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

 

[Hitman928]

M1 Max PugetBench Premiere Pro

View attachment 51752

Ryzen 9 5950X 16-core with GeForce RTX 3080

View attachment 51753

It seems in terms of just the GPU score, the desktop 3080 does much better, but in terms of the overall score they are similar.

Part of the reason is due to what I was talking about earlier, which was the addition of hardcore hardware acceleration in M1 Pro/Max. The robust hardware ProRes acceleration in M1 Pro/Max is going to go a very long way in making this model a darling of video editors.

This benchmark also doesn't really scale beyond 8 cores so you'd get basically the same result against a 5800x. Still impressive, but the 5950x would smoke the m1 max in something that would actually scale to 32 threads.

 

[StinkyPinky]

That's a desktop GPU though, so I don't think any reasonable person is expecting the M1 Max to get close to that.

 

[Mopetar]

Even Apple's comparisons were against a mobile 3080, which is basically a desktop 3070.

Really though, ~66% of a desktop 3080 in a laptop is still incredible. It's probably not going to be able to reach that in gaming workloads for unoptimized titles, but Apple has never been a choice for people who care about gaming.

 

[Eug]

That's a desktop GPU though, so I don't think any reasonable person is expecting the M1 Max to get close to that.

Yes, I mentioned it was a desktop 3080 in my post. However, my point was that M1 Max beats it handily while dealing with ProRes playback. In fact, it's in a totally different league.

For 4K ProRes 422 4X playback, M1 Max is getting 216 fps, whereas the Ryzen machine is getting 48 fps. Cue the "buttery-smooth scrubbing" comments in the YouTuber reviews on Monday.

These chips are designed with very specific pro customers in mind. They don't really give a chit if they score well in gaming benchmarks or whatever.

 

[biostud]

While it is a significant technological improvement in efficiency, is it really that important for a laptop from a competitive standpoint? Most of the time you will be at a desk close to a power source.

 

[Eug]

While it is a significant technological improvement in efficiency, is it really that important for a laptop? Most of the time you will be at a desk close to a power source.

Pretty much every review about the high end 16" Intel MacBook Pro states that it is obnoxiously loud under load. One particular concern is that it becomes problematic for music production for example, because it's so loud.

And yes, power usage matters because of battery life. The MacBook Pro is very popular with creatives on the road.

 

[BorisTheBlade82]

@biostud
Well, that surely depends on if you are the kind of person that likes to work on the outside or in a Café, etc. I would guess that in the last 1,5 years the amount of those people has risen significantly.

 

[Mopetar]

Yes, I mentioned it was a desktop 3080 in my post. However, my point was that M1 Max beats it handily while dealing with ProRes playback. In fact, it's in a totally different league.

The M1 chips have dedicated hardware for ProRes so it's not really a test of the GPU itself and it's unsurprising that Apple performs better. The GPU specific portion of the test is still valid, interesting, and not really out of line with expectations so it's not really a waste.

This is just Apple realizing that most of their users have a specific use case for their MacBooks and trying to accelerate that to give them a reason to buy Apple. Well more of a reason at any rate.

 

[Eug]

The M1 chips have dedicated hardware for ProRes so it's not really a test of the GPU itself and it's unsurprising that Apple performs better. The GPU specific portion of the test is still valid, interesting, and not really out of line with expectations so it's not really a waste.

This is just Apple realizing that most of their users have a specific use case for their MacBooks and trying to accelerate that to give them a reason to buy Apple. Well more of a reason at any rate.

Yes, I already said that part about ProRes acceleration in my original post.

And I'll repeat what I said a few days ago. Those here focusing just on raw general GPU performance and raw general CPU performance are kinda missing the point of M1 Max for Apple's target customers. Apple's hardware ProRes acceleration alone is a bigger deal in the real world for a ton of Apple's target audience than 1000 extra points in Geekbench CPU or 5000 extra points in Geekbench Metal.

 

[insertcarehere]

While it is a significant technological improvement in efficiency, is it really that important for a laptop from a competitive standpoint? Most of the time you will be at a desk close to a power source.

Living somewhere that already deals with horrifically hot and humid summers, being able to do the same work with a few hundred watts less heat dissipated into a small room would have noticeable benefits to the monthly a/c ball.

 

[jeanlain]

1. I doubt Apple would have run the tests in low power mode.

Apple? We're talking about geekbench tests ran by whoever.

 

[Eug]

Apple? We're talking about geekbench tests ran by whoever.

And as I indicated in my post in point #2, I'm talking about Apple's own GPU tests that demonstrate no higher than 1.71X scaling from M1 Pro 16 to M1 Max 32.

Perhaps I should have led with that as point #1, and put the Apple likely not benchmarking with low power mode as point #2.

 

[jeanlain]

2. In one of my previous posts I went through all of the GPU related performance benchmarks that Apple published in its press release, and calculated the scaling between M1 Pro 16 and M1 Max 32. Not a single GPU test published in Apple's press release - from Apple's own lab - scaled at near 2X. The range was 1.39X to 1.71X.

Which is disappointing, as far as scaling is concerned. The geekbench compute test scales better with the number of compute units for Nvidia GPUs, even though they have more CUs than the M1 Max.
Apple's tests may be partly constrained by the CPU, for all we know.

The M1 Max is still impressive in terms of perf/W.

 

[The Hardcard]

Yes, I already said that part about ProRes acceleration in my original post.

And I'll repeat what I said a few days ago. Those here focusing just on raw general GPU performance and raw general CPU performance are kinda missing the point of M1 Max for Apple's target customers. Apple's hardware ProRes acceleration alone is a bigger deal in the real world for a ton of Apple's target audience than 1000 extra points in Geekbench CPU or 5000 extra points in Geekbench Metal.

A huge deal, but I wouldn’t say a bigger deal, since after joyfully scrubbing though multiple streams of ProRes footage, many will be freely adding numerous effects and be looking for the GPU to keep up with processing them in full size in real time. ProRes acceleration will mean much more work for the GPU.

 

[Eug]

Wired: A Look Inside Apple's Silicon Playbook

Srouji is clear on the advantages of rolling out your own chips, as opposed to buying from a vendor like Intel, which was summarily booted from MacBook Pros this week in favor of the M's. “When you're a merchant vendor, a company that delivers off-the-shelf components or silicon to many customers, you have to figure what is the least common denominator—what is it that everyone needs across many years?” he says. “We work as one team—the silicon, the hardware, the software, the industrial design, and other teams—to enable a certain vision. When you translate that to silicon, that gives us a very unique opportunity and freedom because now you're designing something that is not only truly unique, but optimized for a certain product.” In the case of the MacBook Pro, he says, he sat with leaders like Ternus and Craig Federighi several years ago and envisioned what users would be able to get their hands on in 2021. It would all spring from the silicon. “We sit together, and say, ‘Okay, is it gated by physics? Or is it something we can go beyond?’ And then, if it's not gated by physics and it's a matter of time, we go figure out how to build it.”

—-

This illustrates what I was talking about before. The chips are purpose built to serve their target Mac audience, not for raw general purpose GPU and CPU performance targets.

For example, for some to suggest that ProRes performance measures shouldn’t be counted when assessing M1 Pro/Max because it involves a hardware acceleration “cheat” is IMO kinda missing the point. I’d argue having buttery smooth video scrubbing is a lot more useful to pro customers than many other things people here obsess about.

 

[Eug]

BTW, I had forgotten, but someone pointed out the fact that AnandTech got a review unit for the M1 launch. Hopefully the same will be true again for the M1 Max launch.

 

[Red_m]

MAN, M1 Max going head to head with Mobile RTX 3080.

 

[Eug]

MAN, M1 Max going head to head with Mobile RTX 3080.
View attachment 51802

Here is the original article:

Apple's M1 Max Benchmarked in Adobe Premiere Pro: A Mixed Bag

Good in video, bad in GPU compute.

www.tomshardware.com

Apple's M1 Max clearly has an excellent media playback engine that outperforms not only standalone mobile GPUs, but even Nvidia's top-of-the-range GeForce RTX 3090. Those who edit video will certainly appreciate Apple's M1 Max since based on PugetBench for Premiere Pro 0.95.1, the new MacBook Pro systems promise a very smooth experience that will be even better than that on beefy desktops.

Like I said, cue the "buttery smooth scrubbing" reviews...

 

[majord]

Not unless they (qualcomm, you say Microsoft, haven't researched, but I doubt Microsoft is pushing anything other than DirectX) manage to drum up a 9-10 figure gaming business. AMD isn't going to touch niche APIs of any kind unless they show staying power.




Are you saying they lied on their own charts? They claimed "plugged in" performance was the same thing as "battery performance". My prediction: once the real reviews roll in, most of you guys will disappear or stick around to move goal posts. Note that I'm pretty vendor neutral here, but when I see marketing calling the shots, I know what is really going on...



It's a limitation in "Apple products". Geekbench has had limitations in the past, but the OpenCL test has proven to scale above and beyond (it is several years old at this point and scales above even the 3090). Apple has a slow GPU, slow drivers, or both (betting on the hardware. I know from experience). A bunch of you folks are expecting Apple to perform miracles. RTX 3060 mobile performance IS a miracle in a Mac Product! It will NOT set global performance records. It will set SOME, but not many, efficiency records.99% of the records it sets are due to NVIDIA being 2-3 node generations behind Apple. If NVIDIA had GPUs on TSMC 5nm, they would be ahead. This is a fact. I'm not an NVIDIA fan. I actually only own stock in AMD (and a small amount of Apple shares), NVIDIA is very competitive when it comes to perf/watt. There is a reason when AMD, on a superior process, has trouble beating NVIDIA. NVIDIA is a giant and can throw tons of money at R&D. Apple has a good first try, but with both CPU and GPU performance, most of you aren't seeing the bigger picture:

It is easy to build an okay product on a superior node and claim a win. What happens next for all will determine who wins or loses. Thus far, Apple has several losses, including software compatibility, lack of high refresh rate displays, lack of mainstream API support, etc. Their only wins IMO, are on their own, in house, operating system in a closed ecosystem. which I don't personally consider wins, because they don't change the mainstream market. They need to step up to the plate if they want to claim real wins. Neither Windows nor Linux (GUI applications) currently run on Mac hardware unless being emulated. Get back to us when they do. If their emulated solutions can outperform cutting edge native x86 solutions, or if they figured out how to get all the rest of the software in the world to run faster than it does on an x86 CPU, I'll be all ears. As of now, MY company is ditching Macs because the fastest server CPU is x86, not ARM.

Only thing I would argue there is it's not trivial to build on a superior node - it takes a lot of Execution prowess , and risk - and that's where I give Apple credit for most here. They're first at this node - with big die's. That can't be ignored. Architecture wise, yes I think there's a lot of hyperbole. I've already talked about the CPU , which I think is absolute top tier, but nothing more, or magical.. Apple's CPU's have been excellent for years now though. The GPU otoh is less impressive , which stands to reason given it's their first architecture. It follows the same concept of the CPU 'wide' and slow , only 'big and slow' . Beyond ensuring you can scale well to high core/shader counts , the rest is just throwing silicon at it, and dropping the clockspeed and voltage to the floor (1.3Ghz - that's pretty bottom of the barrel @ 5nm) , and bingo, you have amazing Perf/watt. This just makes the SoC expensive - but when you're charging 3K+ for a laptop, who gives a s__t.

I actually wish AMD and Nvidia would start down clocking large die's again for top end laptop GPU's. The trend they both fell into using smaller die's and clocking them high has resulted in pretty piss poor perf/watt. AMD really are in the best position to have done that, as NAVI 21 memory bus is only 256Bit which is practical for a laptop. Instead they've clocked NAvi 22 to the moon for the 6800M, which is very disapointing as it would have been interesting to see a very low power high perf option (at a high price) for premium thin gaming laptops.

A downclocked Navi 21 has VERY good perf/watt. enough to make M1 MAX look not particularly impressive at all. I actually tried to lower a 6800XT down to ~10TF, and you can't actually get that low without running into a BIOS imposed lower Vcore floor well before then. (not undervolting - just stock V/F curve) so there's no way to continue down the VF curve below that.. Will try and get close with a 6800, which should get a bit closer as its lower Vcore is, I think 760mV vs 860

the other thing working in Apples favour is if you don't have higher performance intentions, you don't need to scale to higher frequencies to serve that market.. AMD spent the last few generations getting their architecture's to scale well beyond 2Ghz so they could keep up with Nvidia at the high end. I don't see any evidence or reason Apple would have done the same. As I said before it's a very focused design, as evidence by the fact every variation has identical clockspeeds. If you're limiting the design to these clockspeeds, you can further focus on density and low power at the phy implementation.

 

[biostud]

If the target of m1 pro/max is professionals, shouldn't they be compared to quadro cards?

Or said in another way, Apple has build a SoC which strengths are photo/video work, which makes good sense since that is their core customers for the MacBooks. Wouldn't it be pretty bad if they couldn't beat a card made primarily for gaming?

 

[Doug S]

the other thing working in Apples favour is if you don't have higher performance intentions, you don't need to scale to higher frequencies to serve that market.. AMD spent the last few generations getting their architecture's to scale well beyond 2Ghz so they could keep up with Nvidia at the high end. I don't see any evidence or reason Apple would have done the same. As I said before it's a very focused design, as evidence by the fact every variation has identical clockspeeds. If you're limiting the design to these clockspeeds, you can further focus on density and low power at the phy implementation.

To be fair, the reason they have the same (low) clockspeeds across the line is because they designed their GPU for a phone. That's their big market. The Mac market looks huge when you compare it against most other companies, but against iPhone it looks insignificant. Hell, when the iPad Pro was announced there were more than a few people claiming that the rumors about Macs going ARM were wrong because Apple would drop the Mac line entirely in favor of the iPad Pro!

Even within the Mac line the bulk of their sales are non-Pro laptops/desktops, so from a revenue perspective (let alone units) Macs containing M1 Pro/M1 Max will be less than 1% of iPhone. It is clear what product drives the bus at Apple, so the design for the higher end Macs will always be a compromise. Considering that, the performance they've achieved relative to PC laptops (let alone those in the same power envelope) is a pretty damn good accomplishment.

 

[Doug S]

If the target of m1 pro/max is professionals, shouldn't they be compared to quadro cards?

Or said in another way, Apple has build a SoC which strengths are photo/video work, which makes good sense since that is their core customers for the MacBooks. Wouldn't it be pretty bad if they couldn't beat a card made primarily for gaming?

Can those Quadro cards beat the gamer cards? I was always under the impression the high end cards in both lines were basically the same, aside from the amount of on board DRAM and the driver. But I don't follow the GPU market much, so that impression could be wrong.

Anyway yes I agree, I've said for a long time that Apple will care about how they compare with the workstation GPUs not the gamer GPUs, and on workstation GPU benchmarks not gaming benchmarks. But in a forum like Anandtech most people are gamers so those are the benchmarks they know and use. The people deciding whether to buy a fleet of Macbook Pros or down the road Mac Pros will care how they perform on the applications they use. They don't care if the Macbook Pro sucks at all games, so long as it performs in what they need to do - and those workstation type workloads are what Apple is going to tune their drivers for, not games.

 

[Eug]

Even within the Mac line the bulk of their sales are non-Pro laptops/desktops, so from a revenue perspective (let alone units) Macs containing M1 Pro/M1 Max will be less than 1% of iPhone. It is clear what product drives the bus at Apple, so the design for the higher end Macs will always be a compromise. Considering that, the performance they've achieved relative to PC laptops (let alone those in the same power envelope) is a pretty damn good accomplishment.

iPhone revenue is ginormous, but your 1% claim is off. Mac revenue is sometimes over 20% that of iPhone revenue. Are you suggesting that M1 Pro / Max (or similar future chips) will be less than 5% of Mac sales?

EDIT:

Numbers fixed. Thanks @Leo9.

 

[Leo9]

 

[Eug]

View attachment 51822

Thanks. Even though I was posting to say @Doug S was underestimating Mac sales, it looks like even my number was also a big underestimate too. Post corrected. That even furthers my argument now.