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

 

[jeanlain]

Anandtech's review of the M1 Mini was lacking an in-depth analysis of power consumption. The authors apparently didn't know about the powermetrics command by then. Interestingly, this command dose not report Wattage on intel Macs (though we have other tools for that).
I'm looking forward to another review including perf/W comparisons.

 

[coercitiv]

A lot of AMD fanboys in denial here.

The irony of it all is literally everybody on this forum agreed M1 is a revolutionary piece of silicon.

But that wasn't enough, because having everyone agree means somebody lacks the means to prove their superiority on the Internets. So you push and push until claims inflate, mutate and become ridiculously false. And when they start reacting you can finally shout to release the pressure: FANNNBOYYYS! They can't accept simple facts! Prove me wrong!

And so instead of having a nice thread to follow performance updates on M1 - which is exactly what the OP and other Apple consumers from Anandtech are interested in, we're forced yet again to filter through regurgitated emotional reactions from a few "alpha" posters who need their daily dosage of drama on the forums.

 

[thunng8]

So now we're ok with reported power draw? In any case, that entire segment was a complete non-sequitor, as I mentioned, since we don't have actual power draws under CB for both chips, we aren't apparently using the same measurement system, and so on. Continuing down that path is silly, but I thought interesting enough to consider! I really would love to have more data to make a better comparison, but all we have right now is speculation when comparing with contemporary laptop chips.

Anyway, do you care to address my primary response?

"The example I gave is one of an entire laptop, screen included, under full load - including GPU - using less power (37.2W) than a 3700X under a 2-thread load (39.63W - reference). Which sufficiently makes my point that using the power consumption of a desktop chip to criticize the potential power consumption of its laptop companion is a fool's errand."

Here you go: 4750U draws 27.9-29.9W (CPU package power) under load in cinebench r23. This is double that of the m1 in the mac mini and 4X more than the macbook air when throttled.

AMD laptops have a hidden 10-second performance delay. Here’s why

AMD's flagship Zen 2 laptop CPUs save power by delaying their full performance.

arstechnica.com

 

[DrMrLordX]

Name one "edge case" where you can demonstrate this "level of frustration" you are claiming. I have yet to see anything where the M1 Mac is slower the x86 Mac it is replacing other than something using hand optimized x86 SIMD code.

Congratulations! You just named one. But really it could be anyone using a niche MacOS application that doesn't get updated for native M1 binaries and has to be run via Rosetta 2 in perpetuity. Like, I don't know, games no longer supported by the publisher?

I'm not DrMrLordX, but I'm a Linux user. My level of frustration is not being able to replace MacOS with Linux even though I do like Apple's approach to hardware design most of the time.

Yeah that is a bummer. It is very nice that at least M1 runs under MacOS instead of iOS. Now we can actually sort-of benchmark the thing.

The irony of it all is literally everybody on this forum agreed M1 is a revolutionary piece of silicon.

But that wasn't enough, because having everyone agree means somebody lacks the means to prove their superiority on the Internets. So you push and push until claims inflate, mutate and become ridiculously false. And when they start reacting you can finally shout to release the pressure: FANNNBOYYYS! They can't accept simple facts! Prove me wrong!

And so instead of having a nice thread to follow performance updates on M1 - which is exactly what the OP and other Apple consumers from Anandtech are interested in, we're forced yet again to filter through regurgitated emotional reactions from a few "alpha" posters who need their daily dosage of drama on the forums.

Either that, or someone is trying aggressively to sell a few more M1-based laptops for Apple by claiming they aren't just excellent, but that they're the solution to every user's problems. Need a 5950X? Nah, buy an M1.

edit:

For those wondering about what will and won't run natively on M1, I just happened across this article here:

This handy site tells you if specific apps work on Apple Silicon yet

Are you interested in getting a new M1-powered Mac? Well, make sure you check this site that tells you if your favorite appes work on Apple Silicon before.

thenextweb.com

Here's the actual site you would need to visit to get some information. It's not going to be absolutely comprehensive:

Is Apple Silicon ready?

The Complete Guide to MacOS Apps Optimized for Apple Silicon Macs and the M1/M2 Processors

isapplesiliconready.com

 

[Leeea]

Name one "edge case" where you can demonstrate this "level of frustration" you are claiming.

Is Apple Silicon Ready? - Slashdot

Website list of apps that don't work so far (even with rosetta 2 emulation): LFO Tool plugin Native instruments Nectar plugin Neutron plugin Pro tools Screenflow Google Drive File Stream Virtualbox Revit Parallels VMWare Fusion Haskell I understand why VM software wouldn't work, but google...

tech.slashdot.org

A lot of those are pro-audio tools, used for both music production and multimedia (video) production.


Sure, odds are they will be fixed soon enough, but there are plenty of frustrating "edge" cases of complete non-function at the moment.

 

[Hitman928]

Here you go: 4750U draws 27.9-29.9W (CPU package power) under load in cinebench r23. This is double that of the m1 in the mac mini and 4X more than the macbook air when throttled.

AMD laptops have a hidden 10-second performance delay. Here’s why

AMD's flagship Zen 2 laptop CPUs save power by delaying their full performance.

arstechnica.com

Just FYI, that 4750u has a cTDP of 25W, so it is configured to use more power than the default 15W.

 

[amrnuke]

Here you go: 4750U draws 27.9-29.9W (CPU package power) under load in cinebench r23. This is double that of the m1 in the mac mini and 4X more than the macbook air when throttled.

AMD laptops have a hidden 10-second performance delay. Here’s why

AMD's flagship Zen 2 laptop CPUs save power by delaying their full performance.

arstechnica.com

Perfect - you gave me package power for a 25W TDP variant of the 4750U, not the 15W... and again, we're ok with just accepting the reported power consumption now rather than an actual at-the-wall measurement? Until we get at-the-wall measurements on both a 15W TDP 4750U and MBA/MBP running CB23 (not saying it'll be hugely different), it still won't be a valid evaluation. Then again, like I said, that entire segment was non-sequitor and complete speculation.

In any case, again, back to my original point, which was that someone said because the Zen3 desktop chips draw so much power under single-core loading, the laptop can't be efficient, and my providing evidence that a single core load on 3700X draws more power than a fully CPU and GPU loaded laptop, do you have any input?

 

[jeanlain]

Perfect - you gave me package power for a 25W TDP variant of the 4750U, not the 15W... and again, we're ok with just accepting the reported power consumption now rather than an actual at-the-wall measurement?

I'm no expect, but the shouldn't the reported power consumption be more precise? Plus, it can isolate the CPU package and even individual cores. Anandtech use the reported numbers all the time.

 

[Hitman928]

I'm no expect, but the shouldn't the reported power consumption be more precise? Plus, it can isolate the CPU package and even individual cores. Anandtech use the reported numbers all the time.

His main point is that someone used desktop Ryzen power numbers to point to how much more efficient M1 is, but desktop Ryzen isn't the right comparison as that architecture is essentially designed to scale up to 64 cores and so it makes some compromises in order to be able to do that (one major one is power efficiency at low core counts). It is a much better comparison to use Renoir which has a more similar target market and is designed for that market. Renoir's efficiency, especially at low core counts, will be much higher than desktop Ryzen.

 

[jeanlain]

I was replying to his question related to the measure of power consumption, irrespective of the type of CPU.

 

[amrnuke]

I'm no expect, but the shouldn't the reported power consumption be more precise? Plus, it can isolate the CPU package and even individual cores. Anandtech use the reported numbers all the time.

No, it's not always precise. As one example, there was a huge deal this summer about reported power consumption deviation leading to earlier throttling of the CPU on Ryzen chips.

 

[lobz]

I'm not DrMrLordX, but I'm a Linux user. My level of frustration is not being able to replace MacOS with Linux even though I do like Apple's approach to hardware design most of the time.

Why would you use Linux when Apple's version is just better in every way?

 

[Heartbreaker]

Next Generation AMD APU, does't appear to be that impressive vs M1:

AMD Ryzen 7 5800U Flagship 8 Core Cezanne 'Zen 3' APU Benchmark Leaks Out, Up To 38% Faster Than 4800U in Single-Core Performance

The first benchmark of AMD's upcoming Ryzen 7 5800U APU which is an 8 core chip and part of the Zen 3 based Cezanne family has leaked out.

wccftech.com

ST performance still behind M1, GPU is still 8 core VEGA.

 

[lobz]

I'm no expect, but the shouldn't the reported power consumption be more precise? Plus, it can isolate the CPU package and even individual cores. Anandtech use the reported numbers all the time.

The idea would be to use numbers gathered with the same method from both systems.

 

[Hitman928]

Next Generation AMD APU, does't appear to be that impressive vs M1:

AMD Ryzen 7 5800U Flagship 8 Core Cezanne 'Zen 3' APU Benchmark Leaks Out, Up To 38% Faster Than 4800U in Single-Core Performance

The first benchmark of AMD's upcoming Ryzen 7 5800U APU which is an 8 core chip and part of the Zen 3 based Cezanne family has leaked out.

wccftech.com

ST performance still behind M1, GPU is still 8 core VEGA.

Compared to their 4800u example, that is extremely impressive. That is a 38% ST performance increase on the same node. Obviously the M1 is still faster in ST Geekbench, but the 5800u on Linux should get ~10% performance boost which would only put it about 12% behind the M1, though still using an unknown amount more power. I'm guessing that AMD probably allows the 5800u to use more power in single thread scenarios than they allowed for the 4800u in response to Intel letting their laptop CPUs use so much power in single thread scenarios to win more benchmarks.

The multithreaded improvement is much more muted, but again, we don't know the power configuration of both examples given by WCCF so it's hard to really compare. The Zen3 Geekbench is also still using Clang 9 versus Clang 12 on M1. Might not account for much, but it makes it hard to compare 1:1 as we have no idea how much of a performance difference the updated compiler might make for Zen3.

The GPU cores still being Vega has been known for a while.

 

[jeanlain]

Why would you use Linux when Apple's version is just better in every way?

Better in every way? Nope.
And this is coming from a Mac user.

 

[coercitiv]

Next Generation AMD APU, does't appear to be that impressive vs M1:

That GB5 run used DDR4 2666, Cezanne won't beat M1 in Geekbench but should still score around ~1600 ST with better memory speed.

 

[HurleyBird]

Yup. Windows and slow memory, and M1 is a bit better in Geekbench than apps.

Cezanne should have rough ST parity and a good MT lead while getting spanked in graphics at something like half the efficiency.

 

[Heartbreaker]

The GPU cores still being Vega has been known for a while.

Still a disappointment. It means no movement on x86 integrated graphics for yet another generation.

 

[Antey]

AMD integrated graphics are not x86, it's GCN isa

and GCN is RISC! Don't you love it?

PD: Wait for Van gogh RDNA2 + LPDDR5

 

[moinmoin]

Why would you use Linux when Apple's version is just better in every way?

Apple's version of what?

 

[Heartbreaker]

Apple's version of what?

I would assume Unix, or Unix-like.. Linux being an open source Unix-like OS.

 

[moinmoin]

I would assume Unix, or Unix-like.. Linux being an open source Unix-like OS.

Linux is technically just the kernel. Usually it uses the GNU userland, whereas MacOS respectively uses XNU as kernel and Darwin as userland.
Anyway I want to hear from @lobz what the heck he is referring to with his question.

 

[Heartbreaker]

Linux is technically just the kernel.

You are technically just being needlessly pedantic.

More accurately it's both, but it's more commonly referred to as an OS, and that includes by just about everyone in the Linux industry.

What is Linux?

Linux is an open source operating system that is made up of the kernel, the base component of the OS, and the tools, apps, and services bundled along with it.

www.redhat.com

https://www.linux.com/what-is-linux/

Just like Windows, iOS, and Mac OS, Linux is an operating system.

 

[coercitiv]

Still a disappointment. It means no movement on x86 integrated graphics for yet another generation.

Cezanne marks a more aggressive AMD in the mobile field, not another delay. They aim to compress the gap between server/desktop and mobile, since in the last years they didn't really have the resources to have them ready at the same time. The 5000 series is bound to be very competitive in the x86 ecosystem, albeit less impressive overall now that Apple showed their hand.