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

 

[jdubs03]

Until Intel's foundry is a real thing able to service Apple's volumes with the reliability Apple has become accustomed to from TSMC, that's the only realistic backup plan.

Oh I agree. If it were to come to that, it would be even more encouraging for Apple (and by extension others) to focus even more so on design improvements, such that there would still be some performance improvement. And in that case, what would likely happen is that the cutting edge IP would wind up making its way to Arizona. Not like it would be useful in a burning Taiwan.
Hopefully all this comes to pass though, which I think it will. I’m skeptical that China would make such a foolhardy move.

Irrespective of all that, Intel still needs to impress potential customers to get volume on their upcoming process nodes. And enough to further expand. I think they have a chance. But I’m not getting ahead of myself with optimism.

 

[mvprod123]

There are still no M4 Pro/Max die shots 😕

 

[Doug S]

Oh I agree. If it were to come to that, it would be even more encouraging for Apple (and by extension others) to focus even more so on design improvements, such that there would still be some performance improvement. And in that case, what would likely happen is that the cutting edge IP would wind up making its way to Arizona. Not like it would be useful in a burning Taiwan.

Well I assume if something happens with Taiwan that shuts down TSMC's production there (China invades or a 9.0 earthquake) it will happen far more quickly than the timescale where Apple could redesign anything. If that happened today and they lost access to N3 family nodes they'd have to re-use A16 for iPhone 17. There simply wouldn't be time to iterate that design in any meaningful way - maybe they could tweak the memory controllers to support LPDDR5X but there wouldn't be nearly enough time to try to e.g. port A18P's cores to it.

Even if China did like Russia where they started massing troops on Ukraine's border 3-4 months before they crossed so TSMC knew what was coming, I doubt that would be enough time to pack up and ship EUV scanners to the US and their other locations. Best they could do would be to redirect equipment orders destined for Taiwan, and speed up their timetable for new fab builds in the US and elsewhere.

 

[jdubs03]

Well I assume if something happens with Taiwan that shuts down TSMC's production there (China invades or a 9.0 earthquake) it will happen far more quickly than the timescale where Apple could redesign anything. If that happened today and they lost access to N3 family nodes they'd have to re-use A16 for iPhone 17. There simply wouldn't be time to iterate that design in any meaningful way - maybe they could tweak the memory controllers to support LPDDR5X but there wouldn't be nearly enough time to try to e.g. port A18P's cores to it.

Even if China did like Russia where they started massing troops on Ukraine's border 3-4 months before they crossed so TSMC knew what was coming, I doubt that would be enough time to pack up and ship EUV scanners to the US and their other locations. Best they could do would be to redirect equipment orders destined for Taiwan, and speed up their timetable for new fab builds in the US and elsewhere.

Everyone would be hurting bad in the smartphone space.

Perhaps there would be some large scale rescue mission with cargo planes getting those EUV scanners out. If given enough heads up there should be enough time to evacuate the most important bits.

 

[Doug S]

Everyone would be hurting bad in the smartphone space.

Perhaps there would be some large scale rescue mission with cargo planes getting those EUV scanners out. If given enough heads up there should be enough time to evacuate the most important bits.

Somewhere on the internet there are some videos showing the packing, shipping and unpacking process for those things. A few months would not be enough lead time to make that happen. Maybe you could do a few, but there are only so many ASML employees who know how to do that - and they have probably never broken down and packed up one that's already operational to relocate it. It might not even be possible.

 

[johnsonwax]

Everyone would be hurting bad in the smartphone space.

Perhaps there would be some large scale rescue mission with cargo planes getting those EUV scanners out. If given enough heads up there should be enough time to evacuate the most important bits.

Yeah. Though, we saw a bit of what that might look like after the Japan tsunami. The day of the tsunami Apple packed up executives with cash to fly into Japan and pay suppliers to help their employees recover and reopen. That moved them to the front of the line as a customer and they saw relatively little disruption. In a global crisis, I wouldn't bet against Apple.

And while it can take months to set that equipment up that's because the stakes are low. If the Chinese navy is bearing down, I'm willing to bet they'd figure out how to get them out quickly, even if there was a risk of damage.

 

[FlameTail]

Rumour:

Exclusive:

Apple M5 will be mass-produced by the end of 2025,
Apple A20 will be mass-produced in the second quarter of 2026.

Source: TrendForce

Both products are based on TSMC’s 2nm process.

Jukanlosreve

If M5 is mass produced at the end of 2025, that means no M5 products in 2025.

 

[poke01]

Rumour:


Jukanlosreve

If M5 is mass produced at the end of 2025, that means no M5 products in 2025.

If M5 is based on N2 but comes out in 2026 that’s not bad.

 

[FlameTail]

If M5 is based on N2 but comes out in 2026 that’s not bad.

Seems like it will have a definitive advantage over Snapdragon X Elite Gen 2, which will presumably use TSMC N3P.

 

[Eug]

Apple throttles the CPU before the fan ramps up. However, if you manually increase fan speed, then CPU performance increases.





Power usage and clock speed



The effect of hardware ray tracing

 

[digitaldreamer]

Apple throttles the CPU before the fan ramps up. However, if you manually increase fan speed, then CPU performance increases.

View attachment 112231

View attachment 112229

Power usage and clock speed

View attachment 112230

The effect of hardware ray tracing

View attachment 112232

View attachment 112233

Wow. That 20c M4 beats the 60c M2 in Cinebench 2024 GPU.

 

[igor_kavinski]

Wow. That 20c M4 beats the 60c M2 in Cinebench 2024 GPU.

It's nothing to do with 20C vs. 60C. GPU performance advancements increase at a much faster rate than CPU so kinda expected.

 

[digitaldreamer]

It's nothing to do with 20C vs. 60C. GPU performance advancements increase at a much faster rate than CPU so kinda expected.

But, more than 3 times improvement in 2 generations?

 

[igor_kavinski]

But, more than 3 times improvement in 2 generations?

Probably removed some hardware bottleneck that is preventing the M2 GPU from properly flexing its muscles.

 

[Eug]

Wow. That 20c M4 beats the 60c M2 in Cinebench 2024 GPU.

Probably removed some hardware bottleneck that is preventing the M2 GPU from properly flexing its muscles.

That's why I mentioned the ray tracing. The M2 Ultra does not have hardware ray tracing or hardware mesh shading.

 

[mvprod123]

Probably removed some hardware bottleneck that is preventing the M2 GPU from properly flexing its muscles.

The new shader core, dynamic caching and RT.

 

[The Hardcard]

Rumour:


Jukanlosreve

If M5 is mass produced at the end of 2025, that means no M5 products in 2025.

The rumors continue to line up for the M5. It’s why I still believe there won’t be an M4 Ultra.

The evidence is growing that M5 will be a major step for Apple, and I think they are willing to wait for N2 and the next generation SOIC based interconnect. Both are targeted for 2H 2025.

One, M5 Ultra is important to get right for their Private Cloud Compute. The M5 will go into the second generation Vision Pro. Both of these would benefit from a major step forward in GPU compute power and efficiency.

It appears it will be a long wait for the next Studios, but Apple does have a history of letting their desktops sit for years waiting for updates. I wouldn’t be surprised if that was one incentive to the massive increase in compute for the M4 Pro.

 

[poke01]

毒蘑菇性能测试 - volumeshader_bm - 工具哇在线工具集

可以用来测试手机和电脑的性能，如果设备配置比较差的话，会容易卡死

toolwa.com

Can anyone with an M4 Mac run this test on chrome? and post the recorded video here. You can interact with the ball as well

 

[Eug]

毒蘑菇性能测试 - volumeshader_bm - 工具哇在线工具集

可以用来测试手机和电脑的性能，如果设备配置比较差的话，会容易卡死

toolwa.com

Can anyone with an M4 Mac run this test on chrome? and post the recorded video here. You can interact with the ball as well

This is https://toolwa.com/vsbm/ running in Chrome 131 on the M4 Mac mini with 24 GB RAM, macOS Sequoia 15.1.1.

Note that the screen is a 3840x2560 screen set at 2304x1536 hiDPI (so it's rendering at 4608x3072). However, the browser was running as a window, not full screen.

With this webpage running, desktop navigation is extremely laggy, like a slideshow. Also, interaction with the model is also extremely laggy, with a few seconds of lag after each initiated action.

 

[poke01]

This is https://toolwa.com/vsbm/ running in Chrome 131 on the M4 Mac mini with 24 GB RAM, macOS Sequoia 15.1.1.

Note that the screen is a 3840x2560 screen set at 2304x1536 hiDPI (so it's rendering at 4608x3072). However, the browser was running as a window, not full screen.

With this webpage running, desktop navigation is extremely laggy, like a slideshow. Also, interaction with the model is also extremely laggy, with a few seconds of lag after each initiated action.

thank you.

This is how performs on the 4070 super, very intensive.

 

[Eug]

thank you.

This is how performs on the 4070 super, very intensive.

Haha. Huge difference. We need to see how the M4 Max performs.

 

[FlameTail]

M2 brought a peak power reduction compared to M1!

Avalanche was awesome.

 

[name99]

View attachment 112325
M2 brought a peak power reduction compared to M1!

Avalanche was awesome.

Remember what was said at the time:

Apple M2 SoC Analysis - Worse CPU efficiency compared to the M1

The Apple M2 is here and after our reviews of the MacBook Pro 13 and the MacBook Air M2, we want to have a closer look at the new M2 chip. What changed, which improvements can you expect, and how does the increased power consumption affect the efficiency, especially compared to the previous M1 chip?

www.notebookcheck.net

?


The tech media is mostly clowns. Even Geekerwan is WRONG in how he is presenting this. The correct metric if this is the type of graph you want is performance per joule, ie performance against energy NOT power.
If you double performance and use the same energy, by his graph you are using twice the power! Same energy, double performance - and he would score that as a DOWNGRADE...

There is one metric you should be using to measure this sort of thing, the energy-delay product, which tracks what you want -- less time AND less energy. Naturally no-one in the popular tech press has even heard of it...

 

[igor_kavinski]

Naturally no-one in the popular tech press has even heard of it...

Because they aren't engineers and both Anand and Dr. Ian are no longer in the reviews game.

 

[FlameTail]

Rumour: Apple M5 will be on TSMC 3nm

Apple Orders M5 Chips From TSMC Ahead of Late 2025 Production

Apple has ordered M5 chips from TSMC as the company begins production development of its next-generation processor for future devices, reports...

www.macrumors.com