M3 Pro Chip Barely Faster Than M2 Pro in Unverified Benchmark Result

[FrankySavvy]

It just makes me laugh, I love how far Apple Silicon has come when we are complaining about impressive year over year upgrades. The M3 Pro is faster than the M2 Max, thats a testament to engineering.

For me I am going from a M2 MacBook Air to an M3 Pro seeing +16% performance boost in Single Core and a crazy 52% in multicore performance, plus I am sure the GPU will impressive gains as well.

We get so tied up to year over year upgrades, how many people do you know that upgrade every year?

For example the M3 Pro (12/18) which is now $2199, is faster than the M2 Max (12/30) which was $3099 when it first came out 9 MONTHS AGO! and is 30% faster than the M1 Ultra in Single Core performance which was over $3500 in the Mac Studio last year.

Geekbench Tests below for reference.

M3 Pro (12 Cores) *Single Test*
Singe Core: 3035 +13% Faster than M2 Pro
Multicore: 15,173 +4% Faster than M2 Pro

M2 Max (12 Cores)
Single Core: 2835
Multicore: 14,952

M1 Ultra (20 Cores)
Single Core: 2346
Multicore: 18,168

Plus the M3 Pro is only $200 more to upgrade from the fully binned M3. Which is a 27% better multi-core performance gain for $200 dollars. The M3 Max would be an additional $1200 upgrade for an additional 52% increase, not sure thats worth it.

Look at the M1 -> M3 generation increase in single core is over 30% and multicore is a 42% increase. Thats in 3 years. Extremely impressive!

 

[satchmo]

No. Apple knows that consumers want a physical change. It’s not because “performance gain isn’t enough”

The M3 Max is scoring similarly to the M2 Ultra. That’s a laptop chip performing the same as Apples most powerful desktop.

Users want a physical change, but won't upgrade if there if none is offered.

There's not enough of difference between a M2 Pro and M3 Pro...so Apple added colour as an added incentive.

 

[bcortens]

Plus the M3 Pro is only $200 more to upgrade from the fully binned M3. Which is a 27% better multi-core performance gain for $200 dollars. The M3 Max would be an additional $1200 upgrade for an additional 52% increase, not sure thats worth it.

The base M3 Pro is binned - we don't know the scores for the binned M3 Pro but it will be less than the fully enabled M3 Pro so your $200 price increase for + 27% performance is wrong.

 

[Gudi]

Believe me, I like this made up story. But there's no actual evidence any of it is true, and Apple literally didn't talk about any of these 'improvements' that don't actually exist.

So what do you think the new maximum of 6 E-cores in the M3 Pro will (actually) do to the battery life of that machine?

 

[stars_fan]

Better to call it M3 Plus.

 

[Gudi]

The data available means the following:

M2 -> M2 Pro offers 46% performance increase
M3 -> M3 Pro offers 30% performance increase

46 > 30

Thus the value for money of the M3 Pro is worse than the M2 Pro was.

If your workload can be done on a M2 or M3, then the value of both the M2 Pro and M3 Pro to you equals zero. It’s exactly the same waste of money. Otherwise it only makes sense to compare Pro chips with each other.

 

[name99]

BTW the GeekerWan video (look at 3:12) seems to confirm P cluster sizes are now 6 (presumably likewise for Pro E cluster).
I still don't know what the AMX (or other shared resource) situation is.

BTW the raw inter-core-to-core latencies seem to be about 2/3 relative to M1 Pro, with cross-core latencies slightly higher. Probably a substantial overall win given how aggressively Darwin tries to pack communicating threads on the same cluster.
The Apple intra-cluster numbers are now *comparable* to Intel or AMD core-to-core numbers. There are of course vast differences in the precise details of each, say M3 Max vs Lakefield vs a Zen CCD, but the numbers are not wildly out of sync between the various designs; Apple presumably, as always, optimizing somewhat for energy over raw latency.

Other interesting tech details from the video include that the branch predictor is extraordinary, 10% lower misprediction rate than was already state of the art. (I can't find a recent Intel result to compare, but about half the misprediction rate of a Haswell Xeon)

 

[bcortens]

If your workload can be done on a M2 or M3, then the value of both the M2 Pro and M3 Pro to you equals zero. It’s exactly the same waste of money. Otherwise it only makes sense to compare Pro chips with each other.

Wow.since almost all workloads ‘can’ be done on any mac this means everyone should just buy the base model. Terrible rational.

 

[ProfessionalFan]

Wow.since almost all workloads ‘can’ be done on any mac this means everyone should just buy the base model. Terrible rational.

Well yes, if you don't need the more powerful machines to get work done quicker, there's no reason to get them. Obviously those who need work done quicker, have a good reason to upgrade. They are all capable, it is just a matter of how long it takes to get tasks done. Time is money.

 

[Gudi]

Wow.since almost all workloads ‘can’ be done on any mac this means everyone should just buy the base model. Terrible rational.

Yes, 90% of users are probably fine with an M3. And if you indeed do need a Pro chip, why measure its value versus the M3, who isn’t fit for the job?

 

[bcortens]

Well yes, if you don't need the more powerful machines to get work done quicker, there's no reason to get them. Obviously those who need work done quicker, have a good reason to upgrade. They are all capable, it is just a matter of how long it takes to get tasks done. Time is money.

Then the value for money equation matters. The person I was responding to was claiming that value for money doesn’t matter. I disagree.

 

[bcortens]

Yes, 90% of users are probably fine with an M3. And if you indeed do need a Pro chip, why measure its value versus the M3, who isn’t fit for the job?

The whole point of this thread is about how the M3 pro is disappointing. The M3 Pro can be good value if you need more than the M3 while still being disappointing. Both these things can be true at once.

The M3 Pro is disappointing for several reasons:

Reason 1:
Because it represents worse value for money than the M2 Pro did at the same price.
As I pointed out earlier in the thread:
The M2 -> M2 Pro provides 46% increased CPU multi-core performance, 1.9X GPU performance, 2X memory bandwidth.
The M3 -> M3 pro provides 30% increased CPU multi-core performance, 1.8X GPU performance, 1.5X memory bandwidth.

All the numbers are smaller for the M3 Pro - thus proving, that as an improvement over the base M3, the M3 Pro offers less than the M2 Pro did over the M2.

This makes the M3 Pro a disappointing year on year improvement.


Reason 2:
Minimal improvements in multi-core vs M2 Pro


The M3 Pro:
It is still an improvement.
It is still (probably) good value
It is disappointing because it was decontented without a price decrease

These things can all be true at once.

 

[Gudi]

All the numbers are smaller for the M3 Pro - thus proving, that as an improvement over the base M3, the M3 Pro offers less than the M2 Pro did over the M2.

And I told you that this comparison is pointless. In the end a chip must not win benchmarks, but stand the test of time in real world applications. If you like it or not, most of these chips are going to be built into laptops and those laptops are going to be used plugged or unplugged. So raw performance without battery life is only one way to look at it. And it’s a way which lets Intel x86-64 and even PowerPC look good in comparison. But it hides all the benefits the M3 Pro brings over the M2 Pro. Apple spent so much time during the keynote to highlight the main advantage of the M3 family of chips:

The only way to be disappointed about this, is when you completely ignore Apple's emphasize on performance per watt.

 

[bcortens]

And I told you that this comparison is pointless. In the end a chip must not win benchmarks, but stand the test of time in real world applications. If you like it or not, most of these chips are going to be built into laptops and those laptops are going to be used plugged or unplugged. So raw performance without battery life is only one way to look at it. And it’s a way which lets Intel x86-64 and even PowerPC look good in comparison. But it hides all the benefits the M3 Pro brings over the M2 Pro. Apple spent so much time during the keynote to highlight the main advantage of the M3 family of chips:
…
The only way to be disappointed about this, is when you completely ignore Apple's emphasize on performance per watt.

You are ignoring the point of this thread again.

We can’t assess whether or not the M3 Pro is disappointing without looking at how it fits in relative to how the M2 Pro fit in. They are pitched as being the same class of chip but they aren’t, the M3 Pro is a lower tier chip. In Intel parlance ito as if the M2 Pro was an i7 and the M3 Pro is an i5.

The M2 Pro was fantastic performance per watt chip already. We have no idea whether or not the M3 Pro consumes less power than the M2 Pro at load. We also have no way of knowing what the power consumption of a less disappointing M3 Pro (one that added rather than subtracted transistors) would be.


The thread is about whether or not the M3 pro is a disappointing chip. To do that, we can look at where it fits in the lineup. The fact that it moved down-market without changing the price is disappointing.

 

[FasterQuieter]

Seems like a good idea to have more segmentation between the 3 chips. The cost being this year’s Pro’s paltry gains. If I needed a Mac now it would have pushed me to the standard M3, saving me money. If I had money to burn, I’d have much more incentive to go with the Max.

 

[Gudi]

We can’t assess whether or not the M3 Pro is disappointing without looking at how it fits in relative to how the M2 Pro fit in. They are pitched as being the same class of chip but they aren’t, the M3 Pro is a lower tier chip.

Or a higher tier chip as it is the only one with six efficiency cores.

In Intel parlance it's as if the M2 Pro was an i7 and the M3 Pro is an i5.

Intel, the guys with 4× to 5× the energy consumption? You do realize that i5 and i7 mean absolutely nothing! Those aren't even numbers, just marketing names under which hundreds of different chips were sold.

The M2 Pro was fantastic performance per watt chip already. We have no idea whether or not the M3 Pro consumes less power than the M2 Pro at load.

I do have an idea, I just don't have benchmarks to back it up.

We also have no way of knowing what the power consumption of a less disappointing M3 Pro (one that added rather than subtracted transistors) would be.

More transistors more energy consumption. The reason E-cores are more efficient is because they are built with fewer transistors than P-cores.

The thread is about whether or not the M3 pro is a disappointing chip.

Which is an individual feeling different from person to person.

 

[exon]

PC Mag has a review of the 14" with the lower end M3 Pro (11-core CPU / 14-core GPU) that I didn't see linked previously:

Apple MacBook Pro 14-Inch (2023, M3 Pro) Review

Apple's 2023 14-inch MacBook Pro strikes a winning balance between premium design and M3 Pro-powered performance, for a compact mobile workstation that trounces similarly priced Windows machines.

www.pcmag.com

The benchmarks likely won't change anyone's mind in this thread, but they did get 30(!) hours of battery life in their simple test (looping local 720p file at 50% screen brightness).

 

[chucker23n1]

PC Mag has a review of the 14" with the lower end M3 Pro (11-core CPU / 14-core GPU) that I didn't see linked previously:

Apple MacBook Pro 14-Inch (2023, M3 Pro) Review

Apple's 2023 14-inch MacBook Pro strikes a winning balance between premium design and M3 Pro-powered performance, for a compact mobile workstation that trounces similarly priced Windows machines.

www.pcmag.com

The benchmarks likely won't change anyone's mind in this thread, but they did get 30(!) hours of battery life in their simple test (looping local 720p file at 50% screen brightness).

I wonder how that compares to the M2 Pro (4 e-cores) and M1 Pro (just 2 e-cores). The e-cores got more plentiful and more powerful each generation.

I have the 10-core M1 Pro, and I wouldn’t mind a slight battery life improvement. E-cores seem to be always buy around 50% each.

 

[Jorbanead]

Users want a physical change, but won't upgrade if there if none is offered.

There's not enough of difference between a M2 Pro and M3 Pro...so Apple added colour as an added incentive.

I can assure you 97% of people are not upgrading from M2 to M3 computers. People just don't upgrade their computers that fast nor do most people need to. Only a small 3% of people do that, and usually those people need the top-end performance so they're going with Max/Ultra.

These M3 computers are mostly for those that are still on intel machines, and maybe a smaller portion of people who are on M1 machines. For myself, I personally upgrade fairly often. I have an M1 Max Studio, and I don't plan to upgrade until the M4 Max Studio comes out.

Source

 

[name99]

The whole point of this thread is about how the M3 pro is disappointing. The M3 Pro can be good value if you need more than the M3 while still being disappointing. Both these things can be true at once.

The M3 Pro is disappointing for several reasons:

Reason 1:
Because it represents worse value for money than the M2 Pro did at the same price.
As I pointed out earlier in the thread:
The M2 -> M2 Pro provides 46% increased CPU multi-core performance, 1.9X GPU performance, 2X memory bandwidth.
The M3 -> M3 pro provides 30% increased CPU multi-core performance, 1.8X GPU performance, 1.5X memory bandwidth.

All the numbers are smaller for the M3 Pro - thus proving, that as an improvement over the base M3, the M3 Pro offers less than the M2 Pro did over the M2.

This makes the M3 Pro a disappointing year on year improvement.


Reason 2:
Minimal improvements in multi-core vs M2 Pro


The M3 Pro:
It is still an improvement.
It is still (probably) good value
It is disappointing because it was decontented without a price decrease

These things can all be true at once.

Then don't freaking buy one!
What the hell is wrong with you people that you regard it as a personal insult that someone else might dare to want a machine you don't want? Other people want different amounts of RAM from me, just like they want different amounts of SSD and different CPU or GPU performance levels.

Is the M3 Pro worse than the M2 Pro?
Yes - then buy an M2 Pro mac while they are still available!
No - then buy an M3 Pro mac and be happy!
The same - well, do you HAVE an M2 Pro mac? Then why do you need an M3? If you don't have an M2 Pro mac, then WTF does the comparison matter – we all agree it's better than what you do have.

There is just no universe in which the issue deserves this level of whining!

 

[Chuckeee]

I can assure you 97% of people are not upgrading from M2 to M3 computers. People just don't upgrade their computers that fast nor do most people need to. Only a small 3% of people do that, and usually those people need the top-end performance so they're going with Max/Ultra.

Actually, I was under the impression, that many of those that upgrades that occurring within a year or two; are in response to realizing they need additional RAM as compared to their initial purchase. As a result, those upgrades are one time occurrence as compared to regular upgrade

 

[ksj1]

The same - well, do you HAVE an M2 Pro mac?

This is the part I'm unclear about. Isn't the value argument being thrown out the window already if you're upgrading on less than a 3 year cycle?

 

[jido]

Geekbench is starting to show more results for Mac15,6 and Mac15,7.
They didn’t really change. First bench must have been genuine.

 

[ZebedeeG]

Expect a lot of clickbait videos today along the lines of "Apple Messed Up!" or "Don't buy an M3 Mac until you've seen this..!" etc... etc...

Seems to me that all the cpus are quicker, but with the M3 Pro showing the smallest gain.

M2 to M3 like for like prices have crept up a bit - but then so has just about everything else I buy!!

I will be guilty of watching some of the YouTube reviews, and I'll be interested to see how the drop in memory bandwidth effects performance especially as the actual amount of unified memory seems to have been increased on affected models (16 GB to 18Gb and 32GB to 36GB if I'm reading the specs right). Which may offset it? 🤷‍♀️ Will it make any difference?

The M3 chips seem to make much more sense, with steady increases in cpu & gpu power as you go up the range.

But for me it's really only of academic "geek interest" as the base chips more than meet my needs.

I'm sure the most demanding tasks some run on their Pro & Max machines is... Benchmarks!! (Then get upset if they're not good enough! 😆)

 

[fpnc]

I guess the M3 Pro is fine for what it is, a decent upgrade over the standard M3. However, my disappointment is what it APPARENTLY will do to the performance equation on the next Mac mini.

With the M2 Pro the Mac mini could have been considered as being at the low end of the "pro" experience for Apple silicon. But with the M3 Pro's restricted CPU and memory (only a 4GB increase in maximum RAM) I don't see how it can compare in any way to a Mac Studio with the M3 Max. I guess that could be one of the reasons why Apple made this change and it means that I am no longer considering an M3 Mac mini Pro for my desktop.

That said, might we see an iPad with the M3 Pro? Seems quite possible and that may be another reason for the de-spec on the M3 Pro. Seems like what they have done is to emphasize power efficiency and lower cost and that's exactly what they'd want for an iPad. Plus, it will probably improve margins on the Mac mini Pro or at least allow them to hold steady given the apparent price increases that TSMC has demanded for their advanced fabrication processes.