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]

Re: 65 Watts:

In Cinebench CPU, the M2 Ultra (24-core CPU and 60-core GPU) will peak at a little over 70 watts for CPU.

For the extended 10 minute Cinebench CPU test, the fans remained at baseline (which is silent), and temps did not exceed 85C on the hottest core.

 

[Mopetar]

In Cinebench CPU, the M2 Ultra (24-core CPU and 60-core GPU) will peak at a little over 70 watts for CPU.

Performance that will warm your heart, but not your room.

 

[soresu]

Performance that will warm your heart, but not your room.

85C might not warm your room but I have my doubts about the cooling system workign properly if it insists on staying silent.

I'd rather have my ultra expensive PC brick sound like a jumbo jet than decrease the lifespan of the chip simply to be quiet.

 

[Doug S]

85C might not warm your room but I have my doubts about the cooling system workign properly if it insists on staying silent.

I'd rather have my ultra expensive PC brick sound like a jumbo jet than decrease the lifespan of the chip simply to be quiet.

85C is not a problem for modern chips, and it isn't like Apple doesn't work closely with TSMC who I'm sure does plenty of testing for stuff like this long before any chips are shipped to customers.

If I knew for a fact keeping my PC silent would cause the CPU to fail the day it turned five years old I'd gladly make that choice.

 

[Eug]

85C might not warm your room but I have my doubts about the cooling system workign properly if it insists on staying silent.

I'd rather have my ultra expensive PC brick sound like a jumbo jet than decrease the lifespan of the chip simply to be quiet.

Content creation is a big market for the Mac Studio, and the computer sits on the desk. A big complaint for the M1 Max/Ultra Mac Studio was noise. It wasn't really fan noise related to speed but fan whine. Regardless, it was enough for some musicians and other users to return their Mac Studios. Apple has addressed this in the M2 Max/Ultra Mac Studio by altering the fan ramp up curve, but appropriate for the context of heat tolerances of those chips. I wonder also if they got a new fan supplier too.

The fan WILL ramp up, but only if absolutely necessary. No real point in trying to keep the chip at 65 C or lower. I think 85 C is just fine. Sure it might mean the Mac might only last 10 years instead of 15, but really, who cares? At least it's not 100 C like the previous Intel chips.

BTW, back in the day, I returned a Core i7 iMac because of fan noise. Anything I did that maxed out the CPU would ramp up the fan within 30 seconds. It was incredibly irritating to have a vacuum cleaner on my desk just exporting a simple 1 minute birthday video. So, I returned it and got a slower i5 iMac instead. Much happier. It was about 25% slower, but when the CPU was maxed out, it took ~9 minutes to ramp up the fan to max. That meant that for all shorter CPU-intensive tasks, the computer remained silent. Good tradeoff IMO. One could say the i5 was 18X as good as the i7 for fan noise performance.

 

[soresu]

I wonder also if they got a new fan supplier too

This definitely makes a difference.

Everyone I talk to seems to find Noctua fans fine, but to me they sound unbearable at even middling RPMs.

Unfortunately I'm much too lazy to open up my PC again and install a bequiet fan on the Noctua CPU heatsink so I'm just running my 3950X at much lower clocks than it can ideally manage with a bearable fan noise😅

 

[soresu]

Sure it might mean the Mac might only last 10 years instead of 15

10 years ago maybe - but indications seems to be that chips made on newer fab nodes at smaller pitches fail sooner than the old ones for the same punishment.

The system is basically embedded given that tight space with minimal passive airflow to supplement the cooling when fans are not running, so I strongly doubt that 10 years is a likely lifespan given that operating temperature if it is being used as a weekly workhorse at anything close to the max of that 24C spec which is completely reasonable in the content creation space.

I guess if you can afford yearly full Maya subscriptions and all the other super expensive DCC software it wouldn't bother you much to replace one of these every 5 years, but it's still stupid none the less unless you can actually alter the fan curve yourself.

 

[Eug]

10 years ago maybe - but indications seems to be that chips made on newer fab nodes at smaller pitches fail sooner than the old ones for the same punishment.

The system is basically embedded given that tight space with minimal passive airflow to supplement the cooling when fans are not running, so I strongly doubt that 10 years is a likely lifespan given that operating temperature if it is being used as a weekly workhorse at anything close to the max of that 24C spec which is completely reasonable in the content creation space.

I guess if you can afford yearly full Maya subscriptions and all the other super expensive DCC software it wouldn't bother you much to replace one of these every 5 years, but it's still stupid none the less unless you can actually alter the fan curve yourself.

I'm not sure if you know how the Mac Studio is designed. It’s got a massive heat sink (aluminum for M2 Max and copper for M2 Ultra), and the fans (two of them) are always running. You cannot turn them off. I believe it’s 1100 rpm minimum for the M2 series.

And yes you can alter the fan speeds yourself if you want, with a third party app, but as mentioned, you can't turn the fans off.

The big black bar at the top of the first pic is the heatsink, with the two fans blowing through it.





Basically, nearly half of the interior volume of the Mac Studio is dedicated to cooling.

 

[Mopetar]

85C might not warm your room but I have my doubts about the cooling system workign properly if it insists on staying silent.

I'd rather have my ultra expensive PC brick sound like a jumbo jet than decrease the lifespan of the chip simply to be quiet.

The temperature that the chip reaches doesn't matter. If it's using 70W of power, that eventually gets turned into waste heat that's dumped into the environment.

If they cranked up the fans so that the chip stayed at 35C, it just means the cooling system is more actively distributing that heat.

The only reason to be concerned with a chip sitting at 85C is if it's in a notebook that's only your lap. Even though most chips are built to cut out around 100C, they could continue to operate above that, but hitting that temp usually means the cooling system has failed and there's a risk of thermal runaway and hitting a temperature that is destructive.

 

[soresu]

risk of thermal runaway and hitting a temperature that is destructive

It would be interesting to see a study of what it takes to become destructive for a relative transistor or metal pitch these days.

 

[Doug S]

It would be interesting to see a study of what it takes to become destructive for a relative transistor or metal pitch these days.

IIRC a few years ago I saw something indicating that 130C was where things could become problematic, hence everyone throttles in the 115C-120C range.

 

[Glo.]

Content creation is a big market for the Mac Studio, and the computer sits on the desk. A big complaint for the M1 Max/Ultra Mac Studio was noise. It wasn't really fan noise related to speed but fan whine. Regardless, it was enough for some musicians and other users to return their Mac Studios. Apple has addressed this in the M2 Max/Ultra Mac Studio by altering the fan ramp up curve, but appropriate for the context of heat tolerances of those chips. I wonder also if they got a new fan supplier too.

The fan WILL ramp up, but only if absolutely necessary. No real point in trying to keep the chip at 65 C or lower. I think 85 C is just fine. Sure it might mean the Mac might only last 10 years instead of 15, but really, who cares? At least it's not 100 C like the previous Intel chips.

BTW, back in the day, I returned a Core i7 iMac because of fan noise. Anything I did that maxed out the CPU would ramp up the fan within 30 seconds. It was incredibly irritating to have a vacuum cleaner on my desk just exporting a simple 1 minute birthday video. So, I returned it and got a slower i5 iMac instead. Much happier. It was about 25% slower, but when the CPU was maxed out, it took ~9 minutes to ramp up the fan to max. That meant that for all shorter CPU-intensive tasks, the computer remained silent. Good tradeoff IMO. One could say the i5 was 18X as good as the i7 for fan noise performance.

The GPU itself never exceeds 70 degrees celsius no matter the load and fan speeds.

The SOC uses less than 90W of power at highest possible load. There is absolutely no worry about the longevity of this design, apart from the dust accumulation inside the computer.

 

[richardskrad]

delete

 

[Eug]

This Chinese dude, a chip designer since the Intel Pentium days, claims that A17 is starting on TSMC N3B but will switch to N3E next year.

Mobile phone chip expert's weibo

The A17 used in stock this year's iPhone15 pro and iPhone15 pro max is an N3B process, but the A17 produced at some point next year will be switched to a cost-reducing N3E process, which may be less efficient.

There is some precedent for this, A5 started on 45 nm and then switched to 32 nm. A9 was built on processes from two different companies, TSMC (16 nm) and Samsung (14 nm), and Apple was selling the phones with the different chips at the same time.

 

[Mopetar]

This Chinese dude, a chip designer since the Intel Pentium days, claims that A17 is starting on TSMC N3B but will switch to N3E next year.

Mobile phone chip expert's weibo



There is some precedent for this, A5 started on 45 nm and then switched to 32 nm. A9 was built on processes from two different companies, TSMC (16 nm) and Samsung (14 nm), and Apple was selling the phones with the different chips at the same time.

I'm assuming a jump from TSMC's N3B to N3E is considerably easier than jumping a whole node at the same fab or having two separate fabs build the same chip.

Apple might not even have to change masks unless the two nodes are vastly different despite the name/branding used by TSMC.

 

[trivik12]

N3B is not design compatible with N3E and so they have to redesign it for new node. But I dont know why Macrumors is saying N3E is "less efficient". Didn't TSMC say N3E is more efficient and has higher performance despite lower density.

 

[Eug]

N3B is not design compatible with N3E and so they have to redesign it for new node. But I dont know why Macrumors is saying N3E is "less efficient". Didn't TSMC say N3E is more efficient and has higher performance despite lower density.

Assuming the auto-translation was correct, it was the weibo guy who suggested in his post that N3E was less efficient, but anyhow you are correct according to AnandTech:

TSMC Details 3nm Evolution: N3E On Schedule, N3P and N3X To Deliver 5% Performance Gains

www.anandtech.com

 

[Doug S]

I've seen the claims about switching to N3E for A17 next year - I think what that may entail would be that the A17 used in next year's non-Pro models of iPhone 16 would use this (presumably cheaper to make even if the die is slightly larger) A17.

Though they did have two different A9s with different power draw (people were even returning phones they found to have a Samsung A9 then buying again hoping for a TSMC one) so maybe they could switch mid stream. But that mid stream change may not be worth the trouble for one or at most two quarters' worth (and the quarters with the least high end iPhones sold) depending on when N3E enters mass production and how long it takes for finished chips to be ready to be assembled into iPhones and those iPhones delivered to stores/carriers/customers.

 

[SpudLobby]

I've seen the claims about switching to N3E for A17 next year - I think what that may entail would be that the A17 used in next year's non-Pro models of iPhone 16 would use this (presumably cheaper to make even if the die is slightly larger) A17.

Though they did have two different A9s with different power draw (people were even returning phones they found to have a Samsung A9 then buying again hoping for a TSMC one) so maybe they could switch mid stream. But that mid stream change may not be worth the trouble for one or at most two quarters' worth (and the quarters with the least high end iPhones sold) depending on when N3E enters mass production and how long it takes for finished chips to be ready to be assembled into iPhones and those iPhones delivered to stores/carriers/customers.

IDK. I think it might be worthwhile if they're segmenting Pro/NonPro for the leading/last year A-series chips and anticipate A17 sales going well into 2024, 2025 (might use it in an iPad too). Might also be in time for the next SE. Remember, they still sell the iPhone 13 & the 12, and the SE uses the A15.

N3E is certainly the better node though (overall, put the SRAM aside vs cost, power, yields) what with FinFlex, the superior yields and modest performance boost. I don't think it'll be a big deal, though it'll certainly have improved power characteristics vs the N3B A17.

I'm skeptical though b/c of the tapeout costs, guess we'll see.

 

[Eug]

maybe they could switch mid stream. But that mid stream change may not be worth the trouble for one or at most two quarters' worth (and the quarters with the least high end iPhones sold) depending on when N3E enters mass production and how long it takes for finished chips to be ready to be assembled into iPhones and those iPhones delivered to stores/carriers/customers.

What's the disadvantage of switching chips mid-stream?

ie. What's wrong with having some A15 Pros on N3E A17 and all A16 non-Pros on N3E A17? Presumably CPU and GPU performance would be identical.

 

[A///]

Re: 65 Watts:

In Cinebench CPU, the M2 Ultra (24-core CPU and 60-core GPU) will peak at a little over 70 watts for CPU.

For the extended 10 minute Cinebench CPU test, the fans remained at baseline (which is silent), and temps did not exceed 85C on the hottest core.

at first this sounds outrageous given what apple said about using low power until you consider that this is 24 cores and 60 cores for cpu and gpu respectively operating at full breadth for minutes on end. 70 watts give or take 10 is brain scramblingly impressive. koudous to apple. IDK what the pc equivelant this setup would be to but I'm sure it'll be pushing more than 270 watts in total.

Performance that will warm your heart, but not your room.

like a dram of fine single malt on a wintery evening in front of a crackling fireplace whilst snow and sleet batter the outside of your home.

 

[A///]

Content creation is a big market for the Mac Studio, and the computer sits on the desk. A big complaint for the M1 Max/Ultra Mac Studio was noise. It wasn't really fan noise related to speed but fan whine.

the big complaint I saw with those was whine, coil whine to be exact, or was that a case of mistaken identity?

 

[Eug]

the big complaint I saw with those was whine, coil whine to be exact, or was that a case of mistaken identity?

Not sure but there were differing theories. In any case, one thing that masked the noise was to cover a part of the vent holes. The other thing was to change the fan speed. Neither is ideal for a $2000+ computer.

Initial reports are that for M2 series, it's likely far less common than it was for M1 series. What they did exactly I don't know, except that I do know they decreased the fan idle speed by a couple hundred RPM.

BTW, the official technical specs now have the noise of the M2 series Mac Studio as much quieter. Apple states that from operator position (whatever that is), noise has decreased from 15 dB (2022 model) to 6 dB (2023 model), a difference of 9 dB. That compares well to my uber quiet M1 Mac mini, which is listed at 5 dB.

Mac Studio (2023) - Technical Specifications - Apple Support

support.apple.com

 

[Doug S]

IDK. I think it might be worthwhile if they're segmenting Pro/NonPro for the leading/last year A-series chips and anticipate A17 sales going well into 2024, 2025 (might use it in an iPad too). Might also be in time for the next SE. Remember, they still sell the iPhone 13 & the 12, and the SE uses the A15.

N3E is certainly the better node though (overall, put the SRAM aside vs cost, power, yields) what with FinFlex, the superior yields and modest performance boost. I don't think it'll be a big deal, though it'll certainly have improved power characteristics vs the N3B A17.

I'm skeptical though b/c of the tapeout costs, guess we'll see.

If they do another SE they would have strong incentive to use A16, since while N3E will be cheaper than N3B it will still be more expensive than N4. Heck, they might even do a special version of A16 that uses LPDDR4X if there's still a meaningful difference in its price versus LPDDR5 since there's a lot of pressure on the BOM to make the SE's price point while preserving margins.

 

[Doug S]

What's the disadvantage of switching chips mid-stream?

ie. What's wrong with having some A15 Pros on N3E A17 and all A16 non-Pros on N3E A17? Presumably CPU and GPU performance would be identical.

More SKUs, more qualification, more complexity for repair/warranty. There's a reason why companies selling longer lived products will stick with the same design even when cheaper/better parts become available, unless the cost difference becomes large enough. And when they do it either gets sold under a different name (i.e. like consoles where it is still a "PS5" but they change the form factor and everyone knows if it is the original version or a newer version) or it'll be done as a rev (i.e. all the wireless routers that have rev 1, rev 2, rev 3 over the years that they sell the "same" router and while you might think you can treat them the same you'll see stuff like "only rev 1 and 2 can install DD-WRT, rev 3 uses a different SoC that it isn't compatible with")