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]

A17 Pro SoC
? GHz
CPU: 6-cores, 2 performance, 4 efficiency
10% faster performance cores, and efficiency cores also faster
New redesigned "Pro-class" GPU: 6 cores, 20% faster
GPU has mesh shading, hardware ray tracing (4X as fast as software)
Neural engine 35 trillion OPS
19 billion transistors, N3 (EUV)
? GB memory
ProRes
Pro Display engine
AV1 decoder

 

[jpiniero]

That's a bummer about the CPU. The A16 is 16B so it sounds like the additional transistors were mostly spent on the GPU and the neural engine.

They've basically stopped updating the CPU arch.

 

[gdansk]

10% isn't bad depending on the clock rate.
If it's from higher clocks then it is disappointing.

 

[smalM]

better branch prediction
wider decode and execution

As decode was already 8 wide this means complete new L1I interface and instruction fetch.

Neural engine 35 trillion OPS = half data width ?

 

[Doug S]

A17 Pro SoC
? GHz
CPU: 6-cores, 2 performance, 4 efficiency
10% faster performance cores, and efficiency cores also faster
New redesigned "Pro-class" GPU: 6 cores, 20% faster
GPU has mesh shading, hardware ray tracing (4X as fast as software)
Neural engine 35 trillion OPS
19 billion transistors, N3 (EUV)
? GB memory
ProRes
Pro Display engine
AV1 decoder

These numbers don't make sense.

There was a "leaked" GB6 result recently over 3200 - usually the late leaks are accurate. If it was only 10% it would basically be getting just the process speedup with zero IPC gain. That's possible but it would be pretty underwhelming.

But what really stands out is the claim of 20% faster GPU. Given that it goes from 5 to 6 cores it gets 20% automatically, leaving nothing for process or for the redesigned GPU cores that were supposed to come last year and were supposed to be 20-30% faster but it wasn't finished in time to make the cut. Unless they put ALL performance from the new core (with maybe another year's worth of improvements added on) and process improvements towards power reduction I don't see how they could get only the 20% they're getting from the extra core.

We'll have to wait for benchmarks I guess to clear things up.

 

[Glo.]

These numbers don't make sense.

There was a "leaked" GB6 result recently over 3200 - usually the late leaks are accurate. If it was only 10% it would basically be getting just the process speedup with zero IPC gain. That's possible but it would be pretty underwhelming.

But what really stands out is the claim of 20% faster GPU. Given that it goes from 5 to 6 cores it gets 20% automatically, leaving nothing for process or for the redesigned GPU cores that were supposed to come last year and were supposed to be 20-30% faster but it wasn't finished in time to make the cut. Unless they put ALL performance from the new core (with maybe another year's worth of improvements added on) and process improvements towards power reduction I don't see how they could get only the 20% they're getting from the extra core.

We'll have to wait for benchmarks I guess to clear things up.

That leaked score was complete and utter BS.

 

[Mopetar]

That's a bummer about the CPU. The A16 is 16B so it sounds like the additional transistors were mostly spent on the GPU and the neural engine.

They've basically stopped updating the CPU arch.

For a CPU that's primarily going into a mobile phone, I'm not sure what anyone is expecting. If anything the design for the SoC would prioritize using less power to achieve the same performance. Since the also use the cores in their computers, the other obvious design goal would be to make a core that has more clock range such that it still uses very low power in the mobile SoCs, but can clock much higher when it has the extra TDP.

I don't play any 3D games on my phone that would actually tax the GPU so I can't say how useful the performance uplifts there are going to be, but the added technology does suggest Apple is moving towards feature parity with DX12. More important for their M1 desktop chips than their phone SoC though.

Rumors from earlier this year said the M3 would be released with updated laptops at an event in October.

 

[Glo.]

Welcome to the world of diminishing progress from new, smaller nodes.

 

[Eug]

Well, one piece of good news (besides the GPU stuff) is it finally got a hardware AV1 decoder. That means M3 will also have a hardware AV1 decoder.

 

[eek2121]

These numbers don't make sense.

There was a "leaked" GB6 result recently over 3200 - usually the late leaks are accurate. If it was only 10% it would basically be getting just the process speedup with zero IPC gain. That's possible but it would be pretty underwhelming.

But what really stands out is the claim of 20% faster GPU. Given that it goes from 5 to 6 cores it gets 20% automatically, leaving nothing for process or for the redesigned GPU cores that were supposed to come last year and were supposed to be 20-30% faster but it wasn't finished in time to make the cut. Unless they put ALL performance from the new core (with maybe another year's worth of improvements added on) and process improvements towards power reduction I don't see how they could get only the 20% they're getting from the extra core.

We'll have to wait for benchmarks I guess to clear things up.

It was mentioned here on this very forum not to expect large gains moving forward. Apple is designing a chip for a closed ecosystem. Cost and power consumption are priority for them. Performance is a distant 3rd place.

Anyone expecting more than a 10-15% increase in CPU performance year over year wasn’t seeing Apple’s target market. A faster CPU is not really something iPhone users want/need (though it would be nice)

Things might be a tad more exciting with the M3.

 

[A///]

I missed the Apple event due to a doctor's appointment. Anything exciting I missed before I trawl through Apple articles?

 

[Panino Manino]

Hardly disappoing when you already had the fastest chip.

I'm more interested in that periscope camera.

 

[adroc_thurston]

They've basically stopped updating the CPU arch.

They can't, it's a skeleton crew.

Hardly disappoing when you already had the fastest chip.

The delta versus X4 will be basically nonexistent.

Cost and power consumption are priority for them

Which is why they had the most powerhungry phone cores since 2013 smothered in tons (by phone standards) cache.

Performance is a distant 3rd place.

No it was the first place ever since Swift in A6.

Anyone expecting more than a 10-15% increase in CPU performance year over year wasn’t seeing Apple’s target market

Yes everyone was expecting more than a rounding error IPC bump after 12Q of nothing.

A faster CPU is not really something iPhone users want/need (though it would be nice)

It is exactly what a phone needs.
Those jabbascreept piles aren't gonna move themselves.

 

[gdansk]

The delta versus X4 will be basically nonexistent.

They're so far ahead of the competition I don't see how that could be. GB6 of 2500 for A16. GB6 of 1850 for X3. X4 is supposed to be 15% faster than X3. So their single core lead dwindles from 35% to a bit less than 30%. Boo hoo.

It is exactly what a phone needs.
Those jabbascreept piles aren't gonna move themselves.

Apple forcing everyone to use JSC muddies the water of "CPU performance matters" anyway.

 

[adroc_thurston]

They're so far ahead of the competition I don't see how that could be.

Kinda.

GB6 of 2500 for A16. GB6 of 1850 for X3

Eh there's a clock delta that can shrink any moment.

Apple forcing everyone to use JSC muddies the water of "CPU performance matters" anyway.

It really does.

 

[igor_kavinski]

Anything exciting I missed

Nope.

Unless mesh shading and 4X faster hardware raytracing than doing it in software excites you. 10% faster CPU. 20% faster GPU. Maybe M3 MBA will be 15% faster CPU and 30% faster GPU?

Crappy thing will be, if M1, M2 and M3 laptops all continue to sell, from entry level expensive to ultra expensive.

 

[A///]

Nope.

Unless mesh shading and 4X faster hardware raytracing than doing it in software excites you. 10% faster CPU. 20% faster GPU. Maybe M3 MBA will be 15% faster CPU and 30% faster GPU?

Crappy thing will be, if M1, M2 and M3 laptops all continue to sell, from entry level expensive to ultra expensive.

I am due for a new phone. I'll get a 15 Pro. There was a rumour apple is considering a laptop for college students cheaper than an air. I suspect it will utilise the m1 processor and the air will get the m2 as standard or the m3. It's still a rumour and there's no concerete evidence about it just yet. I'll be picking up an MTL laptop in the next few months as they become available but the M3 mbp is mine to have.

 

[poke01]

I missed the Apple event due to a doctor's appointment. Anything exciting I missed before I trawl through Apple articles?

A17 has hardware based RT and AV1 decoder

 

[poke01]

Nope.

Unless mesh shading and 4X faster hardware raytracing than doing it in software excites you. 10% faster CPU. 20% faster GPU. Maybe M3 MBA will be 15% faster CPU and 30% faster GPU?

Crappy thing will be, if M1, M2 and M3 laptops all continue to sell, from entry level expensive to ultra expensive.

M3 will be mainly about GPU and Ray tracing for 3D applications.

 

[Dayman1225]

But what really stands out is the claim of 20% faster GPU. Given that it goes from 5 to 6 cores it gets 20% automatically, leaving nothing for process or for the redesigned GPU cores that were supposed to come last year and were supposed to be 20-30% faster but it wasn't finished in time to make the cut. Unless they put ALL performance from the new core (with maybe another year's worth of improvements added on) and process improvements towards power reduction I don't see how they could get only the 20% they're getting from the extra core.

We'll have to wait for benchmarks I guess to clear things up.

Yeah they highlighted pretty heavily that this GPU core was basically the biggest redesign they have ever done. Also that 20% figure was peak performance gain, they never mentioned anything about sustained performance boost, so it could end up being lower than 20% in games.

 

[A///]

A17 has hardware based RT and AV1 decoder

sold.

 

[Ajay]

M3 will be mainly about GPU and Ray tracing for 3D applications.

And larger DRAM capacities.

 

[adroc_thurston]

And larger DRAM capacities.

We're still at 24Gbit dies for LPDDR5 so no.

 

[Ajay]

We're still at 24Gbit dies for LPDDR5 so no.

For desktop, not mobile.

 

[adroc_thurston]

For desktop, not mobile.

Every single Apple chip uses piles of LPDDR.