Scaling based on node size has been limited for a long time. It feels like it started at 13nm that every shrink was getting less and less performance uplift, but it's likely it's been going on for longer and we just had much more room for improvement on chip design.

It's a mixture of several problems:

* TSMC N3B being a bit of a flop (yield issues, too expensive)

* Brain drain from Apple's chip design teams over the last few years

* Tim Cook trying to push the average selling price up to keep revenue growth going in the face of sales declines (e.g. hobbling memory bandwidth, reducing the number of performance cores for M3 Pro)

I don't expect there to be a M1 style generational leap for a long time, expect 2010s Intel style yearly performance gains from here on out.

The GPU is increasing a lot though.

Another point is the CPU improvements are constant. In about two years the base M3 is now on par with the M1 Max.

The GPU is critical to future profits though. Apple really wants your monthly subscription to Apple Arcade and they want to expand in other game areas which is why they've been paying AAA companies to optimize for Mac. This also ties into their VR headset where gaming will be one of the core features.

Hardly any AAA companies are optimising for the Apple GPU. MoltenVK is where all the interest is.

Even if they did, there is very little in Apple Arcade which taxes the GPU, most target the lowest common denominator in terms of supported iOS/phone combinations.

The original Apple Arcade strategy was for AAA titles, but for whatever reason that wasn’t pursued, so now we have a tonne of casual games and re-releases of old titles.

Apple just seems to run hot and cold on gaming.

> The original Apple Arcade strategy was for AAA titles

Can you prove that statement? I bet you can't, cause it's simply not true. Every Arcade title is playable from an iPhone to a Mac by way of Apple TV. It was never going to get AAA titles.

I mean, they keep putting desktop class CPU+GPUs in their mobile devices - all just to play an Angry Birds remake?

I own almost every Apple device in the ecosystem, but I never game on them…

Here you go. A simple google would have found that.

Apple wants more Grindstones.

https://9to5mac.com/2020/06/30/apple-arcade-strategy-shift/

How is cancelling contracts with some developers equal to they wanted AAA titles but messed it up.?

There’s no correlation.

It’s true but the problem with the GPU is Apple’s addiction to RAM money. Yes the GPU is improving in performance but it does you no good if it has to share a tiny amount of system RAM with the CPU.

Tiny? My M2 Air has 24g, which is more ram than any other laptop video card I've had.

Sure, if you pay for it. The 14” starts at 8GB. Eight! 200$ more for another 8GB.

> The GPU is increasing a lot though.

Is it? Does it matter? Unified RAM is cool. But I’d rather have an Nvidia 4090.

In a laptop?

I was thinking about the desktop variants.

Apple Silicon is amazing. But if you’re doing heavy GPU it’s not very exciting.

It's very exciting in the local LLMs space, where the unified memory allows fast inference of large models.

I read that the bandwidth is 50% lower on the M3 combined with a lower CPU core coun. This reduction will impact inferrence performance. It maybe better to stick with M2 series if you really really think spending $10k on a laptop mac to do inferrence slowly (but faster than a plain PC of course) make sense.

Weren't all of the new M3 Macs announced the same price as they previously were or lower? Same with the recently announced iPhones? Or am I mis-remembering? Seems like prices not increasing given all the recent inflation are actually a price decrease pretty much across the board not an increase on the average selling price?

Entry level pricing for a MacBook Pro now starts at $1599 rather than $1299.

And it has a downgrade from a Pro chip to a non pro chip.

The 13" MBP only came with a base M2 chip - but once you spec 16 GB RAM on the 14" M3 MBP, the 14" M3 Pro MBP is "only" an extra $200 for a more powerful chip, 18 GB RAM, and better I/O (TB4, ability to run more than one external display).

Apple are masters of product segmentation and upselling.

13" Macbook Pro had M2 not M2 Pro

No it doesn’t. The previous (cheaper) entry level MBPs were non-Pro M2 (and non-Pro M1 before that).

That thing had the same CPU as the MacBook Air. There was nothing "pro" about it. I'm so happy it's finally gone!

the new one also has the non-pro, non-max M3.

> expect 2010s Intel style yearly performance gains from here on out

Intel saw very little gain this year at all in return for a 400 watt power draw under load.

The plain old M3 saw a 20% performance gain along side efficiency gains.

Having a 22 hour battery life is insane and you certainly aren't going to manage that with a 400 watt power draw.

400 watts is on a desktop chip where there is no concept of battery life.

20% increase on performance is compared to M1 not, M2 - which also had 20% increase in performance on M1.

> 20% increase on performance is compared to M1 not, M2

Nope.

> The M3 chip has single-core and multi-core scores of about 3,000 and 11,700, respectively, in the Geekbench 6 database. When you compare these scores to those of the M2's single-core and multi-core scores (around 2,600 and 9,700, respectively), the M3 chip is indeed up to 20% faster like Apple claims.

https://www.laptopmag.com/laptops/macbooks/apple-m3-benchmar...

> 400 watts is on a desktop chip where there is no concept of battery life.

Yes, and in exchange for that ridiculous 400 watt power draw, Intel saw negligible performance gains.

> In some areas, the extra clock speeds available on the Core i9-14900K show some benefit, but generally speaking, it won't make much difference in most areas.

https://www.anandtech.com/show/21084/intel-core-i9-14900k-co...

Intel only wishes they could hit a 20% gain in exchange for all that increased power draw and heat. As that review noted the best improvement they saw in any of the common benchmarks was just 6%.

I wonder how much thermal throttling is going on with these benchmarks? 400W seems ridiculously difficult to cool. The 13900 was difficult if not impossible to cool for throughput without water cooling.

For Apple, yes incremental increase but not for Huawei and recent Elite X.

Apple seems to be focusing on efficiency more than anything, and maybe a little more market segmentation where the lesser chips are less competitive with the bigger chips. The Pro offers fewer performance cores, trading them for efficiency cores. It has reduced memory bandwidth.

From a generational perspective, the M3 Max is offering the same level of performance as the M2 Ultra. That's amazing as the m3 max is 12p/4e vs 16p/8e in the m2 Ultra. The M3 Ultra should be a substantial lift.

It's a tick and some people... especially marketing departments want to sell it as a tock.

Wait, is that the sound the minute hand makes?! I never realized.

Assuming you don’t have a single-hand clock.

I don't know where you get the idea of small performance increase. What I'm seeing until now, also in pre-review unites, is actually the opposite, especially when ray tracing and mesh shading get taken into account. For anything GPU related these new machines are a giant leap.

Are these mainstream use cases? These always feel like they are chosen to demonstrate the processor's strengths, instead of talking about the ways the processor helps in real applications.

Is playing games with insane light effects and mesh shaded scenes on a family computer mainstream enough?

Not really? What are some examples of said games, which are available on macOS?

Stray, Resident Evil Village on iOS, Myst, Lies of P, and a long series of other games that will potentially follow because of this upgrades, and because Apple wants to make iOS - iPadOs - macOS a giant unified cross device platform powered by Metal and proprietary silicon that offers hardware accelerated ray tracing and mesh shading across the entire lineup.

In case you didn’t notice, A17 Pro on iPhone 15 has hardware accelerated ray tracing as well.

Isn't it a 15% increase? I really don't consider any double digit percent increases to be small.

Actual IPC increase is 1-2%. The rest is from ramping the clockspeeds. This is a problem because power consumption goes up exponentially with frequency. Go up too high and they'll be doing what AMD or Intel does where a single core is using 50+ watts to hit those peak numbers.

Power is polynomial (it goes with the square of frequency), not exponential.

How is this a problem? It just looks that they didn't try to improve the general architecture but make one step to ramp up the clock speeds without increasing the power consumption thanks to the process step. Which is a great achievement, because any tape out on a new advanced process - here for the first time a "3nm" - is a big achievement. One has to consider that Apple now has yearly updates in its processor lineup. The next step will probably introduce more architectural changes. Only if those would stop showing up for several years in a row I would get concerned.

With each recent node step, you basically get +10-15% clockspeed or -30% power consumption. They just blew their entire node on a small clockspeed ramp.

Now, if they want a wider core for M4, that means more transistors and more heat. They are then forced to: not go wider, decrease max clockspeed, hold max clockspeed for a pitiful amount of time, or increase power consumption.

On the whole, I'd rather have a wider core and lower clockspeeds then turn the other power savings into either battery life or a few more E-cores.

> Isn't it a 15% increase?

At least according to Geekbench, it's a 20% performance increase.

> The M3 chip has single-core and multi-core scores of about 3,000 and 11,700, respectively, in the Geekbench 6 database. When you compare these scores to those of the M2's single-core and multi-core scores (around 2,600 and 9,700, respectively), the M3 chip is indeed up to 20% faster like Apple claims.

https://www.laptopmag.com/laptops/macbooks/apple-m3-benchmar...

Along side a battery life increase to 22 hours? It's been a pretty good showing.

Benchmarks are optimized for specifically.

Are you alleging that a chip which was in development for years was optimized specifically for a benchmark that was released a few months ago?

The surface area of the M3 Pro vs M2 Pro tells you everything you need to know.

have a handy link and few more sentences for my friend who does not gets it?

Search "M3 Pro" on Twitter and you'll see some decent comparisons.

What's Twitter?

you made the assertion so you provide the link, I am not doing this do your own research BS

Lol it's not my job to inform you. If you care enough do your own research.

Everytime Apple oversells there are apologists putting a spin on it. Nothing new to see.

And there are detractors downplaying improvements. Nothing new to see.

10-15% increase in single threaded code every year is pretty good these days.

I heard from a a hardware developer that the main wins for the M* architecture is really that they integrated the memory onto the cpu die and then the unified memory architecture between the GPU and CPU which was "holy grail" type stuff a decade ago.

I don't think we're going to see the performance increases past the ~10-20% each fab cycle now.

Performance per watt and having dedicated HW for video decoding / other tasks(AI?) I think might end up being the answer to the continuation of Moore's law. Especially considering that we are in the era of heterogenous computing.

The memory is not on the die. It's in the package just like pretty much every other big boy mobile ap out there.