[Music] Hey guys, welcome to the show! Today we're going to be checking out the M4 Max

MacBook Pro. Over here, I've got the M1 Max with 64 GB of RAM; the M4 Max Pro with

64 GB of RAM. We're going to have a bit of a gig off. We're going to be doing

this Llama; we're doing Unreal Engine, we're going to be doing Xcode, Android, crypto; we're going to do some mining; we're going

to do all sorts of things, even unzipping an Excel spreadsheet. How good is it? That's what we're going to find out—how good

this bad boy is for development and all the kind of fun stuff you want to do on these machines. So first off, I've

got this Llama, the 70 billion parameter model—just enough to fit into the 64 GB of memory—and we're going

to be just talking to it. So let's type in 'hello' and let's just see how much faster this bad boy is for figuring

stuff out. Boom! Run at the same time and let's see... On the left, thinking, thinking, thinking... Boom! It's started talking back

to me. The M1 Max hasn't even thought about talking back to me; it's pretty much finished, and the M1 Max is far behind. So

that is a considerable difference. I've already done some tests over here; I've already found that it's up to three

times faster. Most of the time you're not going to notice any difference whatsoever in multiple things you're going

to do, except for some cases it's going to be three times faster. So it's not 20%

faster, 40% faster; it's three times faster—300% faster in some cases.

But I'm going to show you that very soon. So let's just ask another fundamental question about the world, see if you can

confuse it 1 2 3 Let's Go Boom! Straight, we got an answer; we got an answer here.

It's spitting out text. A lot!  F-F on the M4 Max. No fans, but actually something

noticeable. It's coil whine; there's a bit of a zinging sound coming out of this. So, side-by-side, there there is a

bit of coil whine coming out of the M4 Max. It's very, very hard to hear it; it's just

a lot more high-pitched than the one coming out of the M1 Max. M1 Max, like...

the M4 Max is kind of like...obviously, that's extrapolated. You've got to be your ear up to here to notice

it when it's doing this kind of process. But let's go ahead and switch up the show. Let's exit out of this guy and

let's do some image processing. So we're going to be using Llama 3.2 Vision. We're

going to try figuring out who this guy is. Let's see if you can know who this guy is and what the Earth is. So let's

run that show. 3, 2, 1... Who does it faster? It's taken quite a bit, but boom!

We've got a response on the M4 Max. M1 Max is still figuring it out. These tests are variable; there's a bit of randomness

when it comes to giving out answers, but it's spit out a response. It says, "A small American flag pin adorns the background, out of

focus. An American flag hangs on the wall." Very descriptive in describing the image we're about to see, and it's got the

second one is done. M1 Max, a bit behind there. I'm guessing is that...is that more than twice—twice as fast than the M1 Max

on this image recognition situation? M1 Max is still processing; for the first, figuring out who the president is. It came

Up with two, and this one is 40, so yeah, double the speed improvement with the M2 Max.

Which is good. M1 Max, I'm talking—I'm talking away and it still hasn't figured

out what the Earth is. This is the kind of planet it's trying to figure out what it is. That's the second image, and it's

figured it out, finally done it, and it found out that the planet is—it's got atmospheric layers; it's the Earth from

space. So good job. So you are getting considerably faster distances between the two on doing this Llama calculations.

Who needs to run AI on the cloud? You can run it on the—the M1—just AI in your

brains now. Regarding running larger models, for example Llama, they got a 200 MB memory-requiring model. You've got to

use a tool like Xcode. Let's just uh, get that guy running, and run, and we can see

that this M1 MacBook Pro, it goes 21 teraflops—flop, not Bell flop—a teraflop,

and this guy over here, the M2, does 36, so it is more than 50% faster. 20,

15, 15...yeah, 75% faster than the M1 Max. What's cool is we got another MacBook

Pro coming soon, and we're going to use this application to link them all together and see if we can run Llama

40B. Get the Llama running. But for now, you can see that it is estimating—they got 75% faster AI in the brains, and

what I want to do is a raw compute test, and this is unzipping, so we're going to

be unzipping Xcode's, 'cause we're doing some iOS next up. But let's open up these zip

files and see which one compiles it faster, and we're seeing that the M4 Max is slightly faster than the M1 Max, and

the M4 Max is going to win this, but it's just lagged behind. It's not like with

the AI kind of tasks where that one was twice as fast with the GPU.  This is maybe 10-20 seconds

faster than the M1 Max, and that's unzipping a 4 GB compressed file. Boom. Okay, since we've got Xcode, let's go fire

that guy up. I'm going to do a fresh build and run from scratch. I don't have any issues with Xcode; Xcode is just a

nice application to run on these systems, but maybe if you're doing a monolithic application,

probably like if you're trying to compile Chrome, oh my goodness, you're going to need lots of time on your hands for that one.

But anyway, for this basic iOS application, 3, 2, 1, let's hit it. It's already succeeded; it's already built on

the M4 Max. It's already built, so we're just firing up the simulator. Simulator's up; it's built now on the M1 Max.

Simulator is up on there, and oh, which one's going to get it first? Which one's going to get it first?

The M4 Max should be ready, but it's not... come on! Oh, we're done. So we got

29 seconds on the M4 Max, and the M1 Max is 15 seconds behind—45 seconds total. That's 50% faster to build from scratch and launch. I don't really know what's in

this project, because it's only 15 seconds, and rebuild/recompile times... like if I were to run it again from scratch, they're

both going to be instantaneous. 1, 2, 3, let's go. It's just within seconds of

each other, so getting 50% faster in Xcode, two times faster in Vision, and uh, two

times louder in CoreAudio. Yeah, you got that one as well. Let's be fair and let's switch over to the world of Android, so

we're going to go Android because we've got to keep equal allocation times on both of these guys. So Android

Studio. One good thing about this upgrade that I'm doing between the M1 to M4 is I'm not having any issues. The software

is still—even this latest version of macOS Ventura—seems to be running fine as it is. I remember when I

switched from Intel to the M1, I had a bit of tweaking issues, and then they went

away after a few months. But now with this M1 to M4, it seems to be okay. Obviously, there's a few more

applications that run in the background. I saw Private Compute trying to access the internet, and I denied access; denied

it. And inside Privacy, there's a new Private Compute kind of—there are some extra flags that you need to disable, but you know, it

is what it is. Let's jump on to Android. So we've got Gradle running over here; it's

compiling the task; bomb compiling C++ code getting compiled away in Android Studio. Some of that NDK is happening

over here. Just basic crypto mining situations. Something that hard—oh, it's done! Sorry, it took 24 seconds on the M4 Max.

Remember when the M1 Maxes came out? Twitter, they bought everyone M1 Maxes because they said it just compiled—con-  con-

completed their programming tasks. Anyway, so 24 seconds on the M4 Max and 45 seconds on the M1 Max, so it's double the

speed in compilation. You see 20% faster on the M2, 20% faster on the M3, another

20% faster boot that comes up with a 100% faster that is obviously it's just an extra 20 seconds, and that's a reboot

from scratch, so it's not going to change anything, and using the application is still going to be the same. Let's launch

this emulator over here and launch the emulator over there. Connecting to emulator, connect to the emulator, boom

boom boom boom boom. Launch fast on the M4 Max, but it's just a second or so behind on the M1 Mac. So, actually using the

application wasn't much of a difference. Compilation twice as fast, but you can imagine if you have a really heavy

project, you'd probably just be getting a PC. But if you had a project that needed a Mac, there you

go. Basically, if you work for someone and you want a business expense, you're going to be asked for a Mac. When I when I was

working for people, everyone wanted a Mac because they were the best laptops there are to get, and the PC world...

Max, they're portable; you can take them home and play with them as well. So you probably just want MacBook Pros so you

can get work-life balance on the screen all day long. All right, so let's switch over. I said I was going to do crypto, so

I think let's do some crypto. Yeah, let's just see um performance benchmarking. Has the crypto mania gotten much

faster than the M1? Now I'm going to do a torture test on this. I'm going to go

maximum crypto mining on these guys, which I don't recommend you guys doing, cuz Macs, you're meant to love them and

caress them and just make babies of them. This is another application I wrote called Mac Miner. It's I'm just using a few

libraries here and there, but it's pretty much going to do some Ethereum mining. So let's just jump in. I know you're not

meant to be able to mine Ethereum anymore, but there are other coins which still allow it

proof of work, so it says, um, it belongs in the trash, this application. So what you need to do, you need to go into

system settings, and then privacy and security, and then scroll to the bottom, and it comes up saying 'Mac Miner was

blocked.' So click 'Open Anyway,' and that will open it up. So by default, the power mode is set to low on

both, so I'm going to make it maximum. So we are mining away already on M4 Max. It's done its pre-processing and it's

mining at 31.5 megahashes. That is about a third the performance of a 3080, so Nvidia,

you know, those expensive $2,000, $1,000 cards, it's a third of the performance of that.

That's the M4 Max. The M1 Max is still doing the pre-processing, so I'll let

that finish, but from memory I believe that one was around 10. So we got 10-18 and we got 31, so that that means over

three times faster crypto mining performance, and I'm not hearing any... that's amazing! It can probably be

updated to support all the latest features in these chips, so I'm guessing we can go a lot faster. Let's just see.

M4 MAX MacBook Pro - Developer REVIEW | Xcode, Android, UE5, AI, Displays, Coil Whine

M4 MAX MacBook Pro - Developer REVIEW | Xcode, Android, UE5, AI, Displays, Coil Whine

Scripts