Apples new M1 microprocessor

[joseph nicholas]

As quoted from the Wall Street Journal.

Apple’s M1 and the A14 chips
for the latest iPhones and iPads
will be built on TSMC’s 5-nanome-
ter manufacturing process. Intel’s
most advanced chips are built on a
10-nanometer process and delays
with its transition to 7-nanometer
have shaken up the chip maker.

These chips don't require a fan in the new Apple laptop computers.

[tom66]

Apple has already made fanless laptops in the past.  It's not impossible, but the real question is how well does it hold up under heavy workloads.

All well and good making claims like "3.5x faster" but if it can only run a benchmark like that for 10 seconds before thermally throttling...

[Cerebus]

These chips don't require a fan in the new Apple laptop computers.

Not quite so. In the two laptops released so far with this chip, one is fanless (the MacBook Air), the other has a fan (the MacBook Pro).

[hans]

Apple has already made fanless laptops in the past.  It's not impossible, but the real question is how well does it hold up under heavy workloads.

All well and good making claims like "3.5x faster" but if it can only run a benchmark like that for 10 seconds before thermally throttling...

My intuition: fanless designs only cool a couple of W. So max TDP sustained will probably throttle the CPU.
So for the real picture on the M1, ignore the Macbook Air.

I also find the 3.5x faster claim to be unfounded. They only showed 2 curves with no axis labels or anything. They probably compared it to their last-gen Intel devices though, so I would AMD to put a bit more resistance with their 4800U.. but even that chip isn't 3.5x faster than Intel alternatives. So if true, and Geekbench is accurate for once, that would be impressive.

But I'm hard pressed that they actually built a CPU with single-core up there with desktop CPUs that need turbo's with 65W+ TDPs. Or perhaps x86 CISC is so outdated and bloated, that a RISC CPU is the way to go. Let's wait and see.. and with see I mean actual benchmarks, like ffmpeg x265 encoding, WAV>FLAC encoding, chromium compile, Blender renders, things like that.

[Berni]

Yep fanless computers are nothing new, even x86 machines do it, but they tend to be slow.

Apple did some big claims there including "Worlds fastest CPU core" and "Worlds fastest integrated GPU". Doing that with a ARM core is very unlikely if you ask me. Those ARM cores are optimized for power and transistor count while x86 has constantly been pushing for more performance/MHz with power being only enough of a concern to make sure it doesn't melt at the high clock speeds. Apple is also famous for shitty cooling solutions that cook the CPU to death, but at least in that regard the ARM helps

So in order to quality for both of those claims the M1 chip needs to have higher performance than for example the "Ryzen 9 4900H" with its intel matching performance/Mhz zipping along at 3.3GHz base 4.4GHz boost and Vega 8 graphics at 1.7GHz. But they never mentioned the word "mobile" in "Worlds fastest CPU core", so technically for that claim to be true they need to beat the current single threaded performance world champion that is the recently released "Ryzen 9 5950X" desktop CPU that combines the significant performance/MHz improvements of 3rd generation Zen architecture with a 4.9 GHz boost clock. Getting that from an ARM? Hah no way in hell.

However i do think they might be able to go for "Worlds fastest ARM core" or "Worlds highest performance/watt" processor since the new TSMC process node is sure to give a great advantage there. But then just say so apple! Id be plenty impressed if you pull that off, no need to lie and oversell the performance that is most likely not there.

So as usual Apples claims are as usual baseless marketing bullshit that don't provide any meaningful numbers or specs.

[bd139]

Apparently this coordinated release is now officially a shit show. There is no virtualisation support at API level so software devs are shot. On top of that macOS now calls home to check binary signing on app startup. The cluster that handled this fell over yesterday and locked up huge numbers of people’s kit entirely. This drew attention to the fact that it can be used for surveillance. Also iMessage end to end encryption is apparently no longer appearing in documentation. On top of that people are braking NDAs and the performance is definitely exaggerated outside of synthetic benchmarks.

Also i note that to upgrade to 16Gb of RAM from 8Gb will cost you £200 here and you have to do this up front. I paid £248 for the 64Gb in my desktop PC.

Even as a long term Apple user I’m out. They can go to hell. This is a ridiculous proposition.

[Zucca]

Even as a long term Apple user I’m out. They can go to hell. This is a ridiculous proposition.

[bd139]

LMAO half the software dev team at work have broken Macs this morning thanks to Big Turd as well

Friday the 13th is going to be an interesting day.

[JPortici]

Apple is also famous for shitty cooling solutions that cook the CPU to death

Don't worry, 'cause this time they have ACTIVE COOLING (Damn i can't find that ridiculous photo from the presentation)

[bd139]

Do you have to blow on it like your dinner? 

[Berni]

Apple is also famous for shitty cooling solutions that cook the CPU to death

Don't worry, 'cause this time they have ACTIVE COOLING (Damn i can't find that ridiculous photo from the presentation)

Mhm... just like the active cooling in the iMacs that lets the CPU get to >80 °C where it thermal throttles to protect itself to the point of loosing 1/3 of the performance, yet the fans still don't get to 100% speed because the users don't like loud fans. And this is a desktop PC, not a laptop.

So are they going to stick these silly little ARMs into desktops too in order to solve cooling issues.

[tom66]

Many of these benchmarks are multithreaded benchmarks.  Which is great if your workload is multithreaded but so much is not multithreadable.

I have a 2019 Ryzen 3800X and its single-thread performance was only 50% faster than my 2010 Core i5.  But multithread almost 600% faster.

It looks like Apple are using big.LITTLE which is a common ARM architecture but it doesn't make a true octo-core CPU as only the big or LITTLE core in a core set can be active (presumably because they have some shared resources.)  Maybe Apple has unlocked that for specific workloads that aren't data dependent.

I also note Apple are shy about the clock rate of their CPU - now this isn't a useful benchmark between different architectures but might indicate how far behind Apple are.  It takes a lot of knowledge to be able to get an x86 up to 4GHz.  I'd be surprised if they're cracking 2GHz.  Which will mean M1 will be exclusively for silly laptops, no one will use it in a proper workstation.

[bd139]

On benchmarks it's a little more disingenuous on Apple's part as well. It reports around the same as a Ryzen 5950X single threaded on Geekbench, not the most reliable synthetic benchmark either. But that's a pretty crap comparison because as mentioned the M1 only has 4 equal fast cores (versus 16 on the 5950X) and 4 shitty ones. The Ryzen has 16 equal cores, much larger unified caches and also when you take TDP into account, the M1 efficiency per core is still only 40% lower than the Ryzen and AMD are making reliable 20% gains per generation at the moment on performance at the same TDP and apparently are already testing parts on TSMC 5nm process.

Also the M1 is clocking 3.2GHz base frequency. And it is 100% thermal throttled so like hell it'll be able to sustain the workloads they say it will.

I get the feeling what is going to happen is this opens up a whole new benchmarking dick swinging competition rather than looking at the actual gains which so far apparently to really be a loss of privacy, decrease in overall flexibility, increase in cost and general bad story for the end user. Your computer is turned from a general purpose computer into an Apple Toaster.

[madires]

I doubt that Apple will sell the M1 to other manufacturers. It's meant to help Apple to increase profit and to entice more users to enter Apple's walled garden. Hence the aggressive marketing.

[Berni]

Quite a few benchmarks out there have an option to run in single threaded mode.

The large steps that AMD has been taking in architecture improvements has closed the performance/MHz advantage that Intel. Combined with the high clock speeds means that the new Ryzen 5000 series beats intel in both multithreaded and singlethreaded benchmarks. This was pretty much the last advantage that Intel had, so they better do something quick.

For Apple to swoop in out of nowhere and makes a ARM that beats both is pretty much science fiction.

And since they are going to use this in the MacBook Pro means they intend this M1 chip to do heavy workloads like video editing or rendering (so also multithreaded) since the previous generation of MacBooks offers hefty high clock speed i7 processors (That thermal throttle due to overheating anyway but that's besides the point).

[george.b]

I found this to be an interesting read: https://www.anandtech.com/show/16226/apple-silicon-m1-a14-deep-dive

[hans]

Let's also not forget that Apple really profits if they can show huge performance benefits of their own Macbooks vs their previous. They must have had plans for this M1 chip underway for years, so I don't think the cynical narrative of "planned cooling woes" in Intel designs of recent years isn't that unreal. I mean, who designs a freaking laptop with barely any heatsink contact on the CPU, and then adds a fan on the other side of the laptop? Yes, Apple:



So in my opinion, any trumpet blown by Apple will be misleading, even in an Apple-to-Apple comparison (pun intended). I don't think they will show a honest comparison of silicon. I believe LinusTechTips was able to boost the i3 Macbook Air performance by about 13% just by "fixing" the cooling design. What if, magically, they know how to build cooling designs again?

I'm undecided  about the ARM cores coming out of nowhere. Recent ARM Cortex A7x designs are showing more/wider execution units, but I don't think they have both surpassed Zen 3's width nor clockspeed. So in that regard, it sounds like science fiction. The perf/MHz would almost suggest that even a Zen 3 core is stalled half of it's clock cycles. Then again, if you look at individual Geekbench scores, it does suggest that a single core can offer the performance in most tests...

[madires]

IMHO, Apple is comparing a sports car (customized ARM for mobile applications) with an SUV (x86). You can't transport a fridge with the sports cars, but it will fit perfectly into the SUV.

[Cerebus]

On top of that macOS now calls home to check binary signing on app startup. The cluster that handled this fell over yesterday and locked up huge numbers of people’s kit entirely. This drew attention to the fact that it can be used for surveillance.

This needs to be tempered with a little bit of detail. It's just an OCSP certificate check. Most people reading this whatever their browser and operating system are suffering from the same security leak every time they visit a website using TLS if that website isn't serving a stapled OCSP certificate along with the standard X.509 certificate. The difference is that the OCSP check is going to go to the issuer of the site's X.509 certificate, and it's them that get the opportunity to surveil you should they choose to. In this case the OCSP check is being done as part of checking the signature of a binary, and Apple have their own PKI so the OCSP request goes to Apple servers. MacOS, like most others, does some OCSP certificate caching so the apparent 'phone home' only happens when cached certificates expire or are first seen.

Yes, it looks like someone at Apple didn't account for the huge number of OCSP requests that would be caused by a whole bunch of people installing Big Sur* and starting off with empty OCSP caches and so, for a while, Apple's PKI servers choked on the load. It's not like they are the only large scale provider to get their sums wrong and stuff it up.

* What kind of idiot scrambles to install a major OS upgrade on the first day of its release? If we've corporately learned anything, it's "Don't be the first to test run a new operating system release in the wild".

[bd139]

On top of that macOS now calls home to check binary signing on app startup. The cluster that handled this fell over yesterday and locked up huge numbers of people’s kit entirely. This drew attention to the fact that it can be used for surveillance.

This needs to be tempered with a little bit of detail. It's just an OCSP certificate check. Most people reading this whatever their browser and operating system are suffering from the same security leak every time they visit a website using TLS if that website isn't serving a stapled OCSP certificate along with the standard X.509 certificate. The difference is that the OCSP check is going to go to the issuer of the site's X.509 certificate, and it's them that get the opportunity to surveil you should they choose to. In this case the OCSP check is being done as part of checking the signature of a binary, and Apple have their own PKI so the OCSP request goes to Apple servers. MacOS, like most others, does some OCSP certificate caching so the apparent 'phone home' only happens when cached certificates expire or are first seen.

Yes, it looks like someone at Apple didn't account for the huge number of OCSP requests that would be caused by a whole bunch of people installing Big Sur* and starting off with empty OCSP caches and so, for a while, Apple's PKI servers choked on the load. It's not like they are the only large scale provider to get their sums wrong and stuff it up.

Correct but I think it has some ramifications past that which are what need to be considered when you look at the fact that it doesn't actually land on Apple's servers but Akamai's CDN (massively untrusted company) and is apparently transmitted in plain text (I haven't got it on wireshark yet). Also if the network is not available it runs any old binary which is not exactly a winner for the security posture. Thus I can only conclude this is used to revoke developer certs which is, well, basically a vendor trust issue. On top of that with Big Sur, some system applications are given network privileges above the user (and Little Snitch / pf) including OCSP validation. Hmm beacon possibility?

From a security perspective, my conclusion is that Bob's a imperialist wanker, Alice is making baaa'ing sounds while humping the device. The only winner is Mallory who can sit there and collect a lot of who, where and when.

* What kind of idiot scrambles to install a major OS upgrade on the first day of its release? If we've corporately learned anything, it's "Don't be the first to test run a new operating system release in the wild".

Well and interesting one there. A corporate engine forced this one. Someone sets a policy of "your operating system can be only N days out of date". Eventually everyone who isn't regularly updating gets fed up of the SSO solution refusing to let them in unless the node is updated. So they do. Eventually this just turns into a mindless process of "log in, oh look an update which I'm not going to be able to do anything until I upgrade".  This is of course faulty but so are most humans and processes.

[ConKbot]

Dont blindly buy into any  "x nm process" hype blindly.


From the derbauer video series where he took some processors in to a place with  FIB/SEM to get cut them up and actually look at the transistors themselves.



Proof in the pudding is in the eating, whatever mips/MHz/watt capability is demonstrated on a process, that is the thing to be interested in. Otherwise its just a process name that happens to have a nebulous number attached to it.

[bd139]

Spot on. The important bit is not the size but the efficiency gain of the process. It just happens to be labelled with a size.

[Cerebus]

Correct but I think it has some ramifications past that which are what need to be considered when you look at the fact that it doesn't actually land on Apple's servers but Akamai's CDN (massively untrusted company) and is apparently transmitted in plain text (I haven't got it on wireshark yet).

Not sure about that, and here's why. Although Apple do rely on Akamai for a CDN they also have their own which isn't exactly visible and is only, afaik, going to affect their mobile users. Apple have a lot of kit colocated with mobile phone providers to serve a lot of the iPhone<->Apple traffic. What it looks like in standard IP terms may give a misleading picture, and you can only see the full truth if you have access to those networks at the SDN control plane level. This isn't generally well known, not a secret, but not generally well known. I know about it because I saw the infrastructure plan in one huge mobile provider's Confluence while I was consulting there.

[bd139]

Yeah aware of that. Some if it was even on Azure a few months back . Apple own 17.0.0.0/8. You can BYOIP to CDN/cloud providers hence why we bought a /24 recently.  YMMV but occasionally stuff in there is referred to edgesuite.net due to misconfiguration (fucking CloudKit wank shit)

[Cerebus]

Some if it was even on Azure a few months back .

That may even be the beast of a telco that I'm talking about; when I last left them they were gearing up for a poorly planned and rushed network extension that would have brought Azure on-net for them and virtually pushed part of the local network out onto Azure via a horribly mis-thought-out addition of yet another overlay network onto their international backbone. Basically they were going to stretch their VMWare SDWAN o-u-t to commingle with Azure. What could possibly go wrong?