Trick or M3-treat? - Apple’s pre-Halloween “Scary Fast” virtual event

[captain_to_fire]

i9-13900KS remains slightly faster than the M3 Max in both single and multi-threaded tasks but at the expense of 150W, the M3 Max is almost as fast as the i9 but at around 30-45W TDP.

 

Meanwhile, this is the Snapdragon X Elite SKU they used to compare against the M2 Max. So yeah at the moment, the M3 Max remains as the fastest consumer ARM chip...at the moment.

Spoiler

 

[hishnash]

7 minutes ago, captain_to_fire said:

150W, the M3 Max is almost as fast as the i9 but at around 30-45W TDP.

And that 45W is full SOC power draw.. an i9-13900KS will also have chipset power draw memory etc...

[saltycaramel]

https://browser.geekbench.com/v6/cpu/3350282
 

21K 

 

Those twelve 4GHz P-cores make the M3 Max destroy the M1 Ultra and be almost on par with the M2 Ultra in multithreaded workloads. A Max almost on par with Ultras from 5 months ago. In a laptop.
 

I wonder if the A17 Pro and the M3 still use the “Everest” P-cores from the A16.

 

This would make the M3 a “tick” generation on the CPU front. Same architecture, 5nm>3nm shrink.

 

The M4 would be a “tock”: same 3nm, new CPU cores.

 

(The GPU front is a completely different story: this is the year of a new GPU introduction, the sixth gen Apple GPU on both the A17 Pro and the M3.)

 

 

edit: to be clear, if the M3 CPU was based on Everest/Sawtooth cores, it would be a “tick” only compared to the A16. Compared to the M2 (5nm, Avalanche/Blizzard cores from the A15) it would have both a new architecture and a new fabbing node. More appropriately it’s the A17 Pro being a “tick” compared to the A16. 

[hishnash]

1 hour ago, saltycaramel said:

I wonder if the A17 Pro and the M3 still use the “Everest” P-cores from the A16.

 

Given they have a new (9 wide!) decoder and new branch predictor it is clear they are not A16 cores... new decoder is more than enough to consider it an arc change. 


 

 

1 hour ago, saltycaramel said:

if the M3 CPU was based on Everest/Sawtooth cores

Everything is always based on the item before, you could look at Intels 14th gen (13th gen) and say it is based on their 5th gen cpus.. there is commonalty there. Every year you update some bits and replace others, the idea that you throw it all away every year and start from scratch is absude, even Zen had bits and IP from bulldozer it was not completely new. 

 

1 hour ago, saltycaramel said:

(The GPU front is a completely different story: this is the year of a new GPU introduction, the sixth gen Apple GPU on both the A17 Pro and the M3.)

 

GPU changes happen much more often as you are not expect or required to provide any instruction set continuity as apps ship with byte code that the driver targets to the HW in question so Gpu engines are free to make drastic changes if they want... even through however these GPUs are not all new, they will not have re-desigend everything form the ground up that would be stupid and would result in a shit design.  Throwing away what is working well just to have a marketing win and say "All new" is not what happens. 

 

[Dracarris]

On 11/1/2023 at 12:12 AM, leadeater said:

Think I'll save ~$900 and get the ASUS Zenbook 14X OLED 2880x1800 120Hz HDR600.

Such a display and an Aluminium/Magnesium chassis for under 1K - holy nuts. Is there any catch? This sounds a bit too good to be true.

[leadeater]

50 minutes ago, Dracarris said:

Such a display and an Aluminium/Magnesium chassis for under 1K - holy nuts. Is there any catch? This sounds a bit too good to be true.

The chassis while strong doesn't have good edge/surface finish, the Asus ZenBook Pro is noticeably better in that regard. And of course there is the battery life issue being 13500H/13700H CPU. The ZenBook 14 (not X) with Ryzen 7000U has nearly double the battery run time, lower performance of course too being 15W U series.

 

But there is a reason a lot of the reviewers were raving over the ZenBook 14X though, for the price it's damn good but it's not perfect in every area.

 

What I don't like about it and basically every other $1500+ laptop is having a dGPU while makes it great performance in a lot of apps you can't really use it well on battery so it functionally becomes a device that can go between wall outlets for those tasks. So I'd never option it with the dGPU myself as I don't need it and I'll pass on the tradeoffs of simply having it at all.

 

Oh and like many laptops the sound only works properly when it's on a table, Macs don't have the requirement.

[Dracarris]

21 minutes ago, leadeater said:

- snip -

Okay so it's basically comparable to current Macbooks in terms of screen quality and to some extent, that applies to the chassis?

 

Earlier you said, screen and chassis are not the most important differentiation points when it comes to Macbooks. Which (similarly) priced machine(s) in your opinion actually is comparable in both aspects (still ignoring battery life, trackpad, speakers, maybe keyboard)?

[Obioban]

Looks like M3 Max in multi thread is performing about on par with the M2 Ultra in multi thread. Pretty nuts!

 

M3 Ultra is going to be no joke... but imagine if they actually are able to pull off the quad this cycle.

[leadeater]

3 hours ago, Dracarris said:

Okay so it's basically comparable to current Macbooks in terms of screen quality and to some extent, that applies to the chassis?

 

Earlier you said, screen and chassis are not the most important differentiation points when it comes to Macbooks. Which (similarly) priced machine(s) in your opinion actually is comparable in both aspects (still ignoring battery life, trackpad, speakers, maybe keyboard)?

The question asked about the chassis was and I quote "flexless sturdy unibody aluminum build?", surface finish and edge feel are different aspects that some may or may not care about. A product line starting at nearly $1000 is obviously going to be not as good in areas but if you just want to cover off the core things that make a laptop good and function very well you don't have to go past that laptop. I'd still like I said prefer to wait for the updated Ryzen 7000 option since that is coming for it, like the ZenBook 14 (non X) has both options.

 

But above all like I said I'd rather save $900 than get things I don't need, don't care about, and are not really that important. Speakers are just as good as MacBook Air so long as it's on a table yet I would choose to respect people around me and use headphones and that applies to any Mac as well, trackpads are good and the days of MacBooks being way better are long gone and the same for keyboards. If you actually want a critical review of the ZenBook 14X against any MacBook I can write one if you like but I can save you having to read that and give you the conclusion, since I do not do any video editing or audio production I will save the $900 thanks. While I don't feel the ZenBook 14X is computationally equivalent to MacBooks without a dGPU (which you can get in the budget) you have to actually be using those applications and more than just in passing for me to weight it highly.

 

The biggest problem is sticking to the 120Hz, if I were to go with 90Hz then higher quality and higher brightness OLED options become possible but I felt the screen spec listed were more important to the point than a nicer edge finish on the chassis. I also didn't feel going up around $100 over the budget, which there are some very good options there, would help much either and they all from what I remember come with dGPUs which is most of the cost increase rather than going in to other areas. The Razer Blade 14 with the 6900HS is a good example of a higher quality platform but it's that $100 more than asked.

 

The Lenovo Slim Pro 9i would be another that is very comparable other than it's 16" (the 14" is updated to OLED now and not as bright as the ZenBook 14X), has an RTX 4060, and the screen is more gaming focused (16" 3.2K (3200 x 2000), Mini-LED, Glare, Touch, HDR 1000, 100% Adobe RGB, 100%DCI-P3, 1200 nits, 165Hz) also $1749.

 

Pretty good review of the ZenBook 14 and 14X

And a general covering the "best" 14 inch laptops

 

Edited November 2, 2023 by leadeater

[captain_to_fire]

54 minutes ago, Obioban said:

M3 Ultra is going to be no joke... but imagine if they actually are able to pull off the quad this cycle.

The ultra line remains as two max chips linked together with a 2.5TB/s low latency interposer (or as Intel calls it their version an embedded multi-die interconnect bridge/EMIB). The rumors of a quad chip started way back since the M1 ultra. But some leakers say that the reason Apple didn't do a quad chip for the Mac Pro is apparently that simply adding more chips didn't scale the performance and power consumption to what Apple wants, maybe they're trying to achieve a similar performance as a 4th gen EPYC but at 1/2 of the power but Apple can't deliver so they probably said "fuck it" and put an M2 Ultra on a Mac Pro.

 

So yeah the Mac Studio basically killed the Mac Pro. In essence, the Mac Studio is what the the 2013 trash can should've been.

[Obioban]

9 minutes ago, captain_to_fire said:

The ultra line remains as two max chips linked together with a 2.5TB/s low latency interposer (or as Intel calls it their version an embedded multi-die interconnect bridge/EMIB). The rumors of a quad chip started way back since the M1 ultra. But some leakers say that the reason Apple didn't do a quad chip for the Mac Pro is apparently that simply adding more chips didn't scale the performance and power consumption to what Apple wants, maybe they're trying to achieve a similar performance as a 4th gen EPYC but at 1/2 of the power but Apple can't deliver so they probably said "fuck it" and put an M2 Ultra on a Mac Pro.

 

So yeah the Mac Studio basically killed the Mac Pro. In essence, the Mac Studio is what the the 2013 trash can should've been.

Not succeeding on the quad on the M1/M2 doesn't mean they never will-- and it would give the Mac Pro a... purpose. The fact that the Mac Pro continues to exist gives me hope that they think they need it for the thermal headroom a potential future quad chip. As the Mac Pro sits today, it's niche to the point that it seems like it doesn't really need to exist, which is very much not something Apple traditionally does.

[leadeater]

7 minutes ago, captain_to_fire said:

maybe they're trying to achieve a similar performance as a 4th gen EPYC but at 1/2 of the power but Apple can't deliver so they probably said "fuck it" and put an M2 Ultra on a Mac Pro.

It's a lot harder to match the power efficiency of EPYC in scaled out core counts, those products are power efficiency focused unlike the desktop CPUs and don't even allow as high boosting as many of the laptop ones either. AMD has 64 core options in 155W (Zen4) and 225W (Zen3), but these offer no GPU at all where Apple would be. Personally I think Apple took a step back and was evaluating if scaling out past 2 even made sense at all for the product, I don't think they ever were actually looking at doing 4. Maybe in some side thinking but I can't imagine it was considered that much.

 

4 chips would make an exceedingly lager package even without the memory, from what I understand something like that simply isn't viable just due to the physical size.

[igormp]

44 minutes ago, Obioban said:

Not succeeding on the quad on the M1/M2 doesn't mean they never will-- and it would give the Mac Pro a... purpose. The fact that the Mac Pro continues to exist gives me hope that they think they need it for the thermal headroom a potential future quad chip. As the Mac Pro sits today, it's niche to the point that it seems like it doesn't really need to exist, which is very much not something Apple traditionally does.

Well, you just need to wait for any dev to get their hands on a M3 and see if there's enough IRQ controllers for a quad-chip solution or not.

[Paul Thexton]

1 hour ago, igormp said:

Well, you just need to wait for any dev to get their hands on a M3 and see if there's enough IRQ controllers for a quad-chip solution or not.

If there's any clue at all that it's a possibility my guess is Hector Martin will be among the first to let everyone know.

[igormp]

32 minutes ago, Paul Thexton said:

If there's any clue at all that it's a possibility my guess is Hector Martin will be among the first to let everyone know.

Yeah, before posting that I checked his mastodon profile to see if he had anything lol

[saltycaramel]

https://www.tomshardware.com/software/macos/apple-spent-dollar1-billion-to-tape-out-new-m3-processors-analyst
 

Quote

Perhaps the most significant facet of the M3 launch was the fact that they are launching three chips at once. Our initial assumption was that these were the same chip, ‘binned’ in production, but that was wrong. These are three different chips, with the differences manifest in their very different sizes of each and visible to the naked eye in the die photos above. This is somewhat staggering, as we have to assume that tape out costs alone for the three has to be close to $1 billion. Very few companies can afford this large an undertaking.

The M1 Pro was just a cut down M1 Max.

The M2 Pro was just a cut down M2 Max.

 

This time around the M3 Pro is its own thing.

And it’s been rebalanced to be more skewed towards e-cores. (6p+6e or 5p+6e, instead of 8p+4e and 8p+2e of the previous generations)

 

Why is that?

Why was it deemed worth it to be ad-hoc designed this time around?

Why was it made slightly smaller/cooler/less_poweful than its predecessors?
 

Are we gonna see some new products that will use the new&smaller M3 Pro? (A gaming oriented AppleTV Pro? A redesigned slimmer MacMini? The base iMac Pro? Or, heck, even the highest end 13” iPad Pro? Maybe a downclocked M3 Pro could fit in a fanless design?)

 

Or maybe this repositioning of the M3 Pro is simply meant to make the upsell to the M3 Max more enticing? 

[Paul Thexton]

8 minutes ago, saltycaramel said:

Or maybe this repositioning of the M3 Pro is simply meant to make the upsell to the M3 Max more enticing? 

That's what my money is on tbh.  Either that, or maybe the M3 Pro battery life is going to be even better again due to it being scaled back a bit, and that's the market they're trying to amplify their (perceived, at least) advantage in?

[saltycaramel]

On 11/1/2023 at 3:56 PM, saltycaramel said:

 

Trust the process, there's a reason Apple wouldn't touch single-stack OLEDs for large displays with a ten-foot pole and there's a reason they'll suddenly move their entire iPad/Macbook line-up to dual-stack OLEDs in the span of 2-3 years. We'll know more about it (in the form of shiny marketing speak) come next spring.

I need to correct this assertion of mine since I’ve stumbled upon a post from venerable display analyst Ross Young that contradicts it: while the 2024 iPad Pro will use a dual (tandem) stack OLED, there will actually be a 13.4”-large single stack OLED laptop display from Apple next year.

 

 

 

Now, if this was a MacBook Air, we’d be hearing supply chain rumors about a 13.6” and a 15.3”. 
 

But the rumored diagonal of this OLED panel is 13.4” (I know the difference compared to 13.6” sounds small, but Apple is not changing the new Air chassis design after just 2 years..) and there are no rumors about a 15” model.

 

Could this be the under-1000$ “Chromebook rival” MacBook SE?

Or a “remake” of the beloved 2015 12” MacBook?

[saltycaramel]

22 minutes ago, Paul Thexton said:

That's what my money is on tbh.  Either that, or maybe the M3 Pro battery life is going to be even better again due to it being scaled back a bit, and that's the market they're trying to amplify their (perceived, at least) advantage in?

It could certainly allow battery life aficionados to attempt world records in battery life ‘cause with 6 efficiency cores (and of the M3 kind at that) one could even rationalize staying in Low Power Mode at all times and not even “feel” the performance hit (for light general usage)..

[hishnash]

48 minutes ago, saltycaramel said:

Why was it made slightly smaller/cooler/less_poweful than its predecessors?

Form a direct perfomance the M3 pro is still faster than the M2 pro just not as much aster as it would have been had to been just a cut down Max.

As to why I have wandered if the spacing of component islands on the max is a hint to this.. there is an optimal limit for mask projections on any node, maybe N3B has a limit that lead the Max to require multiple seperate islands of projections (rather than a single projection).. if you look at the die shot the max has a LOT of void space between active silicone were the M3 and M3 Pro do not... 

This is not only using up a lot of 3nm wafer $$$$ but if it requires doing seperate projections for each island of logic is going to involved a lot of extra stages in projection $$$$, the max can absorb the increase per die cost of this but maybe a pro chip with that same method would just not be economic.  3NM does not a Lot more than the older 5nm nodes. 

[saltycaramel]

10 minutes ago, hishnash said:

if you look at the die shot the max has a LOT of void space

(I live for nerdy die analysis of those shots but on the other hand I also assume the die shots are intended as “die beauty shots for illustration purposes that may or may not be slightly photoshopped” and Apple is under no obligation of actually showing every last bit of their super-secret IP)

[hishnash]

1 minute ago, saltycaramel said:

(I live for nerdy die analysis of those shots but on the other hand I also assume the die shots are intended as “die beauty shots for illustration purposes that may or may not be slightly photoshopped” and Apple is under no obligation of actually showing every last bit of their super-secret IP)

I agree but I don't see why they would delete stuff on the max between all the islands of what appease to be active silicon so uniformly, unless these games are something to do with the chips fabric and the Max chip has such a more robust internal fabric that it needs all that space?

[HenrySalayne]

4 hours ago, saltycaramel said:

Apple is under no obligation of actually showing every last bit of their super-secret IP)

If you can gather super-secret IP from a die shot, you have a skill worth millions maybe billions.

[captain_to_fire]

Looks like the embargo has been lifted. My YT recommendations are being saturated by M3 Mac reviews.

[Obioban]

On 11/3/2023 at 6:54 AM, saltycaramel said:

https://www.tomshardware.com/software/macos/apple-spent-dollar1-billion-to-tape-out-new-m3-processors-analyst
 

The M1 Pro was just a cut down M1 Max.

The M2 Pro was just a cut down M2 Max.

 

This time around the M3 Pro is its own thing.

And it’s been rebalanced to be more skewed towards e-cores. (6p+6e or 5p+6e, instead of 8p+4e and 8p+2e of the previous generations)

 

Why is that?

Why was it deemed worth it to be ad-hoc designed this time around?

Why was it made slightly smaller/cooler/less_poweful than its predecessors?
 

Are we gonna see some new products that will use the new&smaller M3 Pro? (A gaming oriented AppleTV Pro? A redesigned slimmer MacMini? The base iMac Pro? Or, heck, even the highest end 13” iPad Pro? Maybe a downclocked M3 Pro could fit in a fanless design?)

 

Or maybe this repositioning of the M3 Pro is simply meant to make the upsell to the M3 Max more enticing? 

 

I suspect the M3 Pro is better for the reality of how people that buy the pro level chip actually use their computers.

 

As in, looking at my Activity Monitor, it's shocking how infrequently the P cores even fire up-- whereas the E cores are always doing something. Having more and faster E cores likely improves the actual experience notably, and always.

 

Conversely, The Max chip does come with compromises-- more P cores means more battery usage and more ram means more battery usage. With the price jump from Pro to Max, likely the people that actually need the P cores are the exact ones that'll be buying them, and people getting the Pro level chip will get a chip that's actually better suited to their needs than the older M1/M2 Pros.

 

... Or maybe, as you suggest, they want something less power consuming/heat generating for the Vision Pro. Personally, I think it's far more likely that it just gets an M3, instead of the previously announced M2, but this new pro chip does make it seem more viable for them to use it there.