Discussion Apple Silicon SoC thread

Eug · Nov 10, 2020

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

mikegg · May 6, 2024

Doug S said:
I think that's an outmoded way of looking at things. Back when chip architects were designing a "CPU", having two or three teams leapfrogging each other and when one finishes the 2024 design starting on 2026 or 2027 made sense. That's no longer true today when a chip design includes two different CPU core types, a GPU, an NPU, an ISP, and all the "uncore" like SLC/LPDDR5, USB and so forth.

I would guess they are organized along those lines, so perhaps you have one team doing CPU (or maybe even one doing P core and one doing E core) and another perhaps doing GPU/NPU/ISP, and another doing the uncore/everything else. There would be a team doing the overall A SoC and M SoC designs but they're working at a higher level integrating everything into a whole, and because they're working at a higher level they could work on multiple generations at once without leapfrogging required.

One team cannot be designing 2-3 generations of CPU at the same time. Hence, I said “some kind of system”.

An Apple exec said they plan chips about 4 years in advance. Some team has to be working on the chip set to release in 2028 right now.

poke01 · May 6, 2024

Apple Reportedly Building M2 Ultra and M4-Powered AI Servers

Apple has started building its own AI servers that use the M2 Ultra chip, Haitong analyst Jeff Pu reports. Pu provided the analysis based on supply...

www.macrumors.com

This time a more reputable analysis and supply chain Pu is saying Apple is building M2 Ultrs and later M4 (M4 Ultra probably) servers for AI

Eug · May 6, 2024

poke01 said:
This time a more reputable analysis and supply chain Pu is saying Apple is building M2 Ultrs and later M4 (M4 Ultra probably) servers for AI

Interesting:

1. M2 Ultra
2. Servers - For use themselves, or for retail/institutional sale?
3. "supply chain Pu"

poke01 · May 6, 2024

Eug said:
Interesting:

1. M2 Ultra
2. Servers - For use themselves, or for retail/institutional sale?
3. "supply chain Pu"

I think the AI Server is for themselves, instead of using other companies CPUs/GPUs why not use your own.

Eug · May 6, 2024

poke01 said:
I think the AI Server is for themselves, instead of using other companies CPUs/GPUs why not use your own.

I had read that Apple uses Linux for its iCloud backend.

And when Steve Jobs was running Pixar, the render farms were also Linux (on Xeons). I'm not sure about now, but I'm thinking it is still Linux.

poke01 · May 6, 2024

Eug said:
I had read that Apple uses Linux for its iCloud backend.

And when Steve Jobs was running Pixar, the render farms were also Linux. I'm not sure about now, but I'm thinking it is still Linux.

Linux I think so too. It’s simple and easy to configure. It’s open source, so no fees.

Ajay · May 6, 2024

poke01 said:
I think the AI Server is for themselves, instead of using other companies CPUs/GPUs why not use your own.

That would be odd. x86 servers are likely cheaper and they have at least 8 datacenter locations globally and have invested some $XXB in them.

oak8292 · May 6, 2024

Ajay said:
That would be odd. x86 servers are likely cheaper and they have at least 8 datacenter locations globally and have invested some $XXB in them.

The margins for X86 CPUs and GPUs are fairly clear with both AMD and Nvidia using wafers from TSMC and reporting on gross margins. Intel’s transistor costs have just be broken out into a foundry line item to clarify their costs. Apple is buying enough wafers to get preferred status either at TSMC or Intel. Apple has CPU, GPU and NPU designs with their own software stack. They can license ARM cores for servers just like AWS or Nvidia if they don’t feel their own core is adequate. I wouldn’t second guess them. They know the economics and they have some advantages.

With Google, Microsoft, AWS and now rumors of Apple all designing and building ARM server processors it seems that X86 IP is over priced.

Eug · May 6, 2024

oak8292 said:
The margins for X86 CPUs and GPUs are fairly clear with both AMD and Nvidia using wafers from TSMC and reporting on gross margins. Intel’s transistor costs have just be broken out into a foundry line item to clarify their costs. Apple is buying enough wafers to get preferred status either at TSMC or Intel. Apple has CPU, GPU and NPU designs with their own software stack. They can license ARM cores for servers just like AWS or Nvidia if they don’t feel their own core is adequate. I wouldn’t second guess them. They know the economics and they have some advantages.

With Google, Microsoft, AWS and now rumors of Apple all designing and building ARM server processors it seems that X86 IP is over priced.

It's not really about second-guessing Apple, but about second guessing the analyst Jeff Pu. Do you believe his rumour?

poke01 · May 6, 2024

Eug said:
It's not really about second-guessing Apple, but about second guessing the analyst Jeff Pu. Do you believe his rumour?

Jeff is much more accurate than Kuo. In fact, he reported the iPhone 17 slim model screen size before Ross Young did. So he has some merit.

Eug · May 6, 2024

poke01 said:
Jeff is much more accurate than Kuo. In fact, he reported the iPhone 17 slim model screen size before Ross Young did. So he has some merit.

Jeff Pu isn't more accurate than Kuo, at least historically speaking. He actually has a pretty poor track record. Maybe he's better in recent months than he was in the past, I'm not sure, but in previous years he was worse than the much criticized DigiTimes.

oak8292 · May 7, 2024

Eug said:
It's not really about second-guessing Apple, but about second guessing the analyst Jeff Pu. Do you believe his rumour?

Not really. I think it is a good educated guess but I am not big on rumors. Apple has the need and a cost advantage so it makes sense that Apple would roll their own. Will it be by 2025, who knows? Educated guessing.

Doug S · May 7, 2024

poke01 said:
Apple Reportedly Building M2 Ultra and M4-Powered AI Servers

Apple has started building its own AI servers that use the M2 Ultra chip, Haitong analyst Jeff Pu reports. Pu provided the analysis based on supply...

www.macrumors.com

This time a more reputable analysis and supply chain Pu is saying Apple is building M2 Ultrs and later M4 (M4 Ultra probably) servers for AI

I still don't buy it. I saw other reports that Apple was building custom chips for their AI servers. That I could believe. It makes zero sense to install a bunch of Apple Silicon chips just to use the tiny corner that's labeled 'NPU', or the slightly larger piece that's labeled 'GPU', and waste a bunch of space on excess CPU cores, USB/TB controllers, display controllers and the like. They'd built chips that all AI compute with a little corner with a few CPU cores for management tasks. They'd drop the LPDDR5 and use HBM. To sum up, if they were going to do this they would do it right, not half ass it by trying to leverage one of their existing chips.

Doug S · May 7, 2024

Regarding the Gurman article on using M4, I wonder if his source was the iPadOS 14.5 betas containing indications of iPad 16,x and the T8132. That information was apparently around for a while before Gurman's article, so maybe he doesn't have any actual sources for this beyond that. Others might have put the same 2+2 together, but he's more widely published so he'll get the "credit" if we see M4 iPads unveiled tomorrow.

Are they going to ding his 86.5% "prediction accuracy" when it turns out his 18 month cadence thing was bs? Or will they give him credit, since he was kinda/sorta right based on M1->M2 and M2->M3, if you squint a bit and count the months in the way most favorable to that conclusion?

Eug · May 7, 2024

Doug S said:
Regarding the Gurman article on using M4, I wonder if his source was the iPadOS 14.5 betas containing indications of iPad 16,x and the T8132. That information was apparently around for a while before Gurman's article, so maybe he doesn't have any actual sources for this beyond that. Others might have put the same 2+2 together, but he's more widely published so he'll get the "credit" if we see M4 iPads unveiled tomorrow.

Are they going to ding his 86.5% "prediction accuracy" when it turns out his 18 month cadence thing was bs? Or will they give him credit, since he was kinda/sorta right based on M1->M2 and M2->M3, if you squint a bit and count the months in the way most favorable to that conclusion?

I discussed this earlier in the thread, but yes it would seem that Gurman has recently been reading the same forums and has joined the M4 prediction bandwagon based on that. From his newsletter, it doesn’t sound like he has any additional info on this.

Mopetar · May 7, 2024

Less than half an hour. Get your last minute wild predictions in here.

Anyone want to skip the M4 and go straight to ~~five blades~~ M5?

Tigerick · May 7, 2024

I still think upcoming iPad Pro will use M3 only, not M4. Let's see..

SteinFG · May 7, 2024

Mopetar said:
Less than half an hour. Get your last minute wild predictions in here.

Anyone want to skip the M4 and go straight to ~~five blades~~ M5?

To what? WWDC is still a month away. Maybe i'm out of the loop
edit: separate apple event on may 7, now I know it

IEC · May 7, 2024

Watching live stream now. They just finished on the Air segment:

iPad Air with M2 chips, base model now 128GB storage up to 1TB at high end
iPad Air 11" $599
iPad Air 13" $799

iPad Pro segment next.

Eug · May 7, 2024

iPad Pro has Tandem OLED at 1000 nits and 1600 nits peak.
Called Ultra Retina XDR, for both 11" and 13" models.
Nanotexture

SoC is M4. You guys are right using the identifiers to predict that.

IEC · May 7, 2024

iPad Pro in 11" and 13", silver and space black only

11" Pro = 5.3mm thin, 0.98lb
13" Pro = 5.1mm thin, 1.28lb

Display:
OLED for first time
Uses TWO panels to have enough brightness for 1000 nits, tandem OLED, peak 1600 nits

Mopetar · May 7, 2024

iPad Pro now comes in diet.

Mopetar · May 7, 2024

M4!

4x Performance cores, 6x Efficiency cores, 10 core GPU. New display engine. Uses half the power for same performance as M2.

IEC · May 7, 2024

iPad Pro will use new M4 chip

2nd gen 3nm tech
New display engine, 10-120Hz variable refresh
4 performance cores, 6 efficiency (up to 50% faster than M2 in the previous iPad Pro claimed)
10-core GPU, mesh shading, ray tracing (claimed 4x performance vs M2)
M4 delivers same performance as M2 at 1/2 the power

Eug · May 7, 2024

M4
Up to 10-core CPU - 4P/6E - 50% faster CPU than M2
10-core GPU
Hardware raytracing
NPU is fastest in any AI PC - 38 trillion ops/s
28 billion transistors
120 GB/s unified memory bandwidth
2nd gen 3nm (which likely means N3E)

Discussion Apple Silicon SoC thread

Lifer

Golden Member

Senior member

Lifer

Senior member

Lifer

Senior member

Lifer

Member

Lifer

Senior member

Lifer

Member

Platinum Member

Platinum Member

Lifer

Diamond Member

Senior member

Senior member

Elite Member

Lifer

Elite Member

Diamond Member

Diamond Member

Elite Member

Lifer