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

 

[poke01]

https://twitter.com/x/status/1704131991430738040

 

[uzzi38]

Battery life reduction too.

Talk about a major yikes.

 

[gdansk]

Underwhelming. But need testing to be replicated elsewhere.

 

[Henry swagger]

https://twitter.com/x/status/1704134470776078799

Apple a17 has 9 wide decode.. 50% wider than intel and 2.2x wider than zen4

 

[poke01]

Battery life reduction too.

Talk about a major yikes.

View attachment 85949

Shows how bad 3N is

 

[uzzi38]

Shows how bad 3N is

If N3 was that bad they wouldn't be pushing clocks this hard, would have made sense to keep them at a similar level and reduce power instead if possible.

This is seeming a lot more like a skill issue on Apple's part.

 

[Roger Wilco]

Seems like Apple has hit the wall.

 

[FlameTail]

Woah Geekerwan just released the review of the A17 Pro on their Chinese channel

You can watch it with English subtitles.

This video is so good. They are even running SPEC2017 now. This is crazy. Finally we have a succesor to Andrei's legendary CPU reviews at Anandtech.

Will someone make a new thread on this video?

 

[Mopetar]

If N3 was that bad they wouldn't be pushing clocks this hard, would have made sense to keep them at a similar level and reduce power instead if possible.

This is seeming a lot more like a skill issue on Apple's part.

Based on the results posted, they'd still perform worse in terms of power use even with the lower clocks. They may have been banking on the larger battery compensating for the added power draw, but it seems like they've missed even their most pessimistic targets. Better to have more performance as opposed to the same or lower performance at worse power.

Without having more knowledge it's hard to portion blame between TSMC and Apple fairly. If the node is bad enough than not even Apple at the height of their capabilities would be able to do anything about a bad node. By the time the realized there was going to be trouble, there wasn't a lot of room to maneuver.

 

[Doug S]

If N3 was that bad they wouldn't be pushing clocks this hard, would have made sense to keep them at a similar level and reduce power instead if possible.

This is seeming a lot more like a skill issue on Apple's part.

If the rumors are true that Apple will be switching to N3E as soon as that process is available this makes sense.

When the A9 in the iPhone 6S was made by both TSMC and Samsung, both had the same clock even though the TSMC A9s were more efficient - running cooler and exhibiting better battery life.

So maybe this time around A17 Pros will be made in N3B and N3E, again clocked the same, and the N3E version is the one you want for better efficiency. The only difference is that last time they were sold together so you could "sort of" choose (if you cared enough you could return a Samsung and try again until you got a TSMC) but this time to get the "better" one you'd have to wait until sometime next spring when iPhone Pros start getting N3Es.

There's also the "known good die" thing due to N3B's poor yields, which we don't have any information on as far as what dies qualify. Supposedly yield is in the 55-60% range, which is pretty terrible, and we don't know what sacrifices may have been made to reach even that yield. It may be a situation where dies that are judged "good" have a much larger range of power requirement to pass all the tests at the target frequency than would normally be the case.

The fact he's able to run these tests now, before iPhone 15s have even shipped, means he somehow got his hands on an early one. How early, I wonder? It would be interesting to run power tests on several iPhone 15 Pros arriving 'normally' on this Friday, and maybe check again in December. I suggest testing several to see if the variation I'm suggesting exists and it is a crapshoot whether you get a "good" A17 Pro like the A9 days except you can only tell the difference by testing power draw. I suggest testing again in a few months to see if N3B is improving at all over time.

 

[Ajay]

Seems like Apple has hit the wall.

Lots of talk over the past five years or so of companies 'poaching' engineers, sometimes in bulk. Other have either started startups or been lured away to them. Seems like it's always a problem when a company builds a highly successful engineering team - at the height of their success they get ripped apart. I would think Apple would have enough $$s to offset this a bit, but certainly some engineers are bored and need a new project or want to have more control/impact.

 

[Mopetar]

Lots of talk over the past five years or so of companies 'poaching' engineers, sometimes in bulk. Other have either started startups or been lured away to them. Seems like it's always a problem when a company builds a highly successful engineering team - at the height of their success they get ripped apart. I would think Apple would have enough $$s to offset this a bit, but certainly some engineers are bored and need a new project or want to have more control/impact.

With Apple the all-stars were allegedly the ones that went off and started their own company. The one that Qualcomm spent billions of dollars to acquire. When you think you can get that kind of payday it's hard not to leave even if you get offered a fairly substantial raise.

It's not uncommon for top talent to move around a lot either. Look at someone like Jim Keller. He's been with a lot of companies and even had worked at AMD prior to returning to develop Zen with them.

If Apple is really hurting for some of those key individuals they certainly can afford to sway them back. I don't know how much stake those engineers had in the company when it was acquired, but there are some who just enjoy interesting work.

 

[Ajay]

Battery life reduction too.

Talk about a major yikes.

View attachment 85949

Hmm, wish I understood what benchmark they ran. Mrs. and I both upgraded to the iPhone 15 (15+ in my case). Our carrier was offering a total of $1300 for the upgrade of both older phones (as opposed to a norm of around $4-500). So, kind of a no brainer for us. I'm sure I'm going to like my battery life! Still, no fancy smancy A17 Pro SoC >shrug<

 

[gdansk]

Lots of talk over the past five years or so of companies 'poaching' engineers, sometimes in bulk. Other have either started startups or been lured away to them. Seems like it's always a problem when a company builds a highly successful engineering team - at the height of their success they get ripped apart. I would think Apple would have enough $$s to offset this a bit, but certainly some engineers are bored and need a new project or want to have more control/impact.

Every Silicon Valley company in history has been unable to offer enough to keep engineers from running off to form or work for startups creating interesting products and offering equity.

'Poaching' employees is such a slaveowner term. Even in quotes it disgusts me.

 

[soresu]

Every Silicon Valley company in history has been unable to offer enough to keep engineers from running off to form or work for startups creating interesting products and offering equity.

'Poaching' employees is such a slaveowner term. Even in quotes it disgusts me.

IMHO sticking around is only for corporate ladder climbers at the end of the day.

 

[Buttercream]

Hmm, wish I understood what benchmark they ran. Mrs. and I both upgraded to the iPhone 15 (15+ in my case). Our carrier was offering a total of $1300 for the upgrade of both older phones (as opposed to a norm of around $4-500). So, kind of a no brainer for us. I'm sure I'm going to like my battery life! Still, no fancy smancy A17 Pro SoC >shrug<

It's a loop of:

• WeChat Text Messaging (5 mins)
• WeChat Voice Messaging (5 mins)
• WeChat Video Call (5 mins)
• IT Home (10 mins)
• TikTok (10 mins)
• BiliBili (10 mins)
• Geshin Impact (10 mins)

 

[roger_k]

Woah Geekerwan just released the review of the A17 Pro on their Chinese channel

You can watch it with English subtitles.

This video is so good. They are even running SPEC2017 now. This is crazy. Finally we have a succesor to Andrei's legendary CPU reviews at Anandtech.

Will someone make a new thread on this video?

Something I find confusing is that the power measurements they report for A-series is wildly different from what Anandtech reported. Andrei measured over 4 watts in SPEC for both A14 and A15 (as well for earlier A-series). Geekerwan measured 2.5 watts. In fact, Geekerwan's power draw measurements for A17 Pro are lower than what Andrei reported for A14.

Also, a number of other early sources also report better battery life for 15 pro/max than earlier models (although some of them, like Tom's are highly dubious as they didn't do the tests under equal conditions).

Util these methodological issues are addressed and explained, I think it's a bit too early to render judgement.

 

[Doug S]

Something I find confusing is that the power measurements they report for A-series is wildly different from what Anandtech reported. Andrei measured over 4 watts in SPEC for both A14 and A15 (as well for earlier A-series). Geekerwan measured 2.5 watts. In fact, Geekerwan's power draw measurements for A17 Pro are lower than what Andrei reported for A14.

Also, a number of other early sources also report better battery life for 15 pro/max than earlier models (although some of them, like Tom's are highly dubious as they didn't do the tests under equal conditions).

Util these methodological issues are addressed and explained, I think it's a bit too early to render judgement.

It doesn't matter if his measurements are different than Andrei's, it only matters that he was consistent in how he did those measurements across A14 -> A17 so the generational comparisons are valid.

I still think it is possible there might be some variation in power consumption between parts in a low yielding process like N3B, so I agree we should reserve judgment on that until we get more data. Also, tests that aren't primarily CPU/GPU related (i.e. depending more on display, wifi/5G radio, etc. might be different since it is only the SoC made on N3B. Other parts may have had a shrink to a better process but not THAT process, the display may more efficient, etc.

 

[Henry swagger]

https://twitter.com/x/status/1704413603170824507

A17pro gpu is verry efficient

 

[roger_k]

It doesn't matter if his measurements are different than Andrei's, it only matters that he was consistent in how he did those measurements across A14 -> A17 so the generational comparisons are valid.

I still think it is possible there might be some variation in power consumption between parts in a low yielding process like N3B, so I agree we should reserve judgment on that until we get more data. Also, tests that aren't primarily CPU/GPU related (i.e. depending more on display, wifi/5G radio, etc. might be different since it is only the SoC made on N3B. Other parts may have had a shrink to a better process but not THAT process, the display may more efficient, etc.

I now watched the rest of the video and he also provides a different set of measurements (google translator gives me "mainboard power consumptions" for the Chinese label), which are much closer to Andrei's measurements. And if you look at those it paints a slightly different picture. The peak power consumption is still higher, but what's interesting is that perf/watt is more or less in line with previous models. It seems to me that this new design is more about unlocking higher performance for future desktop applications (where the thermal headroom is pretty much underused). In other words, instead of pursuing increased performance at the same power consumption, they might have went for increasing the peak performance (while preserving the overall power curve).

Which doesn't mean that A17 CPU is not efficient. According to Geekerwan's video, in MC it's more efficient — by a healthy margin — than any other tested SoC.

 

[JoeRambo]

A17pro gpu is verry efficient

I would really love for Apple to make a bold move and produce a "premium" steam deck style console. With whole emulation on Linux thing going and Apple's efforts i think we are not far from revolution in pseudo crossplatform gaming.

 

[eek2121]

I would really love for Apple to make a bold move and produce a "premium" steam deck style console. With whole emulation on Linux thing going and Apple's efforts i think we are not far from revolution in pseudo crossplatform gaming.

They already have a console. It is called the Apple TV and the iPhone for mobile gaming.

You can pair Xbox controllers to iOS devices. Some games actually require a controller.

More Apple + gaming news is coming soon-ish. Apple is supposedly planning on a gaming push at some point in the future. What that entails, I don’t know. Could just be games for the Vision Pro or it could be more. We will have to wait and see.

 

[Nothingness]

I still wonder if A17 is the first Apple Silicon to support ARMv9. Has this been seriously investigated or is too early?

 

[Doug S]

I now watched the rest of the video and he also provides a different set of measurements (google translator gives me "mainboard power consumptions" for the Chinese label), which are much closer to Andrei's measurements. And if you look at those it paints a slightly different picture. The peak power consumption is still higher, but what's interesting is that perf/watt is more or less in line with previous models. It seems to me that this new design is more about unlocking higher performance for future desktop applications (where the thermal headroom is pretty much underused). In other words, instead of pursuing increased performance at the same power consumption, they might have went for increasing the peak performance (while preserving the overall power curve).

Which doesn't mean that A17 CPU is not efficient. According to Geekerwan's video, in MC it's more efficient — by a healthy margin — than any other tested SoC.

That doesn't change that he said he measured a higher temperature on an iPhone 15 Pro than for "any previous iPhone". That's a test independent of how you measure power.

I suppose the change from aluminum to titanium frame could have influenced that though. Aluminum has >10x better thermal conductivity, making it harder for the heat to be dissipated through the entire phone rather than remaining in "hot spots".

 

[Doug S]

One thing I haven't seen discussion of that probably deserves it is the name. Apple specifically chose to name it "A17 Pro".

That kind of implies to me that the non-Pro iPhone 16s will not get this SoC next year. If not, what will they get? Will it be as simple as one less GPU core (which they have binned on already) or will there be larger differences between A17 Pro and next year's "A17 Fusion" or whatever it may be called in the iPhone 16? Is it possible it will be a separate SoC design, and if so what might they differentiate on?