Apples new M1 microprocessor

[Halcyon]

If you don't want to believe you never will

Meanwhile, here's an actual professional cryptographer discussing this exact flaw. A flaw that literal nobody like me expected to exist, obviously. This may give you an idea of how little prestige there is to earn in "researching" it. And of course Apple themselves would never brag about ability to spy on their customers for obvious reasons.
https://www.schneier.com/blog/archives/2015/08/nicholas_weaver_1.html

Interesting article. Unfortunately, you won't find too many details out in the wild. Those in the know are bound by various NDAs and subject to their own Government's secrecy laws.

The subject of warrants is interesting. There are various mechanisms to extract data from companies (not just Apple) and they don't necessarily need a warrant from a court to enact. Most of these cloud companies currently have a Mutual Legal Assistance Treaty (MLAT) in place to enable a Government agency in one country to seek information from a company operating in another country. These are about to be strengthened through new legislation being approved in several countries. Google has a whole help section on how these things work in a broad sense: https://support.google.com/transparencyreport/answer/9713961?hl=en

[Cerebus]

Subject for this topic: "Apples new M1 microprocessor"

Most of the content: Nothing to do with the processor, just the usual quibbling about how bad/good Apple are from the usual suspects with their usual prejudices on the subject.


For those quibbling over whether iMessage is secure, it's not particularly secure (depending on your threat model*), but it is much more secure than the thing it sought to replace - SMS. Anybody who wants actual message security, again depending on one's threat model, uses Signal.

*The giveaway that you're in a poorly informed discussion of security is the failure to define a threat model before making an unqualified assertion that something is 'secure' or 'insecure'.

I will say this from an actual security researchers point of view: Apple devices are among the easiest to extract data from. I'm not just talking about iMessage, but the entire contents of the device.

Belying your prejudices again. The average person's PC - still a device - requires a crosshead screwdriver and something that can read a hard disk or SSD to extract all the data from it, the average iPhone user's data is much better protected.

[Halcyon]

I will say this from an actual security researchers point of view: Apple devices are among the easiest to extract data from. I'm not just talking about iMessage, but the entire contents of the device.

Belying your prejudices again. The average person's PC - still a device - requires a crosshead screwdriver and something that can read a hard disk or SSD to extract all the data from it, the average iPhone user's data is much better protected.

You're asserting that I have been prejudicial in the past. How about sticking with the facts?

Feel free to re-read what I wrote. I did say from the outset that it was unlikely to be a significant issue for most users. BUT, it's an issue that might worry some users. Just because you don't care, doesn't mean other people share the same opinion.
Your argument is the equivalent of a kid getting into trouble then blaming it on someone else because they also did something wrong. It's not a tit-for-tat competition.

The other point I'll make is what might not be possible today, could be possible tomorrow. It wasn't long ago that researchers and experts in the field had to send phones away overseas at great expense to have them unlocked. Now they can do it themselves with a laptop and an interface box. You're also forgetting that many low-end PCs and laptops now come with self-encrypting SSDs and/or TPM devices. The days of simply pulling a hard disk out and reading the data off are dwindling.

[brucehoult]

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 isn't using ARM cores. It's using specially designed *Apple* cores that just happen to use the same Aarch64 instruction set as ARM cores. Apple could have used the POWER9 instruction set or the MIPS instruction set or the RISC-V instruction set and the performance of the same core design would be much the same.

[Berni]

Apple isn't using ARM cores. It's using specially designed *Apple* cores that just happen to use the same Aarch64 instruction set as ARM cores. Apple could have used the POWER9 instruction set or the MIPS instruction set or the RISC-V instruction set and the performance of the same core design would be much the same.

Well yes they might not be just taking ARMs core design and directly slapping it onto a chunk of silicon, but the design goals of the design are very similar. Its running the same RISC flavored ARM instruction set and trying to do it with low transistor count and power efficiency so one might expect to see performance somewhere within the same ballpark. Its running ARM code so it is an "ARM processor". Might even be the fastest ARM processor in the world (and i would certainly find that an impressive feat of engineering) but i find it very unlikely that such a core would suddenly surpass the performance of x86 cores that have been constantly developed and optimized for speed over the last 25 years. Right now these x86 are the fastest cores so being the "worlds fastest CPU core" would require  being faster than those.

It is debatable if the x86 instruction set is actually better than ARM. The CISC approach of x86 certainly made sense in the past, maybe it doesn't make so much sense these days, especially with all the legacy baggage it is dragging behind (I don't design CPUs so i wouldn't know if CISC still helps any, but the x86 set looks pretty messy by now)

[bd139]

If you have a look at the output of ARM compilers versus x86 there are some clear advantages to ARM in that space. I’ve seen up to 50% less instructions per task being executed on ARM due to the conditionals and orthogonality. Per clock they are more efficient execution wiser, power wise and scalability wise.

The only issue is no one has made a 12 core desktop class ARM CPU with 4.7GHz peak clock that runs windows. Yet. It’s is possible.

If someone started selling socketed ARM cores, chipsets and boards then x86 would be dead overnight. Apple doesn’t cut it as its toaster appliances so far.

[brucehoult]

Apple isn't using ARM cores. It's using specially designed *Apple* cores that just happen to use the same Aarch64 instruction set as ARM cores. Apple could have used the POWER9 instruction set or the MIPS instruction set or the RISC-V instruction set and the performance of the same core design would be much the same.

Well yes they might not be just taking ARMs core design and directly slapping it onto a chunk of silicon

Not only did Apple not take ARM's core design and slap it onto a chunk of silicon, they also didn't take ARM's core design and tweak it a little

ARM published the ARMv8-A instruction set specification in October 2011. The first things shipping with ARM-designed cores were in March 2015.

Meanwhile, Apple shipped an ARMx8-A core entirely of their own design in the iPhone 5s in September 2013, a full 1 1/2 years earlier.

It's not physically possible for Apple to have modified one of ARM's cores. Apple were first, by ages.

[Berni]

Apple isn't using ARM cores. It's using specially designed *Apple* cores that just happen to use the same Aarch64 instruction set as ARM cores. Apple could have used the POWER9 instruction set or the MIPS instruction set or the RISC-V instruction set and the performance of the same core design would be much the same.

Well yes they might not be just taking ARMs core design and directly slapping it onto a chunk of silicon

Not only did Apple not take ARM's core design and slap it onto a chunk of silicon, they also didn't take ARM's core design and tweak it a little

ARM published the ARMv8-A instruction set specification in October 2011. The first things shipping with ARM-designed cores were in March 2015.

Meanwhile, Apple shipped an ARMx8-A core entirely of their own design in the iPhone 5s in September 2013, a full 1 1/2 years earlier.

It's not physically possible for Apple to have modified one of ARM's cores. Apple were first, by ages.

I never said that Apple just used the stock ARM design like a lot of other vendors. What i was saying that the design goals for there chip are very similar to the design goals of the stock ARM designs. Both the "higher end stock ARM cores" and "Apple homecooked core" are going for getting a good deal of performance out of a low power, not so high transistor count core design while running the ARM instruction set in the low GHz figures. I'm calling it an ARM processor because it runs the ARM instruction set, not because it would be an designed by ARM chip. Yet still, im sure the stock ARM designs still have a good deal of room for improvement, but in the push for performance the guys at ARM would have already taken all the easy optimization tricks. So Apples chip is probably somewhere near the ballpark unless they came up with some really revolutionary cpu design trick.

Getting the core to run at 5GHz is also part of the cpu design and optimization, long windy convoluted pipelines are the penalty to pay for that. Just going to a fancy new fab process node doesn't just magically it run at in high GHz digits. Tho x86 is these days mostly limited by how much power can be pumped in without the thing melting.

[bd139]

Yeah 5GHz things get interesting. It's really not trivial getting anything to even work up that end of the scale. If you consider the typical pulse can barely transition a the CPU die up there then things get even more hairy in the physics and design department. I think we're going to hit a wall. The 40% or so gain from ARM won't buy us much in that space. The computer scientists will however. I get the feeling that migrating processes and algorithms to many-core systems is going to be the winner and that's ARM's domain which is showing core scalability to 100+ cores whereas x86 seems to drop out around 50 or so at the moment.

[madires]

Since the x86 architecture isn't great anyway and it carries tons of legacy stuff, it's not hard to design something more efficient, smaller and more powerful.

[bd139]

Precisely. It's a 1970s design which has been hacked and hacked and hacked.

ARM however is a 1980s design which has been hacked and hacked and hacked.

Humans at their best

[tszaboo]

I estimated few years ago, that ARM could reach x86 performance and surpass it. I dont think it is quite there yet, but I dont think raw computing performance actually matters. What matters for most users, is how smooth scrolling is, and how much time it takes to start an application, or how fast animations are. That has little to do with the CPU performance, and these things are already faster on a ipad than on a low end laptop.
Apple was in the perfect position to change to ARM, and they had a huge incentive to do this. Majority of their income is coming from the appstore, selling third party applications, that require little to no investment on their side. They are building the same walled garden with their macs now, and they expect similar income. In fact probably the only reason that they are not selling their phones and macs for a nominal fee, because they want to appear premium, and that is easier, if you sell your device for 999 than for 500.

And one thing to get straight: Intel's nodes are better than TSMC's or Samsungs interms of transistor density. Meaning that they have better densities on 10 nm, than others on 7, they are just measuring it differently.

[borjam]

Precisely. It's a 1970s design which has been hacked and hacked and hacked.

ARM however is a 1980s design which has been hacked and hacked and hacked.

Well, maybe it was better designed in the first place and not really “hacked”.

I remember Sun’s SBus, designed in the 80’s. They got it right and later they released a 64 bit version of SBus while the peecee wold had been struggling with appalling hacks such as ISA/EISA/VLBus/AGP until, finally, PCI came out.

When was 64 bit SBus discontinued?

[bd139]

Oh yes ARM was definitely better designed in the first place.

SBus/MBus were rather nice. I miss those machines, particularly my Ultra 2 desktop I had and the 1000E I stupidly used as a desktop for a bit. PCI came in with Ultra 30 if I remember. Hell I even miss SPARC. RIP

[Berni]

Yep i do agree x86 will get buried at some point, but it will take some time to gradually move everything away to something else, most likely ARM or RISC-V or something along those lines. Perhaps something that also includes a bit of hardware acceleration for x86 emulation because its hard to just throw away so much software.

But yes good point that CPUs have gotten so fast these days that they are more than fast enough for everyday tasks. So the average mac user probably wouldn't be bothered even if these laptops are suddenly 50% slower according to benchmarks. The facebook page would still load up just as fast, a spreadsheet opens just as fast and they can listen to music in iTunes just as well. So yeah i think its mostly about creating the same iPhone walled garden on MacOS to pocket the 30% of all mac app sales and conveniently also flipping intel the bird as a bonus side effect. The users wouldn't be bothered since they are used to this app store kind of thing, but the app developers would certainly be bothered.

Microsoft was trying to pull the same thing with the Microsoft Store thing on Win 10, but nobody liked it cause that's not how PC users want to get apps, while they could not just make win10 not run classical win32 apps as then nobody would want to use there OS anymore.

[borjam]

Oh yes ARM was definitely better designed in the first place.

SBus/MBus were rather nice. I miss those machines, particularly my Ultra 2 desktop I had and the 1000E I stupidly used as a desktop for a bit. PCI came in with Ultra 30 if I remember. Hell I even miss SPARC. RIP

I still have an Enterprise 450 doing its duty!

(Although it’s a heretic Sun using PCI)

[tszaboo]

Yep i do agree x86 will get buried at some point, but it will take some time to gradually move everything away to something else, most likely ARM or RISC-V or something along those lines. Perhaps something that also includes a bit of hardware acceleration for x86 emulation because its hard to just throw away so much software.

Probably not, and that's fine. If you look at computing, software is a lot more expensive than hardware. There are a lot of legacy applications that you need to use, there are parts of the Windows system, that nobody touched in the last 10 years. They dont want to rewrite it, and that's fine. If you look at what Intel is doing, they started to produce cores with different architectures. In the past, every CPU core had the same instruction set, the same extensions, the same amount of cache, otherwise the operating system gets confused. Now they started adding Atom cores coupled with high performance cores. Once the OS can handle it, you can do whatever you want. An ATOM core is what, 50 million transistors? The new nVidia GPUs are 50 Billion. The same way you left that ancient x86 core in your CPU, the same way they just going to leave that in, couple cores with a new instruction set, and leave it to the compilers.

[SilverSolder]

Yep i do agree x86 will get buried at some point, but it will take some time to gradually move everything away to something else, most likely ARM or RISC-V or something along those lines. Perhaps something that also includes a bit of hardware acceleration for x86 emulation because its hard to just throw away so much software.

Probably not, and that's fine. If you look at computing, software is a lot more expensive than hardware. There are a lot of legacy applications that you need to use, there are parts of the Windows system, that nobody touched in the last 10 years. They dont want to rewrite it, and that's fine. If you look at what Intel is doing, they started to produce cores with different architectures. In the past, every CPU core had the same instruction set, the same extensions, the same amount of cache, otherwise the operating system gets confused. Now they started adding Atom cores coupled with high performance cores. Once the OS can handle it, you can do whatever you want. An ATOM core is what, 50 million transistors? The new nVidia GPUs are 50 Billion. The same way you left that ancient x86 core in your CPU, the same way they just going to leave that in, couple cores with a new instruction set, and leave it to the compilers.

There have been attempts to get people off x86 code in the past (e.g. Itanium?) but these things flopped in the market for the reason you say -  people's existing expensive code didn't work...

so then the question becomes: what works better - a native x86 core or an emulation running on a different architecture?   My money would be on the native core every time, all else being equal.

[borjam]

There have been attempts to get people off x86 code in the past (e.g. Itanium?) but these things flopped in the market for the reason you say -  people's existing expensive code didn't work...

It already happened. Which mobile devices use Intel CPUs?

[bd139]

so then the question becomes: what works better - a native x86 core or an emulation running on a different architecture?   My money would be on the native core every time, all else being equal.

Turns out transpiling it from x86 to ARM actually might end up faster 

https://wccftech.com/m1-macbook-air-x86-emulation-faster-all-mac-models/

Note that includes i9 and Xeon CPUs that are shipping.

[tszaboo]

so then the question becomes: what works better - a native x86 core or an emulation running on a different architecture?   My money would be on the native core every time, all else being equal.

Turns out transpiling it from x86 to ARM actually might end up faster 

https://wccftech.com/m1-macbook-air-x86-emulation-faster-all-mac-models/

Note that includes i9 and Xeon CPUs that are shipping.

Even Microsoft had x86 for 2 years now, but they dont have x86-64. Does apple have that? What about AVX? What about the doens of other instruction sets that were bolted on?

[bd139]

Apple only does x86-64.

Details here: https://developer.apple.com/documentation/apple_silicon/about_the_rosetta_translation_environment

No AVX, AVX2, AVX512. They are only one, potentially poor if it's AVX512, way to skin a cat. See: https://extensa.tech/blog/avx-throttling-part1/

Again, burst performance is good but anything persistent, narp. The same may be true for the M1 ARM CPUs. Let's see what happens. Battles coming for the next year or so on this front I reckon.

[Cerebus]

I will say this from an actual security researchers point of view: Apple devices are among the easiest to extract data from. I'm not just talking about iMessage, but the entire contents of the device.

Belying your prejudices again. The average person's PC - still a device - requires a crosshead screwdriver and something that can read a hard disk or SSD to extract all the data from it, the average iPhone user's data is much better protected.

You're asserting that I have been prejudicial in the past. How about sticking with the facts?

Feel free to re-read what I wrote. I did say from the outset that it was unlikely to be a significant issue for most users. BUT, it's an issue that might worry some users. Just because you don't care, doesn't mean other people share the same opinion.
Your argument is the equivalent of a kid getting into trouble then blaming it on someone else because they also did something wrong. It's not a tit-for-tat competition.

The other point I'll make is what might not be possible today, could be possible tomorrow. It wasn't long ago that researchers and experts in the field had to send phones away overseas at great expense to have them unlocked. Now they can do it themselves with a laptop and an interface box. You're also forgetting that many low-end PCs and laptops now come with self-encrypting SSDs and/or TPM devices. The days of simply pulling a hard disk out and reading the data off are dwindling.

Well, I've noticed you taking a rather biased approach to Apple in the past, so I think my assertion is justified particularly as it shows here. Your assertion that "Apple devices are among the easiest [of all devices]  to extract data from" seems to add evidence to that, as you make a sweeping statement that isn't supported by the facts or by presenting any evidence. e.g. That the average PC has a data extraction barrier of near zero, much lower than, for example, an iPhone. You don't address the one point I made, wandering off to address points I didn't make, which to me at least is suggestive of dodging the point.

Anyway, I seem to have achieved my objective of getting the thread back on track, so let's not flog the dead donkey eh?

[SilverSolder]

There have been attempts to get people off x86 code in the past (e.g. Itanium?) but these things flopped in the market for the reason you say -  people's existing expensive code didn't work...

It already happened. Which mobile devices use Intel CPUs?

Which mobile devices use or run x86 code?   x86 is pretty much a desktop thing, right?

[SilverSolder]

so then the question becomes: what works better - a native x86 core or an emulation running on a different architecture?   My money would be on the native core every time, all else being equal.

Turns out transpiling it from x86 to ARM actually might end up faster 

https://wccftech.com/m1-macbook-air-x86-emulation-faster-all-mac-models/

Note that includes i9 and Xeon CPUs that are shipping.

In that case, all else is not equal!   An emulation running on a brilliant processor will beat a slow native processor...    but if we are talking about two equally brilliant processors...  ?    My money is on native!