Is the M2 really faster than the M1?

[MayaUser]

The point I haven't come across yet is the poor "optics" of the situation...

18% increase over two years?

Yes, I know about A14 v A15 (or A16), and the number of cores, and the chip shortage, etc -- I'm just talking about the optics of the presentation. It "appears" that Apple has only been able eke out a 9% improvement YoY in actual core speed of their Apple Silicon.

Intel would be ridiculed for this type of announcement. And both sides would be partially right.

Again, I know that it will improve with A16 or A17 (whichever the M3 is based on), even more cores, etc. It just looks bad given the timing and the compromises they had to make to get the M2 launched along with their hardware schedule.

That said, I have a MacBook Pro and Mac Studio, both with M1 Max, and I'm really, really impressed with them. I have to do insane things to them in order to hear the fans come on.

while i agree with you...i have to disagree with you in some points
M1 MBA was announced 18 months and not 24
18% increase in raw cpu is ok for a second 5nm generation, expect a better improvement with the M3 since it will have 3nm and armV9
But also...the gpu is over 30% improvement over 18 months so...and think about scaling this to the M2 Max/Ultra since probably we will have LPDDR5x into those

 

[Kaikidan]

The point I haven't come across yet is the poor "optics" of the situation...

18% increase over two years?

Yes, I know about A14 v A15 (or A16), and the number of cores, and the chip shortage, etc -- I'm just talking about the optics of the presentation. It "appears" that Apple has only been able eke out a 9% improvement YoY in actual core speed of their Apple Silicon.

Intel would be ridiculed for this type of announcement. And both sides would be partially right.

Again, I know that it will improve with A16 or A17 (whichever the M3 is based on), even more cores, etc. It just looks bad given the timing and the compromises they had to make to get the M2 launched along with their hardware schedule.

That said, I have a MacBook Pro and Mac Studio, both with M1 Max, and I'm really, really impressed with them. I have to do insane things to them in order to hear the fans come on.

TSMC 3nm process isn't ready yet, so there is simply no physical way to drastically increase performance keeping the same litography without making the chip physically bigger and hotter and/or bumping cores Ghz to the limit, which is what intel is doing now on their cpus, or using improved components that keeps a marginal gain while keeping the same base blueprint, which is what apple is doing. M2 is basically M1X, "true" M2, in the sense of generational leap is gonna happen when they transition to 3nm and possibly ARMv9, that is when you will see huge performance improvements. M2 is M1 with better eficiency, slight bumped speeds, some bandwith improvements and more GPU cores.

 

[Spock]

it could be faster, but I don't think it's worth upgrading just yet.

Seems to be a point upgrade if you ask me, they should have named it the M 1.5 😅
Still disappointing that it will only support one external display.

 

[jav6454]

Yes. It really is faster.

And consume more power as well.

 

[headlessmike]

no no no its 33.333333%
i think he made how much is 8 compared to the 24 and not to 16

Want to see a magic trick?

24 is 50% more than 16, while 16 is 33.33333...% less than 24.

 

[theluggage]

Seems to be a point upgrade if you ask me, they should have named it the M 1.5 😅
Still disappointing that it will only support one external display.

Extra display support doesn't come for free - providing extra DisplayPort streams means more circuitry, more heat, more chip area and increases demand on the GPU and uses more of the unified RAM for video.

The regular M2 - like the regular M1 - is primarily aimed at ultra-low profile fanless designs, like the MacBook Air and iPad Pro, the vast majority of which sold aren't even going to have a single external display attached. Supporting more screens is part of the distinction of the M1 Pro (which will presumably be followed by a M2 pro).

 

[reallynotnick]

M2 is basically M1X

No, X chips didn't have updated aritectures they just had more CPU and GPU cores with increased memory bandwidth. The M1 Pro is what can be considered an M1X of sorts.

The A15 isn't "basically" an A14X.

I feel like people are getting thrown off because we are starting with low numbers like 2 is twice as big as 1 so the jump must be huge, but 15 is barely larger than 14 so a small update is expected, or something like that.

Not every generation is going to be an equal jump in performance.

 

[DearthnVader]

The M2 isn't leaps and bounds faster, it's an incremental improvement over M1. While as buyers and users of Mac's we want al the performance we can get for our money when we buy new, not many of us are on an 18 month upgrade cycle and those that are still want to maximize resale value of an M1 they may own now to help lessen the cost of the upgrade to M2.

If the M2 were leaps and bounds faster it would lessen the resale value of an M1 you may want to sell.

Also it would compete with the M1 Pro and M1 Max before Apple is ready with the M2 Pro or Max.

The normal M2 as released is a value system, you want leaps and bounds upgrade to the Pro or Max.

 

[darngooddesign]

Apple says that the 8 Core CPU (same number of cores) is up to 18% faster and the 10 core GPU is up to 35% faster. Two extra GPU cores wouldn't account for that much of an improvement. Its not a huge performance upgrade, but I don't think anyone expected it to be a huge difference. I bet the M2-Max is wicked fast tho.

 

[jav6454]

No, X chips didn't have updated aritectures they just had more CPU and GPU cores with increased memory bandwidth. The M1 Pro is what can be considered an M1X of sorts.

The A15 isn't "basically" an A14X.

I feel like people are getting thrown off because we are starting with low numbers like 2 is twice as big as 1 so the jump must be huge, but 15 is barely larger than 14 so a small update is expected, or something like that.

Not every generation is going to be an equal jump in performance.

The M2 has almost the exact same IPC jumps as the A14 (M1 base) and the A15 (M2 base) had when released.

 

[George Dawes]

I always said the M2 was going to be a performance let down

The M1 set the bar so high , it really was a techno great reset

 

[EugW]

it could be faster, but I don't think it's worth upgrading just yet.

If you already have M1, there isn't much point in upgrading to M2 unless you want the new form factor, you need 24 GB RAM, or you work in ProRes.

However, I think the better question is if it's worth it today for those upgrading from older machines to buy an M1 Mac vs. M1 Pro Mac vs. an M2 Mac. (All are available for sale from Apple.)

I always said the M2 was going to be a performance let down

The M1 set the bar so high , it really was a techno great reset

M1 is great, but M2 is even greater. The addition of ProRes acceleration and 24 GB RAM support are big deals IMO. However, if you don't need those features, then it's an incremental upgrade as otherwise expected.

 

[cbum]

TSMC 3nm process isn't ready yet

All these minimal incremental improvements... boring. I want them to finally break the barrier to a -1nm process!

That should give me results to questions I haven't even thought of yet. Can't wait.

 

[turbineseaplane]

Apple says that the 8 Core CPU (same number of cores) is up to 18% faster and the 10 core GPU is up to 35% faster.

Apple was also comparing GPU performance on Studio models to the RTX 3090

(which they don't really usefully compete with in all honesty)

They are wonderful at massaging narratives and coming up with often times odd comparisons to give them a favorable angle or metric to advertise.

 

[darngooddesign]

Apple was also comparing GPU performance on Studio models to the RTX 3090

(which they don't really usefully compete with in all honesty)

They are wonderful at massaging narratives and coming up with often times odd comparisons to give them a favorable angle or metric to advertise.

This is the footnote #4 for both those numbers. Apple was comparing the M1 and M2 13" MBPs. Unlike an RTX comparison the M1 and M2 are basically the same chip so there isn't an issue of whether a common benchmark is better optimized for the M2. Of course, it remains to be seen how the improvements pan out in real world use; but even if the difference isn't as pronounced, the M2 is still faster than the M1.

Testing conducted by Apple in May 2022 using preproduction 13-inch MacBook Pro systems with Apple M2, 8-core CPU, 10-core GPU, and 16GB of RAM, and production 13-inch MacBook Pro systems with Apple M1, 8-core CPU, 8-core GPU, and 16GB of RAM. Performance measured using select industry‑standard benchmarks. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.

 

[turbineseaplane]

This is the footnote #4 for both those numbers. Apple was comparing the M1 and M2 13" MBPs, not the MacStudio.

Testing conducted by Apple in May 2022 using preproduction 13-inch MacBook Pro systems with Apple M2, 8-core CPU, 10-core GPU, and 16GB of RAM, and production 13-inch MacBook Pro systems with Apple M1, 8-core CPU, 8-core GPU, and 16GB of RAM. Performance measured using select industry‑standard benchmarks. Performance tests are conducted using specific computer systems and reflect the approximate performance of MacBook Pro.

"oh"

Sorry, I didn't read the footnotes
I suspect most don't, and Apple knows that when crafting often misleading marketing headlines and charts.

 

[darngooddesign]

"oh"

Sorry, I didn't read the footnotes
I suspect most don't, and Apple knows that when crafting often misleading marketing headlines and charts.

Why are you making performance counter-claims if you don't even understand what is being compared?

 

[turbineseaplane]

Why are you making performance counter-claims if you don't even understand what is being compared?

I didn't really need footnotes to know that nothing about Apple claims should even mention a RTX 3090..

One BIG reason for that, that isn't even technical, is the types of usages for a 3090 aren't even something you can do on macOS

1. Because a lot of it is Windows gaming
2. Because you can't even use a 3090 with macOS to do anything anyways

My point is that it was a nonsensical comparsion to be making by Apple

Let's move on - we are going nowhere here

 

[darngooddesign]

I didn't really need footnotes to know that nothing about Apple claims should even mention a RTX 3090..

One BIG reason for that, that isn't even technical, is the types of usages for a 3090 aren't even something you can do on macOS

1. Because a lot of it is Windows gaming
2. Because you can't even use a 3090 with macOS to do anything anyways

My point is that it was a nonsensical comparsion to be making by Apple

Let's move on - we are going nowhere here

You realize that this thread is about the M2 vs the M1, not the M2 vs the RTX. Why would Apple compare its general purpose, efficiency-optimized M2 chip against a high performance power-hungry RTX? Apple will make those kinds of comparisons with the M2-Ultra.

 

[deconstruct60]

So am I reading this correctly -

https://www.apple.com/macbook-air-m2/

Apple tested a M2 Air with 24GB Ram and 8 Core CPU and 10 Core GPU, against an M1 Air with 16GB of Ram, 8 Core CPU and 8 core GPU, and are claiming that the M2 is up to 1.4x faster than the M1?

M2 memory speed is 50% faster. (the Max/Pro/Ultra already were faster). So anything that is doing large bulk memory transfers is likely going to be faster than the M1 even if there was no difference in the cores on each SoC . So with even GPU cores there would still be a difference. Won't be 1.4x faster. but also extremely likely not going to be zero faster. It is a number small enough that Apple doesn't want to brag about it on their sales marketing page.

Apple did a multi thread CPU demo at the keynote. That was probably a highly 'cherry picked' metric also but was in the mid teens ( 15-18% ) range. Not huge but also not zero either.

There is no huge jump in single thread metrics. This is more of a just do a safe, incremental upgrade the SoC. Apple isn't trying to make folks who relatively recently bought an M1 throw it away and by an M2. It is more aimed at the folks who are still on Intel ( on macOS or Windows) to jump to the "new world". It isn't a "version 1.0" product. It likely has some issues work out, that the M2 users will skip.

The charts on the marketing page that are more relevant to sales are the ones Apple makes against the intel MBA model. Those all look really good. The ones against the M1 are much more narrowly targeted. Anyone who does video that ProRes difference is game changing. Apple is warming up to targeting more casual gamers with Macs so that "fair fight" on GPU count.... for those folks who cares. They are not looking for a "parity" they are looking for higher frame rates with the limited budget they have. That is it.

 

[deconstruct60]

The point I haven't come across yet is the poor "optics" of the situation...

18% increase over two years?

The M2 is probably late. The model ids for these two laptops leaked out back in March. Does that cut it down to exactly one year? No. But it appears to be the same baseline tech that is in the A15 which did appear one year after the A14/M1 appeared. It is being a shipped on a node that TSMC started producing back in Q4 2020 also (at risk production at that point but shipping. Volume was in middle 2021. So basically the fab process is about year also also). There is few bleeding edge options about this chip.

It is more the case that it is trickling out to these Macs in a delayed fashion, that what the Silicon team is doing.

Apple Silicon team is spread out over more dies now. Supposedly the AR/VR system also has a custom SoC in it. The Watch hasn't moved in a long while now (still on N7 node and A13 E cores , while Mac is moving to year three on N5. ). Honestly, not really surprising. Part of Apple's silicon sprint to the lead was that they were doing far , far , far fewer products that most of the major competitors were. Make a new A-series SoC and toss it into 5 products. While other folks are making 10 different SoCs. As the product front gets broader it is going to be harder to keep a sprint pace in all directions. Throw on top the pandemic and not sure what "optics" would expect large jumps.


That's said it is a SoC not a uni-dimensional "Just a CPU". Apple isn't not far behind on primary CPU cores performance. So most of the additional transistor budget did not go there. So 18% isn't really all that bad. Apple got bigger performance/watt bang for the buck in other areas ( e.g. Video de/encode. ). They are not trying to build a single thread "hot rod" drag racing engine. The top proirity goal is perf/watt. That is where they spend most of their time and resources. Single and mulithread improvements fall out of that focus. but those are not the top priority goals.

Yes, I know about A14 v A15 (or A16), and the number of cores, and the chip shortage, etc -- I'm just talking about the optics of the presentation. It "appears" that Apple has only been able eke out a 9% improvement YoY in actual core speed of their Apple Silicon.

Intel would be ridiculed for this type of announcement. And both sides would be partially right.

That is really more about the myopathy of the narrow focus on CPU core performance that the optics breadth of what Apple is presenting. That is more an issue with the viewers of the scene than of the scene itself.

Again, I know that it will improve with A16 or A17 (whichever the M3 is based on), even more cores, etc. It just looks bad given the timing and the compromises they had to make to get the M2 launched along with their hardware schedule.

Somewhat doubtful Apple is going to push CPU P core count way higher for the 'plain' Mx products going forward. The P and E cores will get more budget and bigger caches but unlikely Apple is going to get into push the 'core count wars' down lower into the product line. As the memory bandwidth gets higher the P core might get a bigger share of the bandwidth, but it is still going to be balanced against the demands of the other cores. Apple will also more likely throw bigger budget at AMX (which is in the P core complex but not a user visible 'core'. ). Big picture lots of new transistor budget spent on more specialized compute than on very generic compute that delivers much better perf/watt.

Also moving the data being worked on closer to the processor ( or vice versa) is better than just blindly cranking core count.

The M2 is still not Thunderbolt 4 in part because it isn't complete on display processing. Should fix that when have more transistor budget before moving on to more CPU cores.

 

[AltecX]

People here complain about 18%/35% when normally Intel's upgrades are like 8-10% CPU MAYBE 15% on a GPU.

Even if on average the M2 is 8-12% faster in most tasks and "only" 15-20% on GPU that's still a very solid improvement for a second iteration of what is essentially a new product.

 

[Misheemee]

People here complain about 18%/35% when normally Intel's upgrades are like 8-10% CPU MAYBE 15% on a GPU.

Even if on average the M2 is 8-12% faster in most tasks and "only" 15-20% on GPU that's still a very solid improvement for a second iteration of what is essentially a new product.

I’m not complaining at all - I just wonder whether they should be marketing that it’s 40% (or 1.4x) faster than it’s predecessor, or whether the number needs to be more along the lines of 18% faster (or 1.18x)

 

[darngooddesign]

I’m not complaining at all - I just wonder whether they should be marketing that it’s 40% (or 1.4x) faster than it’s predecessor, or whether the number needs to be more along the lines of 18% faster (or 1.18x)

Like adult diapers it depends. Some tasks 18% faster, while other tasks are 40% faster.

 

[EugW]

I’m not complaining at all - I just wonder whether they should be marketing that it’s 40% (or 1.4x) faster than it’s predecessor, or whether the number needs to be more along the lines of 18% faster (or 1.18x)

The 40% faster is likely true... for some ProRes workloads as they indicate, cuz M2 includes a ProRes hardware accelerator that is absent with M1.