News Apple M4 benchmarks suggest it is the new single-core performance champ, beating Intel's Core i9-14900KS — incredible results of 3,800+ posted

[Admin]

Apple M4 processors are convincingly breaking Geekbench single-core records. Multiple scores of roughly 3,800 points or over have appeared in the Geekbench online database over recent hours.

Apple M4 benchmarks suggest it is the new single-core performance champ, beating Intel's Core i9-14900KS — incredible results of 3,800+ posted : Read more

 

[Dementoss]

That's all very well but, demanding applications are typically multi-threaded and, M4's multi-threaded score is rather less impressive.

 

[diminishedfifth]

Why is the multi thread score ignored?

 

[kealii123]

Am I reading this correctly? A tablet handily beats Intels flagship desktop CPU in single threaded?

"But muh multicore" THIS IS A FREAKING TABLET VS A DESKTOP CPU THAT CONSUMES OVER 300 WATTS

 

[CelicaGT]

Am I reading this correctly? A tablet handily beats Intels flagship desktop CPU in single threaded?

"But muh multicore" THIS IS A FREAKING TABLET VS A DESKTOP CPU THAT CONSUMES OVER 300 WATTS

This says more about the i9 than the M4. AFAIK lesser Intel chips and AMDs stuff come closer in perf/W.

 

[JamesJones44]

That's all very well but, demanding applications are typically multi-threaded and, M4's multi-threaded score is rather less impressive.

It only has 4 performance cores and 6 e cores. The fact that its performance is 69% of a processor that has 4 more performance cores and 12 more e cores is pretty impressive. If the M4 architecture can scale, an M4 Pro or Max could beat a 14900K in both single and multi thread benchmarks while using way less power to do so.

 

[ezst036]

Everybody gawks at the performance.

Not one of you wants to be trapped in that walled garden.

 

[magbarn]

Everybody gawks at the performance.

Not one of you wants to be trapped in that walled garden.

Or get stuck paying SSD/Memory prices from over a decade ago...

 

[hushnecampus]

That's all very well but, demanding applications are typically multi-threaded and, M4's multi-threaded score is rather less impressive.

This is the base M4. Compared to Intel's top of the range. One would assume the M4 Pro alone will stomp Intel, let alone the models up from that. Unless of course Intel bring something new out before the M4 Pro.

 

[hushnecampus]

Everybody gawks at the performance.

Not one of you wants to be trapped in that walled garden.

I do.

 

[CarlosO]

Well, it is all nice and dandy in the Apple world. At the end of the day, does such performance of the tablet for watching Youtube videos or browsing webpages, like the 95% of people will do with such tablet, which has hampered multitasking an limited access to limited on board space.... Any orher tablet for 1/3 can do almost the same thing. Apple has fallen into the loop. Their software garden is going to die off unless they open it to other insects. And selling these devices for such a premium prices, when you can buy a laptop and do some serious work.

 

[helper800]

I do.

Heretic!

 

[twotokensandablue]

I do.

How come? Seems like you're severely hindering yourself by opting to use an Apple computer.

 

[HaninTH]

Well, it is all nice and dandy in the Apple world. At the end of the day, does such performance of the tablet for watching Youtube videos or browsing webpages, like the 95% of people will do with such tablet, which has hampered multitasking an limited access to limited on board space.... Any orher tablet for 1/3 can do almost the same thing. Apple has fallen into the loop. Their software garden is going to die off unless they open it to other insects. And selling these devices for such a premium prices, when you can buy a laptop and do some serious work.

And I would find it surprising to see anyone else get the same performance out of the hardware as Apple even if they opened it up to alternative OSes. Apple would keep the best to themselves, nor am I convinced the MPAA/RIAA/Content Cartels would let it go without highly restrictive DRM which Apple will not document.

So many possibilities, locked in by greed.

 

[kealii123]

How come? Seems like you're severely hindering yourself by opting to use an Apple computer.

Ho so? I've always hated Apple. I hate their hypocritical preaching on "sustainability". I hate their smugness. I hate the walled garden, but the only thing that can't be done on a macOS device that can be done better in Windows is gaming. (Also, Microsoft has turned just as preachy & hypocritical. Windows literally tells me to turn down my screen brightness to save on "carbon emissions".)

 

[PEnns]

Everybody gawks at the performance.

Not one of you wants to be trapped in that walled garden.

When your horse is beaten badly by another horse......"Oh look, a flying saucer!"

Walled garden? As I see it: Millions of Apple users around the globe are wrong but the Apple-hating crowd is right!

 

[helper800]

When your horse is beaten badly by another horse......"Oh look, a flying saucer!"

Walled garden? As I see it: Millions of Apple users around the globe are wrong but the Apple-hating crowd is right!

Billions of Android/Windows/Linux users are right you mean?

 

[Findecanor]

Apple has not disclosed much about the microarchitecture of the P-cores in its press release, but it should have "wider decode" ... compared to M2 of course.

Both P-cores and E-cores have a "machine learning" enhancement, which probably means SME.
SME is available for ARMv8.6, and there was a dot-product extension for ARMv8.4 already. The M3 is ARMv8.6.
If the CPU was ARMv9 then it would have supported SVE, which is an enhanced SIMD instruction set. It is a large step so I think Apple would have mentioned that, and we might have seen even more interesting benchmarks.

 

[PEnns]

Billions of Android/Windows/Linux users are right you mean?

Are they all complaining like you??

 

[twotokensandablue]

When your horse is beaten badly by another horse......"Oh look, a flying saucer!"

Walled garden? As I see it: Millions of Apple users around the globe are wrong but the Apple-hating crowd is right!

I don't think most of Apple's users realize it's a walled garden, most pick Apple because it's convenient.

 

[scottslayer]

Am I supposed to be impressed by Apple beating 10nm+++++++++++?!

 

[ezst036]

I do.

Probably because you are already there, am I right?

 

[JamesJones44]

Probably because you are already there, am I right?

I'll bite...

As someone who uses Linux, Windows and Mac the answer for me is no, I wasn't. I started my life off using Dos and Windows. Later I had to start using Linux for work and backing development, I used to do this in a VM, but then when the original OS X was released based on Unix the ability to be able to run linux scripts and workflows made development a lot easier. When Mobile came along I needed to be able to do Windows, Linux/Android and Mac/iOS. Mac is the only one that can do all 3. These days with ML it's all Linux based and macOS runs almost anything Linux does thus it's way easier to maintain and update tools are required for the work I do. For testing small datasets you don't need a farm H100s or even a 409x series. Large datasets are trained in our server farms. WSL can do most of this as well today, but WSL is still stuck somewhere between running something in a VM and somewhat integrated in with the rest of the Windows environment. Enough so that it's not a seamless experience the way terminal is on the macOS.

Other than performance issues with Windows 11, I don't have a strong opinion on any of the OSs, they all have their pros and cons. macOS at the moment just aligns best with what I do on a daily bases while still maintaining fairly easy to manage system (Linux desktops get a little funky, but have gotten way better in recent years).

All that being said, I wouldn't waste a penny on these iPad Pros. iPadOS in no way can leverage most of what an M4 offers.

 

[TerryLaze]

Highest result for an intel cpu under windows is a single score of 4580 with an i9-14900KS, now I'm no mathematician but I think that is higher than 3800 points.
Running android or linux on x86 results in much higher results as well up to 8650....probably geekbench compensates for differences in archs which means that results are not comparable between different types of cpus.
https://browser.geekbench.com/v6/cpu/5568973

 

[JamesJones44]

Highest result for an intel cpu under windows is a single score of 4580 with an i9-14900KS, now I'm no mathematician but I think that is higher than 3800 points.
Running android or linux on x86 results in much higher results as well up to 8650....probably geekbench compensates for differences in archs which means that results are not comparable between different types of cpus.
https://browser.geekbench.com/v6/cpu/5568973

I've never trusted the individual submissions to GeekBench, the first 4 entries on the single core tests don't even look real, let alone all of the other anomalies in other submissions. They do publish an average of the scores which is probably more useful when comparing processors, but I've never taken GeekBench scores as the end all be all of benchmarks.

The issue for me is they don't really separate out overclocking so it's hard to say what is factory and what is overclocked.