Become a MacRumors Supporter for $50/year with no ads, ability to filter front page stories, and private forums.
The benchmarks show an only modest increase in machine learning capabilities over the M3. Apple’s AI features will likely run on any device with an A14/M1-class Neural Engine or higher.

The M3 was built on a flawed, expensive first-gen 3nm process. No one except Apple was using it. The N3E process the M4 is based on is significantly better in terms of yields and overall costs. It makes sense for Apple to get chip production off N3 and onto N3E as quickly as possible.
“Flawed” process?

More expensive, but not flawed.
 
The benchmarks show an only modest increase in machine learning capabilities over the M3. Apple’s AI features will likely run on any device with an A14/M1-class Neural Engine or higher.

The M3 was built on a flawed, expensive first-gen 3nm process. No one except Apple was using it. The N3E process the M4 is based on is significantly better in terms of yields and overall costs. It makes sense for Apple to get chip production off N3 and onto N3E as quickly as possible.
Intel uses N3B for upcoming lunar lake chips.
 
  • Like
Reactions: Chuckeee
I mean, yeah, when the original thread is about Geekbench results, comparisons to other vendors should be as well.

Here's the problem with your claim: with the CPUs I'm listing,

  • Apple's SoCs are M1 @ 3.2 GHz, M2 @ 3.5 GHz, M3 @ 4.05 GHz, M4 @ 4.4 GHz. That's already not great. Apple is approaching a ceiling here.
  • Intel's CPUs, meanwhile, are Ice Lake @ 2.3 GHz, Tiger Lake @ 2.9 GHz, Alder Lake @ 1.8 GHz, Raptor Lake @ 1.8 GHz, Meteor Lake @ 1.7 GHz. So Intel has been keeping the clock relatively low and even reducing it in some generational shifts (Ice Lake, Alder Lake, now Meteor Lake), which is why their IPC is doing so well.
  • Finally, AMD goes 1.8 GHz, 1.8 GHz, 2.7 GHz, 2.7 GHz, 3.3 GHz, 3.3 GHz. So while the 8840U scores 49% higher than the 4800U did, it also needs 83% more clock to do so.
Which makes me wonder why you would even bring up IPC at all. It clearly isn't AMD's strength, at least on mobile.
Are you taking Turbo Boost into account when calculating the 1 GHz scores? The clock speeds you’re mentioning sound like the base frequencies, but ST Geekbench would run at full speed.
 
Are you taking Turbo Boost into account when calculating the 1 GHz scores?

I thought about it and decided against it, opting for clock "as advertised" instead.

The clock speeds you’re mentioning sound like the base frequencies, but ST Geekbench would run at full speed.

Here's the same chart assuming max boost all the time (which I don't think is realistic):

1715449991270.png


This gets us a little closer to the claimed 10% yoy.

2020 to 2021, Intel goes up 16%, and AMD goes down 1%.

2021 to 2022, Intel goes up 2%, and AMD goes up 9%. Apple goes up (from 2020) 2%.

2022 to 2023, Intel goes up 18%, AMD 18%, and Apple 2%.

2023 to 2024, Apple goes up 12%.
 
I thought about it and decided against it, opting for clock "as advertised" instead.



Here's the same chart assuming max boost all the time (which I don't think is realistic):

View attachment 2377012

This gets us a little closer to the claimed 10% yoy.

2020 to 2021, Intel goes up 16%, and AMD goes down 1%.

2021 to 2022, Intel goes up 2%, and AMD goes up 9%. Apple goes up (from 2020) 2%.

2022 to 2023, Intel goes up 18%, AMD 18%, and Apple 2%.

2023 to 2024, Apple goes up 12%.

Thank you! I didn't realize that Geekbench actually does take several minutes to conclude, so it won't run the entire time at full turbo on the Intel and AMD chips, and the turbo duration will also be dependent on the cooling solution, but this seems more realistic than dividing by the idle frequencies, which are intentionally as low as possible to preserve power.
 
Modern chips, like the M4, have multiple individual processors in them. The M4 has up to 10 processors in the chip. Software can be written to run its code in one processor or in multiple. Running software in multiple processors can speed up the work but it is more complicated. Many apps only run on one processor at a time.

the benchmarks will tell you how fast a app would run using a single processor and also how fast it would run using multiple processors.

There are more nuances and complexities around this topic but that is the basics and probably the most helpful to you to understand the meaning of benchmarks.
Thank you very much!

So basically, to sum up a lot hehe, single core and multi core, it tells you how fast a "device" is?
I suppose both are kinda the same but it just depends if said app runs on single or multi, is that correct? (because I always see the multi core being way higher than the single core score)
 
  • Like
Reactions: Tagbert
Please note the keywords: “UP TO”.
(means you have to purchase the 1TB or over version to get the improvement.
 
A lot of people get hung up on benchmark scores, but if something that currently takes 0.6 seconds were to take only 0.5 would you notice? What if your new computer was twice as fast and the operation only took 0.3 seconds. Would you notice that? Many of the things our computers do are in those time frames. For consumers a computer spends most of its time waiting for data from the internet. Getting even a 10x faster CPU isn't going to magically make YouTube or Amazon load faster.

Professionals, people whose workflows include tasks that actually take minutes to complete, notice when they get a faster CPU or GPU.

I'm still on an Intel iMac because it does everything I need it to and the 27" 5K display that shows me a lot of information at once helps to maximize the speed of the slowest part of every computer system: the human using it.

I'm expecting to run into a wall with respect to support for Intel Macs in the 2026/27 time frame. So my next Mac will likely be rocking an M6 processor. Given that it should be 3x as fast as my current machine it will probably feel a little snappier.
 
Just traded in my IPad Pro M1 256GB WiFi and 5G, got over AU$600 for it. Ordered two iPad Pro's from Optus (cheaper on contract including data instead of buying outright and paying for data on top of the outright price over two years). 1 iPad Pro 13" 512GB in silver - WiFi and 5G (non-matte display) / 1x iPad Pro 13" 1TB in black - WiFi and 5G (non-matte display). Looking forward to these and giving them a spin (not literally). Will be interesting to see the CPU / GPU / RAM difference as well (if noticeable). Total cost of 2 years including data for the 512GB - AU$2,933.72. Total cost over 2 years including data for the 1TB - AU$3,633.56. I also primarily upgraded to experience the new OLED display, which hopefully comes to other MacBooks and other Apple hardware in the not-too-distant future.

OT: Will eventually replace my 15" MBP from 2015 which is still going with no issues, however waiting till after tax return and for the M4 models for the largest MPB to be released later in the year of next year. I am sick of hearing about iOS being gimped on such a powerful device, if you want the full desktop experience, a MPB is better as it has a larger screen as well as other hardware components. The iPad is not designed as a desktop replacement, merely and additional hardware device, to be used with other hardware and is designed for different use scenarios - they are to complement each other - not to replace the other.
 
  • Like
Reactions: NetMage
Thank you very much!

So basically, to sum up a lot hehe, single core and multi core, it tells you how fast a "device" is?
I suppose both are kinda the same but it just depends if said app runs on single or multi, is that correct? (because I always see the multi core being way higher than the single core score)

One way to think of it is “the multi-core score will tell you how fast a device is overall, and the single-core score how fast, in most cases, the current frontmost app will be”.

That’s only an approximation because it depends on factors like what kind of work you do, how heavily the app uses multiple cores, etc.
 
Guess my M3 Pro MacBook Pro I just got is obsolete already.
If you are even half-serious, I think you need to look up the definition of "obsolete" and the related adjective "obsolescent":

obsolete | ˈɒbsəliːt | : no longer produced or used; out of date; (grown old, worn out)

obsolescent | ˌɒbsəˈlɛsnt | : becoming obsolete; (falling into disuse)


None of these apply to your M3 Pro MBP - they are still produced, and even when an M4 version is released I doubt that your M3 will be "worn out", or "out of date" in the sense that it can no longer run current software, or will "fall into disuse" - at least not for several years.

My interpretation of these terms (for Mac computers) is that it becomes obsolescent when performance with current software becomes unacceptably slow, and it becomes obsolete when it no longer runs the required software.

Of course obsolescence is somewhat subjective. Some people may be absolutely fine with the performance of an old Intel Mac, whereas others find even an M1 or M2 Max unacceptably slow for their tasks.

I still run 14 year-old Intel Macs that are technically "obsolete" (they won't run all apps), but still useful for a large subset of computing tasks.
 
If you are even half-serious, I think you need to look up the definition of "obsolete" and the related adjective "obsolescent":

obsolete | ˈɒbsəliːt | : no longer produced or used; out of date; (grown old, worn out)

obsolescent | ˌɒbsəˈlɛsnt | : becoming obsolete; (falling into disuse)


None of these apply to your M3 Pro MBP - they are still produced, and even when an M4 version is released I doubt that your M3 will be "worn out", or "out of date" in the sense that it can no longer run current software, or will "fall into disuse" - at least not for several years.

My interpretation of these terms (for Mac computers) is that it becomes obsolescent when performance with current software becomes unacceptably slow, and it becomes obsolete when it no longer runs the required software.

Of course obsolescence is somewhat subjective. Some people may be absolutely fine with the performance of an old Intel Mac, whereas others find even an M1 or M2 Max unacceptably slow for their tasks.

I still run 14 year-old Intel Macs that are technically "obsolete" (they won't run all apps), but still useful for a large subset of computing tasks.

They're worried about Strix Point; full stop. M3 was a dud and their architecture can't scale because apus don't scale. They need to work on their backend and add a real pcie lanes with tunneling
 
Register on MacRumors! This sidebar will go away, and you'll see fewer ads.