Apple M4 LTT review prediction

[flocked]

Apparently the new M4 chip has an increase of 25% multicore, 100% on the neural engine (Apple was btw. one of the first company building in a neural engine in a consumer CPU for machine learning since the A11 2016).

When you e.g.. search for Apple or Pizza in photos, it finds you the pics with that object. It runs locally compared to the cloud in android  (and sure you can now maybe find me some app that does it locally, but it isn't the same to have it integrated locally first party)

 

The graphics performance apparently got also again 30% faster. And single core also around 20%.

 

This all is faster for an upgrade after only 7 months after the M3 release which already was an upgrade of 80% in GPU compared to M2.

 

My LTT review prediction: Instead of mentioning this even once in the review, Linus will rant the whole video review who Apple compares the M4 to an M1 and an 200% increase. They set a false picture, they cheat on the consumer.

 

Apple for sure 100% does this for marketing reason. But also because nobody buys a new computer or iPad every year, but more likely every 3-4 years if not longer… The M4 is as fast as the M3 Pro, the M3 as fast as the M2 Pro, etc. (except of the GB/s ram to video ram speed).

 

Yes Apple needs to be criticized. But don't spend half the review on how bad the marketing is.

----

edit: changed the fact that apple is the first company to integrate a neural engine to "one of the first"

[whispous]

Imagine getting angry at something that currently only exists in your imagination, at least give it a chance to be made and published before bashing it lol

[Aleph256]

17 minutes ago, flocked said:

But also because nobody buys a new computer or iPad every year, but more likely every 3-4 years if not longer…

It's impressive how you underestimate the Apple addicts' ability to want the new shiny thing.

There's a reason second-hand last-year iDevices are consistently so numerous and so cheap (compared to their base price).

[dogwitch]

36 minutes ago, Aleph256 said:

It's impressive how you underestimate the Apple addicts' ability to want the new shiny thing.

There's a reason second-hand last-year iDevices are consistently so numerous and so cheap (compared to their base price).

its cheaper to build a cult base. it been proven that in multi countries for a brand.

[manikyath]

1 hour ago, flocked said:

Apple was btw. the first company building in a neural engine in a consumer CPU for machine learning since the A11 2016

they were the first since the last time it was done?

 

except.. qualcomm has an entire product stack around this you didnt even bother to google?

https://www.qualcomm.com/products/mobile/snapdragon/smartphones/mobile-ai

samsung has announced one:

https://www.gizchina.com/2024/05/05/samsung-unveils-first-3nm-mobile-processor-designed-with-ai/

even mediatek has AI accelerated stuff:

https://www.mediatek.com/technology/ai-for-smartphones-6th-gen

 

also.. yes, they will rip on comparing performance to not-the-last-gen to get a bigger number. imagine if AMD would compare ZEN5 performance to the K6. it's a ridiculous metric where you can just cherry pick the number you want. the rate at which users dispose of their devices is irrelevant to this debate.

 

also, you're angry about a review that isnt even out yet.. spring is coming, please use your energy for observing greens outside instead of ranting about an imagined negative review for a multi-billion dollar company you fanboy so hard you made a new account just for this rant.

[flocked]

12 hours ago, manikyath said:

they were the first since the last time it was done?

 

except.. qualcomm has an entire product stack around this you didnt even bother to google?

https://www.qualcomm.com/products/mobile/snapdragon/smartphones/mobile-ai

samsung has announced one:

https://www.gizchina.com/2024/05/05/samsung-unveils-first-3nm-mobile-processor-designed-with-ai/

even mediatek has AI accelerated stuff:

https://www.mediatek.com/technology/ai-for-smartphones-6th-gen

 

also.. yes, they will rip on comparing performance to not-the-last-gen to get a bigger number. imagine if AMD would compare ZEN5 performance to the K6. it's a ridiculous metric where you can just cherry pick the number you want. the rate at which users dispose of their devices is irrelevant to this debate.

 

also, you're angry about a review that isnt even out yet.. spring is coming, please use your energy for observing greens outside instead of ranting about an imagined negative review for a multi-billion dollar company you fanboy so hard you made a new account just for this rant.

Did you read my post? Apple has a neural engine since 2017. You post a chip from 2024.

[flocked]

People critique me for being an Apple fanboy. I hate Apple. But can we please keep getting back argumenting instead of calling me something I not am. I still didn't read a single argument.

[starsmine]

1 hour ago, flocked said:

Did you read my post? Apple has a neural engine since 2017. You post a chip from 2024.

K

https://www.qualcomm.com/news/releases/2016/05/qualcomm-helps-make-your-mobile-devices-smarter-new-snapdragon-machine

 

1 hour ago, flocked said:

People critique me for being an Apple fanboy. I hate Apple. But can we please keep getting back argumenting instead of calling me something I not am. I still didn't read a single argument.

Because you posted a hate thread about LTT not being nice enough to the m4 in a review that is not even out yet with arguments that do not track. 

if the M4 is better then the M3, show that it is better then the M3, not the M1, being better hen the M1 is assumed. If the M3 was not good enough to get a person to upgrade, show how much better the M4 is to that M3.

[flocked]

5 minutes ago, starsmine said:

K

https://www.qualcomm.com/news/releases/2016/05/qualcomm-helps-make-your-mobile-devices-smarter-new-snapdragon-machine

 

Because you posted a hate thread about LTT not being nice enough to the m4 in a review that is not even out yet with arguments that do not track. 

if the M4 is better then the M3, show that it is better then the M3, not the M1, being better hen the M1 is assumed. If the M3 was not good enough to get a person to upgrade, show how much better the M4 is to tthat M3.

I'm just predicting. Also based on all the reviews LTE so far did. Just check the review and it is half about the marketing instead of the chip (Linus actually does the opposite to Apple. Show the right graphics while ranting about some other fact :D) I'm not denying that is showing graphics of 3-4 old chips. The show it to chip that people actually often previously owned. Many people I know often keep their laptops 3-4 years.

But can we not talk only about one of the points I said.

----

You are btw. right. Qualcom was half a year faster. Still Apple is one of first using it stuff. All the pictures on the iPhone e.g. are processed using it so you can e.g. search for Pizza or dog and find the pics with this content. - It all runs locally compared to the cloud how it works in android (and sure you can now maybe find me some app that does it locally, but it isn't the same to have it integrated locally first party)

[flocked]

Geekbench score of the M4 is out:

• M4 with 3778 (Single Core) and 14.518 (Multi-Core)
• M3 Pro (the old better model) with 3100 (Single Core) and 13.600 (Multi-Core)

Remember this is an improvement after just 7 months. Sure this are just synthetic tests, but give a clear picture about Apples yearly advancement.

[flocked]

On 5/12/2024 at 3:24 AM, starsmine said:

K

https://www.qualcomm.com/news/releases/2016/05/qualcomm-helps-make-your-mobile-devices-smarter-new-snapdragon-machine

 

Because you posted a hate thread about LTT not being nice enough to the m4 in a review that is not even out yet with arguments that do not track. 

if the M4 is better then the M3, show that it is better then the M3, not the M1, being better hen the M1 is assumed. If the M3 was not good enough to get a person to upgrade, show how much better the M4 is to that M3.

What you are linking is by the way a software SDK. Take a deep read. n software you can do AI since 30 years.

[igormp]

9 hours ago, flocked said:

What you are linking is by the way a software SDK. Take a deep read. n software you can do AI since 30 years.

The SD 820 already had a Hexagon DSP that managed to run some conv nets fairly good. I've worked developing stuff for it personally. Hexagon itself with proper vector instructions has been available since 2013 or so (not sure if previous versions were any good for that), that announcement is just an SDK to make it easier to develop neural networks for it, but it was already possible before then.

 

On 5/11/2024 at 10:37 PM, flocked said:

All the pictures on the iPhone e.g. are processed using it so you can e.g. search for Pizza or dog and find the pics with this content. - It all runs locally compared to the cloud how it works in android (and sure you can now maybe find me some app that does it locally, but it isn't the same to have it integrated locally first party)

Android is a really fragmented ecosystem. With that said, the default gallery app in my android phone can do exactly what you said locally, that's not really an impressive feature.