The ARM chip race is getting wild… Apple M4 unveiled 

cause hes me

That clip is definitely ai generated

Its AI, look at the fingers

AI generated 😂

He thinks diffrent

Tenor, Giphy and more. Probably easy to find if you just search their expression on Google

RU-vid

meta-llama

AI is cheaper than stock footage.

And who decided to make stock footage of a couple of guys in suits fighting each other with gardening tools?

expectation : America and the gang vs Russia and the gang reality : American vs another American while Taiwan supporting everyone

The hot wars

It's like that translator joke with Les armes

Apple has basically had the lead for so many years, i doubt they will lose it

@@Lenfer-hp3ic taiwan is literally just 80 miles from china, so without america taiwan is f'kd in few hours.

Our new racing horse can achieve speeds of up to Mach 4!** **if it had legs and could actually survive running at those speeds

38 TOPS, ray-tracing, but still can't run macOS fucking lol

@@civilroman still don't have file manager with file system access!

What even are you gonna do with all that power? Makeshift oven? AI generate all your art?

That cracked me up real good.

It's not an ARMs race, it's just hotwheels with a hot charge, USB 4.XXL

it wasnt

@@Moli05 dude, Apple is nothing, ever since jobs died. Now it's just another brand, take it easy. 🤡🤡

@@Moli05 Apparently you’ve never seen an apple magic mouse

😂

😢

It's the future

Honestly, I would buy it just because of the name (I'm broke)

woah there buddy you forgot the X

I'm waiting for the Super V2

nice

no pain no game reversed

Deep , definitely an interesting topic

I had quasedillas + buritto styled salad and I am currently in the same proccess you were in

Are you lactose intolerant or something? I don't get people who say that tacos make them poop

Those two are irrelevant

you can, what you smoke?

..,,::;;!!!????​@@gigamoment

@@weiss588 shhhh, he's trying to make the lulz.

I can run an llm on my s20 fe, that isn't special, people have ran llms on hacked switches.

now i really wanna see a rifle made by Apple

@@planetrenox featuring the thinnest scope bezels apple has ever made!

it's not far fetched, samsung make fridges, tanks etc

its an M4 rifle pointed at the snapdragon

I'm waiting until the Apple M16 comes out.

How bad is it?

@@PapiDey-dv4gw it's bad out here man

@@PapiDey-dv4gw IpadOS is like a nerfed MacOS. It's better than iOS but worse than MacOS. Apple refuses to put MacOS on it, I guess because they would have to kill the iPadOS. And the lack of supported apps makes you asks the question, why not buy a MacBook instead.

​@@wlockuz4467 My guess is MacOS is not on iPad because MacOS allows apps outside of App Store, so there's less revenue.

​@@wlockuz4467 they don't put MacOS because people wouldn't buy MacBooks anymore...

should've stayed out of the road

@@iverssonkaufert7619 lmao. looks like the car got him again.

timestamp?

​@@subhashgottumukkala 0:11

is going to have 4 GB of unified memory and each 4GB is going to cost $400.

@@segiraldovi Sounds like a deal to meeeee (this message is sponsored by apple)

I already prepreordered mine.

But I already sold my kidneys and liver

Sounds like super soaker water

because AI gets better

Because he's using AI

He did previously mention that his AI sounds more human than him, so...

😅😅😅😅😅😅😅

AI's getting better :)

I mean it's quite obvious that Nvidia's chips are go to for AI related stuff, but it's still impressive what performance M chips can achieve with how little power they need. Sadly, they are OS exclusive.

@@mikadofxx9030 They aren't. Apple has specifically made sure that third-party OSs can run and can do so safely and without sabotaging the system's security model The drivers and some other important bits of code being private is of course a big obstacle, but it's one that can be and defeated, as the Asahi Linux team has proven.

To add, Apple themselves encouraged Microsoft to port Windows to Apple silicon... why? the answer is that macOS makes them practically no money. Apple is a hardware company.

@@U20E0, sources? Note, links may not work. So, just name the sources.

@@CarlBach-ol9zb Apple's documentation on the boot process as well as documents from the Asahi Linux project

Wrong! It’s exactly this CISC decoder unit which consumes a lot of energy. Furthermore there was the famous IA64 experiment which failed. Even further seen you have to distinguish between software which runs for companies versus for private consumers. They have different development life cycles and budgets. There is a reason why COBOL exists till today. I don’t know of any large migration project after Y2K. Usually a software stalls in further development if its code is 1) unreadable 2) test coverage is bad of low.

​@@javastream5015the decoder is only about 5% of die area in modern CPU cores. Most transistors and energy are spent on branch prediction, handling out of order execution and cache.

​@@javastream5015OK, I get the point about CISC to RiSC adding to overhead, but why bring up Itanium? Yeah, it failed. But so did all the other times Intel tried to "replace" their x86 based cores. From what I understand x86s is about removing/depreciating old 16 and 32 bit instructions. In other words making x86 more focused on x64 and beyond. That is not an attempt to replace the current architecture, just a revision.

@@MichaelDeHaven Back to x86s: Yes, it can be done, because Apple did the same with the M1. Apple removed the support for 32 Bit API even one OS version before. 👉A lot of old games (like Civilization 4) stopped working. To run old software emulators are needed. Then you have stupid companies like Parallels who didn’t want to build a x86 emulator into Parallels Desktop so that old x86 (and x64) virtual machines could be used further. ☹️However, as I have read now, Windows for ARM supports emulation of 32 Bit x86 software (on 64 Bit ARM). 👍 👉Long story short: It could work technically. But you would make a lot of owners of legacy applications angry without a proper emulation support. Just look what Apple did.

@@MichaelDeHaven x86S isn't really about deprecating instructions (as that could be devastating for compatibility for older software - Intel's proposal would require changes only from the operating systems), but about deprecating unused features. For example, x86 chips first boot in 16bit mode, then they can be switched to 32bit and then finally to 64bit adding a lot of unnecessary complexity. There's also the entire real and protected mode ordeal, together with unused executions rings 1 and 2 and finally some things about the segmentation model. As for the entire RISC and CISC problem on x86, in my opinion, the decoder and all the execution units (APU, FPU,...) still have to be prepared to deal with a lot of instructions. x86 contains more than 186 opcodes for just the ADD instruction, having the option to provide an immediate value or address from memory, each of those for 8, 16, 32 and 64bit values. Adding a MOV instruction just before that and supporting only the registers would have caused a huge reduction in the opcode count, but it's a little too late for that. It shows, that x86 was designed to be programmed directly in assembly and to offload more complexity to the hardware rather than the software, which is just not how we use computers today with many abstractions above the hardware. Intel's official documentation to x86 in its current form has exactly 5060 Portrait Letter pages full of really well constructed and written text (one of the best documentations I've ever read) and yet it's still incredibly big. Sometimes less is more.

They're not here because they're having issues compiling their web browser for RISC-V

well at least there won't be monopoly in this arms race

@@TheDanielLivingston I laughed entirely too much at your comment

Yeah Apple’s gotta stop

O B S O L E T E 😂

You could throw it in my bin though

@@veenmikki27 Nah this is the fault of the person throwing it out From my experience old Macs (and almost every higher end computer and phone for that part) work perfectly fine multiple years into the future

​​@@IDKisReal2401True. I would have been watching on this my 2017 MBP if it weren't for the fact that I accidently dropped and broke it a few weeks ago. That's considering that those models criticised for keyboard, battery, and fan problems. Still sucks a bit that they announced M4 when I just bought a M3 MBA as a replacement.

Some old phones were built like that. I have one at home

Remember time, when Samsung phones had charger port and jack from the top? Because I do.

@@Favour.A.Emmason-pv1mk yeah but where is the camera and why are the speakers on top then lol

it's ai generated stock videos

​@@Favour.A.Emmason-pv1mk i think i would prefer that

"x86" CPUs have not been "x86" internally for a long time, it's just a layer on the outside. You should listen to CPU architect Jim Keller who says that x86 is not really a substantial disadvantage. Case in point, AMD is goint to release Zen5 based StrixPoint and StrixPoint Halo, which to me seem targeted at Apple's "M" SOCs.

@@swdev245yeah it’s not the architecture, it’s just Intel and AMDs disregard of mobile chips. Now they’re scrambling to deliver much more efficient chips with Strix and Arrow Lake

Was it an AI generated comment?

@@BernardoLeon yeah looks like it, click on his profile and see his comment history

@@BernardoLeon Which comment do you mean? I wrote mine with natural stupidity ;)

This. Everyone forgot the first windows tablet (Windows XP tablet PC edition anyone?), and shat bricks when the iPad launched. Same with arm (Windows RT anyone?), no one wanted to hop on the arm train back then. Some tech needs apple’s hypetrain.

I have a windows on snapdragon laptop and its great. I do work for Qualcomm as a software engineer though, so I may be biased.

Common Microsoft L

i like your profile picture

@@TheDeveloperGuy I heard that Windows RT sucked ass and maybe that's why nobody really wanted to jump ships. It also had no x86 compatibility layer (which the existing software library is the main selling point of Windows these days anyway). Without Rosetta functioning as well as it did, I don't think the jump would have been such a no-brainer for many people as it was.

underrated comment

bump

Also, it kept running without the power supply for a while.

@@brunesifirst arm processor on showcase to bbc was even unplugged to electricity. It was enough of power from returned electricity, from connected monitor to motherboard, to run it😂

@@brunesi not without power supply. But it was able to run from input diode pin leakage current. Still impressive though.

~flashbacks of farming gazers with tamed dragons...or of stealing peoples shit from their bags with a thief. Or fighting them, disarming them and then stealing the weapon from their bag. Still one of the funniest things I remember doing in a game.

Vas flam

@@hmthatsniceiguess2828 Or being chased by people wanting to kill you only to provoke nearby dragons and wisps to attack and chase them. It really was a great game

@@johannesandersson9477_”Cor Por”_ The standard greeting UO before 99% of the population moved to Trammel.

im just here for the UO dragons

I was hoping someone said something. Genius!

I didn't get it, what was the joke? The app?

@@lucasjritter I don't think there was one. Just the fact that Grindr exists is hilarious apparently 🤔

@@BrysonThill gay people are really funny. Thats probably why theyre so happy

@@lucasjritter The very fact that Grinder was the featured app was the joke. As a gay man, this joke gave me all the good feels.

I had to check that figure. Amazing! It only takes an electron 3 billionths of a second to do one orbit. How can they do 10,000 operations in the time it takes an electron to do an orbit?

not qualcomm, but windows with their shit os

@@weiss588 I dont care. If not windows what will you do with the CPU? Either they do it with windows or at least with android and Linux.

there is always linux to save the day, but windows have the corporate user OS so i think there is no going to be a real diference.

@@omarjimenezromero3463 True. Corporate users dont care about ARM chips tho. All they need is performance not efficiency.

Ah yes, making bitcoin worthless, fun.

Why llms per second just go with nano like the other one or go even lower, picosecond

All that computing power and the first thing you think about is bitcoin

​@@ctrl_x1770 the first thing he thinks about is probably milf videos

Haha love this comment 😂

yea, seeing that was a blast from the past!

That's correct. Most people don't understand this. Qualcomm's 45 TOPs number is based on INT4 precision. That's not considered accurate enough for most use cases. Apple's numbers for the M4 are based on INT8. In the past, they've used numbers based on FP16. The point being, TOPS is a meaningless metric unless you know exactly what you are comparing. Either way, Qualcomm's X Elite's NPU is clearly LESS powerful than the M4's NPU.

Are the newer ones released already or you using a 2023 model?

Seems Google takes the middle ground and submits it to the cloud, while taking a little longer, it comes back on the first try as well...

I was going to say, “now if only software would evolve so much”, but with AI like these, who needs software?

fr fr

@@TheScottShepard isn"t AI itself a software?

yea seems like we haven't hit the peak of the curve yet 😀

Project Lavender, just another advancement we can all learn from.

A real OG 🫡

Glad that someone mentioned risc-v lol

Buhu dude The ipad is a great device. Try using it once, you will see

​@@annoswet1576while the iPad may be a good device apple is the worse and most scum company on the planet so no doubt it was used on the iPad because their M4 isn't as good as they say it is as usual

@@annoswet1576no 😂

Most old instructions unsued intructions are just emulated in microcode, nobody will notice that some software written for i486 30 years ago is running 10x slower. They only add complexity on the decoder, wich uses only about 5% of a core's transistors anyway.

No they not, x86 problems has nothing to do with old instruction set or backward compatibility. So it would be cool if you stop spreading lies because you are clearly incompetent on the matter.

​@@randomrandom7208he isn't wrong though

@@randomrandom7208 r u telling tat to urself

he did

says every bottom ever

Really tops those charts I’ll see myself out

She ghz on my tops till' i M4

I wish I was topsed

GHz development stopped 15-20 years ago.

If only applicable to NPU operations per second then ok meh...

Ha! But you are not going to tell us 🤣😂

which ones?

@@chiluco2000 On my first watch and I caught these errors. (1) The article quoted is that claims x86 is dead is 90% wrong only the last paragraph is even remotely correct. (2) The structures x86 is not based on transistor count but the number of instructions the cisc chip can use to perform its micro operations. (3) Arm was poorly described in the video, he based the argument on instruction sets(Risc) and then went to describe it by transistor count. This is not really the reason why arm is considered better for mobile devices and devices with simple specialized functionality,

maybe speed difference

privacy matters, like govts fine tuning LLMs on top secret military data, you get the point

because you don't want to send your private and company data to FAANG and NSA ?

Because AI is a lot more than ChatGPT. For example, removing unwanted objects from pictures is a relatively "mainstream" AI task. There are services that do it online, but most require a subscription or a purchase of credits. Being able to run AI locally means you don't have to pay for things like that.

Speed, cost, privacy. For simple tasks a high-performing fine-tuned small model running locally will be sufficient. Apple developing both their own hardware and software can achieve this very effectively.

fish here

ok?

@@lucassilvas1 😂

@@soejrd24978 made my day 😭

👍

:(

" Hell, It's 'bout time. "

If the 100+ TOFS is true, then I don't think so

We wouldn't need arm SoCs if intel made good energy efficient SoCs

Bu... but is has AI!!! DON'T YOU LIKE AI!? 😂

Wow, just looked it up and it really doesn't have av1 encode... Kinda wack.

@@Velocifero Yeah the M2 not having hardware decode for AV1 was an L, the M3 not having AV1 encode was an even a bigger L but the M4 STILL not having AV1 encode? What is Apple even doing anymore

for studio users the M2 Ultra was 31 tops + nobody cared or used it. M4 ultra can be anything, because it will get its own die, people say, but if it is as before 2x, then look for 76 TOPS. The key will be, if apple has it doing anything useful. Frankly, I doubt it will be critical on Mac OS, but suspect they will on iPhone OS. Because that is where some of the opportunities are obvious, like replacing Siri.

Unless you run your 4090 in the cloud to access its compute power remotely your point is invalid.

also, afaik there's no API to interact with it so other than apple no one else that can run code on it.

Now compare power consumption.

it also uses 660 watts of power geg

@@2pingu937 Cloud GPUs like the H200 can to 4000 TFLOPS of FP8

I’m using one right now

Why? What should run on those ARM chips? Compare them with the support for M1 and you can see the winner quickly.

Based on? If you don’t mind explaining your take