second thing to point out is that all of the M1 SoCs share the same process node. That's currently our only M-series data point, so we can't draw any conclusions, but I've arranged the releases in a way that highlights the fact the M-series has replaced the X variants.

Next, a lot of speculation has been made about future 3nm TSMC process nodes. But TSMC has two still-upcoming 5nm nodes: gen4 "N4P" (P = Plus) and gen5 "N4X" (X = Extreme). The N4X node is especially intriguing with respect to a new Mac Pro: https://pr.tsmc.com/english/news/2895 -- I don't know about you, but this press release, with its emphasis on HPC, advanced packaging, and "the common design rules of the N5 process," reads like a recipe for M2 Ultra/Extreme SoCs.

In conclusion: if the past history of the A-series is any indication, the long, likely eight- or nine-month gap between the M2 and the M2 Pro+ probably indicates a process node change. It still seems too early for TSMC's 3nm tech, so that leaves the N4P and N4X nodes. Here's a wild guess for what the end of my list above will look like prior to the start of a new cycle with the M3.

M2 Pro/Max (March 2023) iMac 24" :: Mac mini 6 :: MacBook Pro 14" 16" [TSMC 5nm gen4 "N4P"]
M2 Ultra/Extreme (June 2023) Mac Studio :: Mac Pro [TSMC 5nm gen5 "N4X"]

I had totally lost sight of that process difference between the A10 and A10X. I had pretty much given up hope that the M2 Pro etc chips would be on anything other than 5nm, but you point to a really interesting historical data point!

I'm still not very optimistic, though. There was a pretty long gap between the M1 and the M1 Pro/Max, too.

Ugh, I forgot another obvious possibility for this hypothetical M2 Pro/Max+ process node change: N4 (already in use for the A16)…

I purposely avoided the question of where M2 Pro/Max would be placed in the history, under A15 or possibly under A16. The split between the iPhone 14 (A15) and the iPhone 14 Pro (A16) is unprecedented, so who knows what that means going forward.

The long gap between the M1 and M1 Pro/Max was likely due to multiple factors—we don’t know the original Apple Silicon release timetable, so we don’t know how long the MacBook Pro 14" 16" launch was actually delayed, not to mention the M1 Ultra and the Mac Studio and Studio Display.

blastdoor · December 1, 2022 1:38PM

tenthousandthings said:

blastdoor said:

tenthousandthings said:

I've been waiting for a thread to post this in, I guess it can go here. The tl;dr version is there has been one other unusual, long time interval in the past history of Apple Silicon, between the A10 and the A10X, and that is the only case where Apple switched process nodes in between. This fact could indicate that the reason for the similarly long interval between the M2 and the M2Pro/Max (at least) is because of a process node change.

Below is a history of the A-Series and M-series to date. This is selective, but everything here is accurate and known, not speculative. I don't track every device. For Macs, I only note those that are not transitional. For example, for the M1 I only list the iMac 24" -- every other M1 Mac was transitional. I also track the TSMC process nodes.

A4 (March 2010) iPad 1 :: iPhone 4
A5 (March 2011) iPad 2 :: iPhone 4S
A5X (March 2012) iPad 3

A6 (September 2012) iPhone 5
A6X (October 2012) iPad 4

A7 (September 2013) iPhone 5S :: iPad Air 1

A8 (September 2014) iPhone 6 [TSMC 20nm]
A8X (October 2014) iPad Air 2 [TSMC 20nm]

A9 (September 2015) iPhone 6S :: iPad 5 [TSMC 16nm]
A9X (November 2015) iPad Pro 1 [TSMC 16nm]

A10 (September 2016) iPhone 7 :: iPad 6 :: iPad 7 [TSMC 16nm gen3]
A10X (June 2017) iPad Pro 2 [TSMC 10nm]

A11 (September 2017) iPhone 8, iPhone X [TSMC 10nm]

A12 (September 2018) iPhone XS :: iPad Air 3 :: iPad 8 [TSMC 7nm gen1 "N7"]
A12X (October 2018) iPad Pro 3 [TSMC 7nm gen1 "N7"]
A12Z (March 2020) iPad Pro 4 :: Developer Transition Kit [TSMC 7nm gen1 "N7"]

A13 (September 2019) iPhone 11 :: iPad 9 [TSMC 7nm gen2 "N7P" (P = Plus)]

A14 (October 2020) iPhone 12 :: iPad Air 4 :: iPad 10 [TSMC 5nm gen1 "N5"]
M1 (November 2020) iMac 24" :: iPad Pro 5 [TSMC 5nm gen1 "N5"]
M1 Pro/Max (October 2021) MacBook Pro 14" 16" [TSMC 5nm gen1 "N5"]
M1 Ultra (March 2022) Mac Studio [TSMC 5nm gen1 "N5"]

A15 (September 2021) iPhone 13 :: iPhone 14 [TSMC 5nm gen2 "N5P"]
M2 (June 2022) MacBook Air 4 :: iPad Pro 6 [TSMC 5nm gen2 "N5P"]

A16 (September 2022) iPhone 14 Pro [TSMC 5nm gen3 "N4"]

Notice the long time interval between the A10 and the A10X. In every other case (the A6/A6X, A8/A8X, A9/A9X, and A12/A12X), there is one month between the release of the flagship iPhone SoC and the flagship iPad SoC. Only the A10 and A10X have a nine-month gap between them. How is that different from the others? It's the only one with a process node change. Moreover, it's a big jump, from the third generation of TSMC's 16nm node to its new 10nm node.

The second thing to point out is that all of the M1 SoCs share the same process node. That's currently our only M-series data point, so we can't draw any conclusions, but I've arranged the releases in a way that highlights the fact the M-series has replaced the X variants.

Next, a lot of speculation has been made about future 3nm TSMC process nodes. But TSMC has two still-upcoming 5nm nodes: gen4 "N4P" (P = Plus) and gen5 "N4X" (X = Extreme). The N4X node is especially intriguing with respect to a new Mac Pro: https://pr.tsmc.com/english/news/2895 -- I don't know about you, but this press release, with its emphasis on HPC, advanced packaging, and "the common design rules of the N5 process," reads like a recipe for M2 Ultra/Extreme SoCs.

In conclusion: if the past history of the A-series is any indication, the long, likely eight- or nine-month gap between the M2 and the M2 Pro+ probably indicates a process node change. It still seems too early for TSMC's 3nm tech, so that leaves the N4P and N4X nodes. Here's a wild guess for what the end of my list above will look like prior to the start of a new cycle with the M3.

M2 Pro/Max (March 2023) iMac 24" :: Mac mini 6 :: MacBook Pro 14" 16" [TSMC 5nm gen4 "N4P"]
M2 Ultra/Extreme (June 2023) Mac Studio :: Mac Pro [TSMC 5nm gen5 "N4X"]

I had totally lost sight of that process difference between the A10 and A10X. I had pretty much given up hope that the M2 Pro etc chips would be on anything other than 5nm, but you point to a really interesting historical data point!

I'm still not very optimistic, though. There was a pretty long gap between the M1 and the M1 Pro/Max, too.

Ugh, I forgot another obvious possibility for this hypothetical M2 Pro/Max+ process node change: N4 (already in use for the A16)…

I purposely avoided the question of where M2 Pro/Max would be placed in the history, under A15 or possibly under A16. The split between the iPhone 14 (A15) and the iPhone 14 Pro (A16) is unprecedented, so who knows know what that means going forward.

The long gap between the M1 and M1 Pro/Max was likely due to multiple factors—we don’t know the original Apple Silicon release timetable, so we don’t know how long the MacBook Pro 14" 16" launch was actually delayed, not to mention the M1 Ultra and the Mac Studio and Studio Display.

True, the pandemic messed up a lot of schedules (and still is).

I guess at this point it’s pretty much a crapshoot

TheObannonFile · December 1, 2022 3:54PM

The numbers don’t check out here on CPU.

if they improved the single core, and then added 2 more CPU cores (high performance I assume), the multi core should be a good bit higher than that.

somebody help me out in the math here.

tht · December 1, 2022 4:29PM

TheObannonFile said:

The numbers don’t check out here on CPU.

if they improved the single core, and then added 2 more CPU cores (high performance I assume), the multi core should be a good bit higher than that.

somebody help me out in the math here.

It's 8+4 not 10+2. Just 8 performance cores. They added 2 more efficiency cores according to rumors.

TheObannonFile · December 1, 2022 4:39PM

tht said:

TheObannonFile said:

The numbers don’t check out here on CPU.

if they improved the single core, and then added 2 more CPU cores (high performance I assume), the multi core should be a good bit higher than that.

somebody help me out in the math here.

It's 8+4 not 10+2. Just 8 performance cores. They added 2 more efficiency cores according to rumors.

Aaaaaahhhhhhhhh, ok, that makes a lot more sense on the math.

I’m curious, are the additional high efficiency cores there to help so as many tasks as possible on less power to extend overall battery life? I can’t think of another reason why to add additional efficiency cores.

Early M2 Max benchmarks may have just leaked online

Comments