iPad Pro M4 Geekbench scores wow…

[Analog Kid]

I'm confused with the Metal score, my M1 Ultra gives me 119033 (which is much faster than the M4 here at 53630).

Any chances this is Geekbench 5?

You're comparing two M1 Maxes to an M4 Vanilla...

 

[ric22]

I'm looking forward to the next M4 Pro! Let's hope it's a good one after the crap M3 Pro (crap gains vs M2 Pro).

 

[gsal]

It’s interesting that they made the Apple logo out of copper for thermal cooling. Do you think they will release an active cooling attachment that will enable sustained performance when at a desk?

 

[Kylo83]

But the m1 is over kill for iPadOS, what does the app do to need that power, even a 2018 iPad Pro is fast enough

 

[magbarn]

But the m1 is over kill for iPadOS, what does the app do to need that power, even a 2018 iPad Pro is fast enough

Hopefully WWDC has some good surprises, and hopefully Apple doesn't artificially nerf the M1/M2 iPads just to upsell..

 

[DrWojtek]

This is so weird. What is the M4 Pro going to look like then? Just 2 more perf cores and more GPU cores and more display driver? I don't know if I like this product segmentation. The normal and Pro chips are so close together and then there is a massive gap to the Max.

For M4, it might be that every P core does 2500 points and every E core half, 1250. Roughly.

That would put a 6+6 M4 Pro at 20.000 multicore score, going on par with the M3 Max. And a 8+6 Max at 25000 points. Great product segmentation IMO.

 

[sparksd]

But the m1 is over kill for iPadOS, what does the app do to need that power, even a 2018 iPad Pro is fast enough

I saw improvements in 4K video processing & rendering when I went to the M1.

 

[Eracles]

The iMac deserves this M4

 

[NastyMatt]

That is a nice bump. Hoping this translates to more AAA games on the iPad.

I wonder if having ray tracing will help too..

 

[ttyRazor]

This is so weird. What is the M4 Pro going to look like then? Just 2 more perf cores and more GPU cores and more display driver? I don't know if I like this product segmentation. The normal and Pro chips are so close together and then there is a massive gap to the Max.

I’m guessing the M4 got 6 efficiency cores to make up for possibly binning performance cores. Hopefully the pro gets a similar bump in one of the core types, or it’ll end up like how the M1 Pro and Max were only different in GPU cores and memory channels. A base M4 MacBook Pro seems more attractive now, at least until you add RAM and storage to match the stock Pro chip configuration.

 

[klasma]

That is a nice bump. Hoping this translates to more AAA games on the iPad.

The market is probably too small, also in light of the additional development work needed e.g. to add touch input.

 

[klasma]

They might skip m4 pro and go straight to m4, m4 max and m4 ultra. An overclocked (as per mac apple silicon fashion) m4 with few more gpu cores can easily match m3 pro performance while delivering 2x more npu power thanks to the inclusion of half precision instructions. This can allow them to deliver m3 pro performance in the 1.599 base mx mbp sku with hopefully 16gb ram this time. I'm not sad that I bought an m3 pro model, I'm just in awe of this generational bump as m3 family is already stupid fast, and i'm excited for my 2027-8 macbook or ipad pro purchase

The M4 apparently only supports one external display. They will likely make an M4 Pro for the base supports-two-external-displays model.

 

[klasma]

I believe those scores are from the one terabyte version of the new iPad That cost a whole lot more money. I’d like to see the numbers to the 256 and 512 MB version.

Single-core performance should be the same though.

 

[CookItOff]

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

Compare that to a Core i9-14900KS scoring about 3,300 points.

www.tomshardware.com

3810! 🤯

 

[klasma]

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

Compare that to a Core i9-14900KS scoring about 3,300 points.

www.tomshardware.com

3810! 🤯

It seems those single-core results have to be taken with a grain of salt:

“The big single-core gains on Geekbench could be fueled by newly added support for Scalable Matrix Extensions (SME) — some of the subtests, like object detection and image blurring, see massive gains (~200% for object detection). Support would imply that Apple is using an ARMv9 architecture, but this isn't yet confirmed. In either case, SME would give a strong boost to some of the tests that form Geekbench’s CPU suite, bumping up the overall single-core score. However, Geekbench 6 only recently introduced support for SME with version 6.3, and Intel's competing AMX isn't supported. That's largely because matrix workloads are a better fit for other forms of compute, like the NPU or GPU, than the CPU cores. As such, it isn't clear how much real-world benefit SME would deliver if run on the CPU cores, if any, in daily usage.”

“The remainder of the benchmark score indicates a roughly 3% increase in performance over the prior gen M3”

This is a reminder that the Geekbench score isn’t a single linear measure.

 

[CookItOff]

It seems those single-core results have to be taken with a grain of salt:

I'll take some of that salt cause even if you started deducting for what you mentioned I don't think it would add up to the massive 500 point advantage of the M4 in an iPad, and the M4 would still be in the lead over the desktop PC world heaviest hitter.

 

[ric22]

It seems those single-core results have to be taken with a grain of salt:

“The big single-core gains on Geekbench could be fueled by newly added support for Scalable Matrix Extensions (SME) — some of the subtests, like object detection and image blurring, see massive gains (~200% for object detection). Support would imply that Apple is using an ARMv9 architecture, but this isn't yet confirmed. In either case, SME would give a strong boost to some of the tests that form Geekbench’s CPU suite, bumping up the overall single-core score. However, Geekbench 6 only recently introduced support for SME with version 6.3, and Intel's competing AMX isn't supported. That's largely because matrix workloads are a better fit for other forms of compute, like the NPU or GPU, than the CPU cores. As such, it isn't clear how much real-world benefit SME would deliver if run on the CPU cores, if any, in daily usage.”

“The remainder of the benchmark score indicates a roughly 3% increase in performance over the prior gen M3”

This is a reminder that the Geekbench score isn’t a single linear measure.

If it's really 3% only, which is hard to believe, that would be a gigantic fail.

 

[klasma]

I'll take some of that salt cause even if you started deducting for what you mentioned I don't think it would add up to the massive 500 point advantage of the M4 in an iPad, and the M4 would still be in the lead over the desktop PC world heaviest hitter.

Tom’s Hardware says, as I quoted, that disregarding the special-use SME, single-core performance only improved 3% over the M3. Of course you can believe what you like.

 

[klasma]

If it's really 3% only, which is hard to believe, that would be a gigantic fail.

Power efficiency apparently improved a lot, so that’s still a significant upgrade.

 

[CookItOff]

Tom’s Hardware says, as I quoted, that disregarding the special-use SME, single-core performance only improved 3% over the M3. Of course you can believe what you like.

Ok so the M4 is only 3% better than the M3. Got it.

 

[ric22]

Seems like Tom's Hardware really are out to get the M4, as they also claimed the ML progress is actually only 5%, and Apple are just calculating the numbers differently to before:

Alleged Apple M4 Geekbench scores show incremental improvement in machine learning over last gen

Incremental improvement over last gen

www.tomshardware.com

Note: A user on our forum is refuting their understanding of the situation, but has yet to present evidence.

 

[klasma]

Ok so the M4 is only 3% better than the M3. Got it.

It would explain why Apple so far has provided zero performance comparisons to the M3.

 

[Mark.g4]

A beast, impressive scores.

 

[Zdigital2015]

They might skip m4 pro and go straight to m4, m4 max and m4 ultra. An overclocked (as per mac apple silicon fashion) m4 with few more gpu cores can easily match m3 pro performance while delivering 2x more npu power thanks to the inclusion of half precision instructions. This can allow them to deliver m3 pro performance in the 1.599 base mx mbp sku with hopefully 16gb ram this time. I'm not sad that I bought an m3 pro model, I'm just in awe of this generational bump as m3 family is already stupid fast, and i'm excited for my 2027-8 macbook or ipad pro purchase

Apple isn’t going to skip the M4 Pro, it accounts for a sizable chunk of MBP sales at the retail level.

 

[Confused-User]

If it's really 3% only, which is hard to believe, that would be a gigantic fail.

The author is confused in a number of ways, which have already been discussed in other threads here.

Even if IPC gains are as low as 3% - even if they were 0%, which they're clearly not - this would not be a fail. Driving clocks 9% higher without *losing* IPC is a huge win, and not one that should be taken for granted.

If you look at the various subtests you'll see 15-20% gains are common. Factor out the 9% for clocks and you still have a really serious uplift.