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

 

[Eug]

Holy crap! Up until today, you guys had me believing M4 cores would essentially be tweaked M3 cores.

I wonder if the binned 9-core variant is roughly the same multi-core CPU speed as M3 at about 12000 in GB 6.3.

We will likely have to wait until next week for those benches though. All the current benches are not surprisingly 10-core.

 

[poke01]

I think at this point I have to say GB6 is no longer a good benchmark.

It should have been called GB7 if something as important as SME was added. It skews the results.

I'm waiting for more workloads but nice to see Apple include SME.

 

[Nothingness]

I hope it has SVE2 also then.

Up until now there has been no impetus for OSS to actually start playing with SVE2, especially if they already have comprehensive NEON code paths.

It's not that obvious. I guess they implemented SME as they did for AMX: as a separate unit shared by multiple cores. So even if they added SVE, it would likely be the subset required for streaming mode and only be run on the SME separate unit. That still could be good for ST workloads, but less so for MT ones.

We need more information, and first, as @gdansk said, runs with older versions of GB.

 

[Nothingness]

I think at this point I have to say GB6 is no longer a good benchmark.

It should have been called GB7 if something as important as SME was added. It skews the results.

I'm waiting for more workloads but nice to see Apple include SME.

Nah! Just do what @SarahKerrigan did: compare subtests 😀

But I agree it's unfortunate.

 

[SarahKerrigan]

Nah! Just do what @SarahKerrigan did: compare subtests 😀

But I agree it's unfortunate.

Probably a inevitable issue, though - new extensions are added by vendors all the time. I'm not sure what the solution is. SPEC doesn't prohibit submission with whatever ops the user wants, but vendor SPEC submissions are also kinda worthless.

Perhaps, if Geekbench is wedded to the model of doing binary distribution, it might be good to have "benchmark version" and "executable version" as separate things, and the inclusion of new extensions would bump the executable version? I know this is sorta what they already do (with minor version bumps denoting a new build) but it seems like it's causing confusion.

 

[Eug]

I think at this point I have to say GB6 is no longer a good benchmark.

It should have been called GB7 if something as important as SME was added. It skews the results.

I'm waiting for more workloads but nice to see Apple include SME.

FWIW, this is what Geekbench had to say about it:

Geekbench Blog

www.geekbench.com

Geekbench 6.3
April 11, 2024 - John Poole

The latest version of Primate Labs’ cross-platform benchmark features the following changes:

• Introduce support for Arm Scalable Matrix Extensions (SME) instructions. Geekbench 6.3 includes SME implementations of the matrix multiplication kernels used by the Geekbench 6 machine learning workloads.
• Reduce Horizon Detection run-to-run variability. On Android devices running Geekbench 6.2 and earlier, Horizon Detection demonstrated significant run-to-run variability.. This issue is fixed in Geekbench 6.3.
• Disable OpenCLOn12. OpenCLOn12 causes the Geekbench application to crash at launch on recent hardware (e.g., Intel Meteor Lake laptops). Support for OpenCLOn12 is disabled until the underlying issues in OpenCLOn12 are fixed.

For systems without SME instructions, Geekbench 6.3 CPU Benchmark scores are comparable with Geekbench 6.1 and Geekbench 6.2 scores. Systems with SME instructions enabled will score higher in Geekbench 6.3 than in earlier Geekbench versions.. Geekbench 6.3 GPU Benchmark scores are compatible with Geekbench 6.2 for all systems.

 

[Eug]

Not necessarily the most appropriate comparison since it's M3 Max (with active cooling) against M4 (with passive cooling), but here you go.

Clock speed increase 8.7%.
IPC increase overall 7.6%.
IPC increase excluding SME 3.0%.

https://twitter.com/x/status/1788364108737466609

 

[trivik12]

Still crazy impressive numbers considering M3 barely released 7 months back and N3E is less dense than N3B (but more efficient). But it would be interesting to see apples to apples comp when M4 MBA releases and benched. I am not sure if that will happen this year as M3 MBA barely launched. Is Apple known to refresh their laptops that soon?

still they are a benchmark for PPW. I wonder how X Elite, Strix Point and Lunar Lake does in this regard.

 

[poke01]

here is GB5

iPad16,6 - Geekbench

Benchmark results for an iPad16,6 with an ARM processor.

browser.geekbench.com

 

[Eug]

Still crazy impressive numbers considering M3 barely released 7 months back and N3E is less dense than N3B (but more efficient). But it would be interesting to see apples to apples comp when M4 MBA releases and benched. I am not sure if that will happen this year as M3 MBA barely launched. Is Apple known to refresh their laptops that soon?

Yes and no. I posed that question earlier but there are some considerations. Just rambling here...

1. Apple has occasionally updated their laptops really quickly, in well under a year. This has more typically been the MacBook Pros though IIRC.

2. OTOH, in the past Apple has sometimes let the MacBook Air languish with no CPU updates for years, even though it was always a strong seller due to price. I guess they could get away with it, because a lot of the low end SKUs for the MBA get purchased by students and grandmas and such, who don't usually care much about performance.

3. Apple just announced that the 13" MBA and 15" MBA are both the world's bestsellers in those size tiers. Keeping it fresh might help it continue that sales momentum.

4. But what does "keeping it fresh" mean? They don't necessarily have to be updated within the year. Even if they don't update it until mid 2025, it would likely continue to sell well. And in fact, in terms of performance, for those who care, M3 CPU likely performs roughly as well as M4 9-core anyway.

5. If N3E yields at 4.4 GHz are an issue, and if Apple did ship M4 MacBook Airs, they'd also start at 9-core, probably for the 8 GB tier. They'd reserve 10-core for the higher end configs.

6. Is N3B yield bad enough that they want to scrap it altogether? Would they really kill off M3 MacBook Air, M3 MacBook Pro, M3 Pro MacBook Pro, and M4 Max MacBook Pro all this soon?

7. Does a CPU core binned M4 negate the need for M4 Pro? Probably not because of the display support and video encoding speed.

I could see a scenario where they update the MacBook Pro quickly, killing off the production of M3 Pro and M3 Max, but keeping the M3 MacBook Air around longer. The MacBook Pro would get M4 10c, M4 Pro, and M4 Max, whereas the MacBook Air would stay on M3. This wouldn't necessarily be cheaper, but would be useful for market segmentation.

Thus I could see something like this happening:

2024 Q4 / 2025 Q1
MacBook Pro M4 10c, M4 Pro, M4 Max
iMac M4 9c & 10c
Mac mini M4 9c & 10c, M4 Pro

2025 Q2
Mac Studio M4 Max, M4 Ultra
Mac Pro M4 Ultra

2025 H2
MacBook Air M4 9c & 10c

However, some of the pundits disagree and think the MBAs would get refreshed before the Mac Studio and Mac Pro. That does have some support, because if TSMC wants to stop all N3B production asap, then it would make sense to just kill off M3 at about the same time they kill off M3 Pro and M3 Max.

An analogous situation to M3 series Macs would be the iPad 3, sort of. iPad 3 was a performance gimped release with A5X, and then 7 months later they released an ungimped iPad 4 with A6X. A5X was a very short lived chip. M3 isn't gimped in terms of performance, but it's likely still problematic in terms of cost/yield and TSMC's possible reluctance in continuing its manufacturing on N3B.

If M3 is discontinued, what would they do for the "cheap" US$999 MacBook Air then? They could just keep selling the M2 version. So, in the scenario where they kill off all M3 series chips, they could just sell the M2 MacBook Air along side the M4 MacBook Air.

 

[Henry swagger]

https://twitter.com/x/status/1788338408445919698

Even better

iPad16,6 - Geekbench

Benchmark results for an iPad16,6 with an ARM processor.

browser.geekbench.com

Impressive.. apple has a 65% ipc advantage on a per clock basis over intel and amd.. intel and amd need wider cores.. because if apple reach 5ghz its gonna be gard to catch up

 

[Doug S]

Impressive.. apple has a 65% ipc advantage on a per clock basis over intel and amd.. intel and amd need wider cores.. because if apple reach 5ghz its gonna be gard to catch up

Given the way they've been ramping clocks the last few years, they'll exceed 5 GHz in two years. At some point I have to think this is going to be problematic for Apple's focus on low power. Curious to see how this does power-wise when running at max clock vs M3.

 

[Doug S]

However, some of the pundits disagree and think the MBAs would get refreshed before the Mac Studio and Mac Pro. That does have some support, because if TSMC wants to stop all N3B production asap, then it would make sense to just kill off M3 at about the same time they kill off M3 Pro and M3 Max.

If they want to kill off M3 as quickly as possible why did they just release the new MBA two months ago?

I think they always want a little distance between MBP and MBA, so the MBA will tend to be upgraded a bit later after the initial surge in sales from the new MBPs. So if they do new Macbook Pro this summer like I assume they will (and M4 based Studio/Pro will also arrive then or this fall at the latest) they could do an M4 MBA this fall. The thing is though, they could have achieved the same thing leaving MBA on M2 for a few months more so while I can see why they wouldn't want to sell M4 MBAs before they started selling M4 MBPs, why did they think they had to upgrade MBA in March?

The truth is we have no idea what the Apple/TSMC deal looks like, especially for N3B with that KGD pricing and TSMC's clear incentives to move everyone off N3B and forget it ever happened. Maybe Apple prepurchased some amount of N3B capacity and they weren't going to use enough wafers between iPhone 15 Pro/Max and M3 Mac sales, so they added a higher volume product in MBA to use it up? While I have heard exactly zero rumors suggesting something like this and am just using that scenario as an example, sometimes decisions that make little sense viewed from the outside make perfect sense if you know all the details behind it. There's clearly a missing piece or two from this puzzle.

 

[Eug]

If they want to kill off M3 as quickly as possible why did they just release the new MBA two months ago?

I think they always want a little distance between MBP and MBA, so the MBA will tend to be upgraded a bit later after the initial surge in sales from the new MBPs. So if they do new Macbook Pro this summer like I assume they will (and M4 based Studio/Pro will also arrive then or this fall at the latest) they could do an M4 MBA this fall. The thing is though, they could have achieved the same thing leaving MBA on M2 for a few months more so while I can see why they wouldn't want to sell M4 MBAs before they started selling M4 MBPs, why did they think they had to upgrade MBA in March?

The truth is we have no idea what the Apple/TSMC deal looks like, especially for N3B with that KGD pricing and TSMC's clear incentives to move everyone off N3B and forget it ever happened. Maybe Apple prepurchased some amount of N3B capacity and they weren't going to use enough wafers between iPhone 15 Pro/Max and M3 Mac sales, so they added a higher volume product in MBA to use it up? While I have heard exactly zero rumors suggesting something like this and am just using that scenario as an example, sometimes decisions that make little sense viewed from the outside make perfect sense if you know all the details behind it. There's clearly a missing piece or two from this puzzle.

Agreed. The rumours of M4 coming out this early made me wonder, if true then why does M3 exist at all?

 

[junjie1475]

amarioguy (@amarioguy@treehouse.systems)

@bshanks@mastodon.social @never_released@mastodon.social @marcan @dougall@mastodon.social @saagar@saagarjha.com Donan is *definitely* the M4, there's even new P and E core generations to go with it in another Xcode header file.

social.treehouse.systems

Can't wait for your tests @junjie1475 via Geekerwan

Was thinking of including GPU microarchitecture stuff(both compute and graphics), but may take some times.

 

[mikegg]

Seems legit. M2 was Avalanche + Blizzard on TSMC N5P, while M3 was Everest + Sawtooth on N3. So you have µarch plus process improvements with that generation. Apple claimed the Everest performance cores in the M3 were up to 15% faster than Avalanche in the M2, and the Sawtooth efficiency cores were up to 30% faster than Blizzard. They also made the claim then that:

Probably a similar deal with the comparison between M4 and M2, I'm sure you can imagine what the marketing slide comparing the curves would look like. But the M4 has 50% more of the cores that were 30% faster and gets another process improvement good for +5% speed at the same power or -7% power at the same speed.


Apple's not talking about single-threaded performance, they're talking about peak multi-threaded. M3 already beat M2 by more than 25% in that scenario because the e-cores saw more of an uplift than the p-cores, and the M4 has two more of them and all of the cores can clock higher.


Apple doesn't bin for artificial market segmentation. Compared to other semiconductor companies, they barely bin at all. Apple has always quietly binned their SoCs based on power / leakage. Disabling cores is 100% about yields. In this case the lead products are iPads: passively cooled, ultra-thin tablets that have serious constraints when it comes to skin temperatures. Disabling a single performance core makes practically zero difference to the vast majority of end users, but probably allows them to salvage a massive number of chips that would otherwise not have met power / thermal targets when all of the cores were loaded. If they had started the M4 transition with the MacBook Pro and Mac mini, they might have been able to bin low leakage chips until they had enough to use in the iPad Pros. On the other hand, having led with the iPads, the M4s destined for future Mac minis and MacBook Airs may have all their CPU cores enabled but instead have one or two GPU cores disabled, allowing Apple to salvage yet more dies for use in products where GPU performance is less critical.


I already responded to your initial question, but I just wanted to address the bit about process improvements. The only claims that matter are the ones TSMC is making, and as I have pointed out several times in this thread now, TSMC says N3 to N3E is +5% speed at the same power or -7% power at the same speed. Random people spew nonsense on the internet all the time, but if a credible source with first-hand knowledge of TSMC's N3 processes has publicly disputed TSMC's performance claims, feel free to provide links.

GB6 of 3800+

It’s a new core or they optimized the hell out of M3 core.

Holy cow.

 

[soresu]

because if apple reach 5ghz its gonna be gard to catch up

For a lot of people this doesn't matter simply because Apple is never going to produce a Windows PC with their hardware.

 

[mikegg]

Not necessarily the most appropriate comparison since it's M3 Max (with active cooling) against M4 (with passive cooling), but here you go.

Clock speed increase 8.7%.
IPC increase overall 7.6%.
IPC increase excluding SME 3.0%.

https://twitter.com/x/status/1788364108737466609

Why did this person use M3 Max? Max always has higher ST scores than base. Need to compare to M3 Air instead which is also fanless.

Also, there are higher M4 scores than the one he used.

 

[amosliu137]

the gb5 of m4 is 2641 、
https://browser.geekbench.com/v5/cpu/22466510

 

[Eug]

the gb5 of m4 is 2641 、
https://browser.geekbench.com/v5/cpu/22466510

I compared that M4 Geekbench 5.5.1 result to this M3 MacBook Air's:

Mac15,12 - Geekbench

Benchmark results for a Mac15,12 with an Apple M3 processor.

browser.geekbench.com

That specific M3 MacBook Air's Geekbench 5.5.1 scores are better than average on their website but are not at the absolute top. IPC increase with M4 over M3 is roughly 3% single-core and 9% multi-core.

 

[Nothingness]

Thanks! Geekbench 5 made no use of either SVE (first in GB 6.1.0) or SME (first in GB 6.3.0) so this gives a better point of comparison.
So basically no improvement. I guess variations could be due to memory only. It seems the core did not have many changes.

 

[Hitman928]

I compared that M4 Geekbench 5.5.1 result to this M3 MacBook Air's:

Mac15,12 - Geekbench

Benchmark results for a Mac15,12 with an Apple M3 processor.

browser.geekbench.com

That specific M3 MacBook Air's Geekbench 5.5.1 scores are better than average on their website but are not at the absolute top. IPC increase with M4 over M3 is roughly 3% single-core and 9% multi-core.

View attachment 98646

Lines up with the GB6 results when SME is taken out, about 3% IPC improvement (maybe due to memory increase?), the rest is clock speed increase.

 

[SpudLobby]

Th

Holy crap! Up until today, you guys had me believing M4 cores would essentially be tweaked M3 cores.

I wonder if the binned 9-core variant is roughly the same multi-core CPU speed as M3 at about 12000 in GB 6.3.

We will likely have to wait until next week for those benches though. All the current benches are not surprisingly 10-core.

They are. It’s just SME added and support for SME from GB6.

I had the same reaction but when you look at integer messy subtests or GB5 which has no SME, the perf/GHz gains are just 8-9% since M1 — also known as roughly 0% since A17/M3 AKA the same 8-9% since M1 for those.

It’s a tweak for N3E design rules with higher clocks and SME instead of AMX.

 

[gdansk]

Shipping a ~10-15% faster chip 6 months after shipping another ~15% faster chip is pretty good in any case. Who knows why that happened, it couldn't have been the plan.
I know some people are disappointed it's mainly in clock rate instead of IPC but what's the difference when they claim to be keeping power draw down too?

 

[SpudLobby]

The big L from this seems to be
A) people who always said Apple can’t tweak for higher clocks due to their width and size — iso-node it’s true they use denser cells, and they ride out node gains here instead of using less area efficient stuff, but in terms of timing they’ve been able to make tweaks and make the cutoff.

I’ve heard that cope line for years now and we went from Apple at the low 1GHz to 4.4GHz. Denser nodes ease timing constraints for wide designs and allow Apple to hit higher clocks while keeping power low. They just don’t push clocks as high as AMD/Intel, but looking here they’re going to be about 80% of the way to AMD’s top Zen 5 mobile product.


Similarly, AMD and Intel have gotten wider too as usual and soon Lion Cove and Zen 5 will look much more like an Apple P core (Golden Cove was already a step that way), minus the cache stuff.


2. That Apple wouldn’t adopt SME or SVE and would stick with AMX. This is proven totally wrong now. They adopted and built AMX before Arm had their own solution, but worked with them (it’s rumored) to produce SME, and now they’ve adopted that, and probably ditched AMX.