M3 Chip Generation - Discussion Megathread

[Mac_fan75]

I think this makes a lot of sense from the marketing perspective. They need to generate a lot of hype and get developers on board to ensure that there will be enough high quality software compatible with their platform on release. Right now Windows on ARM is in somewhat sad state, so releasing a high-end product outright can be very risky. Qualcomm seems to be playing it by targeting the creatives first, which is a smart move, as there is a limited number of players in the field and their hardware will work well for this kind of work.



It most certainly won’t be. Qualcomm’s target market are premium laptops. I also doubt that they will be cheap. They will be competing directly with Apples MBP. I think Qualcomms target customer is a creative or a tech enthusiast/developer who wants a fancy fast energy efficient laptop but prefers Windows.

I am curious to see if their strategy will work out. They’d certainly do fine against M2 Pro, but I wonder how good their value proposition will be against M3?

I think the M3 single core will disappoint a bit and scary fast is the multicore of the new M3 Max because of the extra cores i’m afraid.

Will those new Q chips not be available in the Microsoft hardware? Might again boost them to push more ARM stuff for windows but yeah if it’s up market prolly not

 

[APCX]

I think the M3 single core will disappoint a bit and scary fast is the multicore of the new M3 Max because of the extra cores i’m afraid.

Will those new Q chips not be available in the Microsoft hardware? Might again boost them to push more ARM stuff for windows but yeah if it’s up market prolly not

Why would the single core disappoint?

 

[leman]

I think the M3 single core will disappoint a bit and scary fast is the multicore of the new M3 Max because of the extra cores i’m afraid.

I would think that the M3 will be at least 10% faster than the iPhone15 Pro. That alone will make it pretty much the fastest CPU core on the market.

 

[MRMSFC]

Does it seem like this snapdragon processor is following the procedure that Intel typically does?

Lots of PR about the chip itself and a continuous stream of PR as that ship gets further and further along but it will be six months to a year before there’s an actual consumer product that hosts the chip.

Also, do you think they will be following the current x86 procedure that when (and if) a commercial board is available, will this chip be mounted via a socket?

The technology market has so much bs graphs around now it takes ages to determine what they’re trying to say anymore, and Apple is guilty too.

I call it “Marketing through obfuscation”

As for whether it will be socketed, I very much doubt it. It’s apparent that Qualcomm is targeting the laptop segment, which uses socketed cpus rarely. Usually that’s more of a desktop thing.

 

[Confused-User]

Also, do you think they will be following the current x86 procedure that when (and if) a commercial board is available, will this chip be mounted via a socket?

I don't know if that was addressed but it's very unlikely. That's not actually "current procedure" at all. It's current for desktop processors, but not for laptop chips, which are not socketed. I would expect this chip to be surface-mounted, like everything else in that market segment.

 

[Confused-User]

I would think that the M3 will be at least 10% faster than the iPhone15 Pro. That alone will make it pretty much the fastest CPU core on the market.

...not to mention, at least 6 months earlier than the new Qualcomm chip comes out - and probably more like 8-10.

The really interesting question is, will Apple do a yearly cadence with M chips, as it does with A chips? The popular opinion seems to be that they won't, but I disagree - I think with covid-related supply chain issues and TSMC's N3 delay in the rear-view mirror, Apple will get the Macs on a yearly track, just like the iPhones. And if they do, then M4 could be only a few months behind the SnapX. Of course, we have no idea what the M4 will be like, but it's safe to say it'll be superior to the M3. (Perhaps not by all that much, if they move to N3E, but they won't regress!)

 

[caribbeanblue]

Huh, interesting. Where did you get these? I looked here and they said there was no mention of single core performance.

Qualcomm Previews Snapdragon X Elite SoC: Oryon CPU Starts in Laptops

www.anandtech.com

These weren't in the press slides, Qualcomm revealed them later at the keynote.

Qualcomm showed a slide where they claim that their new chip is 50% faster than the M2 in multi-threaded tasks. Which makes it about as fast as an M2Pro/Max (according to Geekbench 6, which is the tool they apparently use).
I understand why they used the M2 as a reference if they can't beat the "pro" Apple SoC.
Still, that's not bad if their chip actually use 30% less power than an M2 Max for that.

Being only as fast as the M2 Pro, which has 8 P-cores and 4 E-cores, while having 12 P-cores, is surprisingly underwhelming? And how is that even possible when the single core score they get at 4.3GHz seems to be similar to Apple's IPC at this point? Is GB6 that memory bandwidth dependent?

Also, it is interesting that the 2-core boost clock of the Oryon CPU is where people estimate the M3 Pro's single core boost clock will be, if A16 and A17's frequency improvements stack up.

Edit: Having read the rest of the thread it seems like my confusion is shared by everyone, so I apologize if my comment sounded repetitive.

 

[Confused-User]

Being only as fast as the M2 Pro, which has 8 P-cores and 4 E-cores, while having 12 P-cores, is surprisingly underwhelming? And how is that even possible when the single core score they get at 4.3GHz seems to be similar to Apple's IPC at this point? Is GB6 that memory bandwidth dependent?

Also, it is interesting that the 2-core boost clock of the Oryon CPU is where people estimate the M3 Pro's single core boost clock will be, if A16 and A17's frequency improvements stack up.

Edit: Having read the rest of the thread it seems like my confusion is shared by everyone, so I apologize if my comment sounded repetitive.

(Thanks for catching up, it's really annoying when people come in just repeating the same things because they can't be bothered.)

AFAIK, the 4.3GHz speed that you refer to as the "single-core boost frequency" some have guessed at for the M3 is not a boost frequency at all. It's what seems plausible for all-core performance. In fact, Mx chips don't really do "boost" speeds, as any throttling (so far as I know) so far observed in phones and Macs is done at a system level, not the SoC level - that is, an iPhone with no cooling may get too hot and slow down its A17, just as a MacBook Air may throttle its M2, but that's an issue of system cooling. Those SoCs could both plausibly run at those pre-throttled speeds indefinitely on all cores with better system cooling. That's more like "all-cores boost" speed in the x86 world.

(Edit to add: I'm sure Apple SoCs can throttle based on energy and thermals, I just think they generally don't run cores fast/hot enough to require the chip do it for thermal safety or max power draw issues, in any of their current designs. Not even the Studio/Pro. This is possibly due to fundamental design issues that limit top clocks.)

AFAIK GB6 is not hugely bandwidth-constrained. (Not that it sits entirely in cache either.) So yeah, Oryon numbers are all extremely mysterious at the moment. If I had to bet, it would be that their quoted single-core numbers are true but atypical, and probably also their multicore performance scales poorly. But that's just a guess.

 

[Xiao_Xi]

probably also their multicore performance scales poorly

It is difficult to know if a SoC scales well using a benchmark that was designed to not scale linearly with the number of cores.

The multi-core benchmark tests in Geekbench 6 have also undergone a significant overhaul. Rather than assigning separate tasks to each core, the tests now measure how cores cooperate to complete a shared task. This approach improves the relevance of the multi-core tests and is better suited to measuring heterogeneous core performance.

Geekbench 6 - Geekbench Blog

www.geekbench.com

For example:
- 24-core M2 Ultra scores around 21k in GB6
- 12-core M2 Max scores around 14k in GB6

- 24-core M2 Ultra scores around 28k in GB5
- 12-core M2 Max scores around 15k in GB5

 

[Confused-User]

It is difficult to know if a SoC scales well using a benchmark that was designed to not scale linearly with the number of cores.

Or, perhaps, you might say that that benchmark is designed well for probing a SoC's ability to handle problems that are not "embarrassingly parallel".

I think we have a baseline expectation that most chips will scale fairly well with code like GB5's - though it would certainly be interesting if the SnapX didn't. The question is, what about other types of code? And obviously there are going to lots of different kinds of "more difficult" code, and GB6 may not be a great example. But it's *an* example, and it's not crazy to wish for scaling that gets a bit closer to linear than the M2 Max gets - or than the SnapX gets.

 

[Confused-User]

Somewhat relatedly, has anyone run GB6 against *only* the P cores in a Max? That is, keeping the E cores out of it?

 

[name99]

The technology market has so much bs graphs around now it takes ages to determine what they’re trying to say anymore, and Apple is guilty too.

I call it “Marketing through obfuscation”

As for whether it will be socketed, I very much doubt it. It’s apparent that Qualcomm is targeting the laptop segment, which uses socketed cpus rarely. Usually that’s more of a desktop thing.

Example of Apple doing this???
I've tracked many many years of these things and have yet to see such an example. EVERY fscking time people claim this (most recently the iPhone 15 Pro nonsense) it always collapses into some combination of
- stupidity (not knowing difference between energy and power)
- incompetence (running benchmarks one minute after new iPhone is opened, while it is engaging in the usual background setup overhead)
- blatant lies

 

[name99]

It is difficult to know if a SoC scales well using a benchmark that was designed to not scale linearly with the number of cores.

"Scaling well" is a property of CODE not of an SoC...

An SoC can provide more (or less) memory bandwidth, which will throttle code that has extreme memory access, but realistically it's easier to see this on the GPU side. The chip QC announced looks like it will be throttled on the GPU side, and could (possibly, but not by most realistic code) suffer memory bandwidth limitations on the CPU side.

An SoC can also provide better (or worse) synchronization/communication performance and primitives. This is mostly what GB6 is testing in the "non-scaling" code and it is an important and realistic test.

An SoC PLUS THE OPERATING SYSTEM can also provide provide better (or worse) synchronization/communication performance by co-scheduling (same time, and on the same cluster) threads that are mutually communicating. Apple certainly has code in Darwin plus counters being monitored in the HW to optimize for this; no idea about either the QC chip or Linux/Windows. BUT it's unclear if GB6 usefully tests for this. To test for it you would have to do things like create many pairs of producer/consumer threads, more than the number of cores so that the hardware is oversubscribed, to see if the OS is doing an optimal job of placing together (in time and space) threads that are mutually communicating.

 

[neinjohn]

I think we are either getting all Macs with the base M3 updated on Monday (Air, MBP13, iMac and Mini) or no M3 at all. If they update only the iMac to a M3 it'll certainly hurt the sale of their most sold model as the Air. Of course, assuming that the event has anything to say about iMacs.

I think the $1,299 and $1,499 iMac 24'' are getting upgraded to the M2 with maybe a price and specs adjustment and they are releasing an iMac 24'' with the M3 Pro, price and overall specs unknown.

 

[thenewperson]

If they update only the iMac to a M3 it'll certainly hurt the sale of their most sold model as the Air

Eh, they’re completely different kinds of computer so I doubt it.

Placing the Air is weird. After this we have 2 more event times before the iPhone: March/April and June. They’ve stuck to June for the past 2 years to reveal an Air, but that kinda means if they are going for a yearly cadence for M chips (and assuming they want it out around this time) then the Air can’t be the flagship device for the M series, which is weird. i wonder what they’ll do about it.

The others could probably be updated in March/April + iPads + VP.

 

[dugbug]

"Scaling well" is a property of CODE not of an SoC...

Just to be clear the chip design can inherently allow or hinder code to scale. To say this is all code is not correct

For example look at the gpus in the m1 ultra. Theoretical code could not achieve perfect scaling across those cores

 

[neinjohn]

Eh, they’re completely different kinds of computer so I doubt it.

Hurt not in the sense they are competing on the same market.

I assume it would hurt the sale if the M3 proves to be a worthy upgrade from the M2. If the performance difference goes mainstream on media, consumers can wait longer for the M3 Air.

Whenever there is a upgrade on the base M-SoC I would expect the Air to be first Apple's device to get it.

 

[thenewperson]

Hurt not in the sense they are competing on the same market.

I assume it would hurt the sale if the M3 proves to be a worthy upgrade from the M2. If the performance difference goes mainstream on media, consumers can wait longer for the M3 Air.

Whenever there is a upgrade on the base M-SoC I would expect the Air to be first Apple's device to get it.

I really don't think it matters to them ultimately. If iMacs are affecting MBA sales it's still their computers. I'm not sure how that's supposed to happen anyway.

And I agree that the Air would be expected to be the first with a new Mx chip as the last 2 versions have seen but then they went ahead and unveiled the 15" in June, so I have a hard time seeing that happen this time.

 

[Chuckeee]

Hurt not in the sense they are competing on the same market.

I assume it would hurt the sale if the M3 proves to be a worthy upgrade from the M2. If the performance difference goes mainstream on media, consumers can wait longer for the M3 Air.

Whenever there is a upgrade on the base M-SoC I would expect the Air to be first Apple's device to get it.

I see it more as pushing people who are on the fence to get a MBP instead of the MBA. And generate the corresponding larger margin.

 

[jdb8167]

I see it more as pushing people who are on the fence to get a MBP instead of the MBA. And generate the corresponding larger margin.

Also the Osborne effect. How many people are going to buy a 16 month old M2 MacBook Air when it is clear that an M3 version is imminent.

To not update their biggest selling computer in a timely fashion has to hurt the bottom line. It’s one thing if it no one knows when the M3 line is going to be introduced but a completely different thing when you know that an M3 MacBook Air is coming in less than 6 months.

 

[senttoschool]

Also the Osborne effect. How many people are going to buy a 16 month old M2 MacBook Air when it is clear that an M3 version is imminent.

To not update their biggest selling computer in a timely fashion has to hurt the bottom line. It’s one thing if it no one knows when the M3 line is going to be introduced but a completely different thing when you know that an M3 MacBook Air is coming in less than 6 months.

There are probably 2 reasons:

1. MacBook Air 15" was just released 4 months ago. It's way too early to update it to M3. Buyers of the M2 15" Air would feel terrible. It doesn't make sense to update the 13" to M3 and not the 15" at the same time.

2. Because of the lower yields of N3B, Apple may not have enough base M3 chips to launch Air updates. This is assuming M3 is using N3B because A17 Pro is using N3B and because N3E is scheduled for volume production in "H2 2023" which is a little too late for volume M3/Pro/Max launch.

I personally think that Apple was planning to use M3 for the 15" Air launch at WWDC. Something unexpected must have happened and they couldn't pull it off. It's hard to explain why the M3, M3 Pro, and M3 Max are launching together at the same time.

 

[jdb8167]

1. MacBook Air 15" was just released 4 months ago. It's way too early to update it to M3. Buyers of the M2 15" Air would feel terrible. It doesn't make sense to update the 13" to M3 and not the 15" at the same time.

If Apple does release an M3 SoC on Monday then every purchaser of a 15” MacBook Air is already going to know that they have an older model. What Apple could do to appease this audience is release the 13” and 15” M3 MacBook Airs while keeping the M2 versions around at a slightly reduced price and introduce the M3 versions at a slightly higher price.

2. Because of the lower yields of N3B, Apple may not have enough base M3 chips to launch Air updates. This is assuming M3 is using N3B because A17 Pro is using N3B and because N3E is scheduled for volume production in "H2 2023" which is a little too late for volume M3/Pro/Max launch.

This is a real possibility. The Air is their biggest seller (for Macs). But it would have to be a very severe shortage considering the number of SoCs that they need for the iPhone pro. The number of M3 SoCs needed would be a rounding error even for the MacBook Air in comparison.

I personally think that Apple was planning to use M3 for the 15" Air launch at WWDC. Something unexpected must have happened and they couldn't pull it off. It's hard to explain why the M3, M3 Pro, and M3 Max are launching together at the same time.

It’s unlikely that Apple could change a new product that quickly. I think that the 15” was always going to be an M2 but its release got delayed. It would have made more sense to release it with the updated MacBook Pros in January.

I’m still in let’s wait and see mode for Monday’s announcements. Something seems off to me too.

 

[thenewperson]

I personally think that Apple was planning to use M3 for the 15" Air launch at WWDC. Something unexpected must have happened and they couldn't pull it off.

Would be weird if that were the case. Why wouldn't they wait till they can get M3 instead of going with M2 and throwing the Airs off-balance wrt update schedule?

 

[Chancha]

Exactly, if an M3 15" were to appear now, it wouldn't hurt that much to not release it with M2 just 4 months ago? Dropping both 13" and 15" with M3 now would then make a lot more sense.

Gurman's latest report also claims the 13" Touch Bar despite not getting any major design change is also not dropping now. This model unlike the iMac 24" has no room of which chip ends up being in there, it has to be an M3. This to me means the M3 may not even appear on the iMac, and the whole thing is pushed to being N3E where N3B is exclusively for M3 Pro and up. The iPad Pro, iPad Air these all are waiting for an M3 as well, they too are pushed to next spring.

But then it will be quite odd to throw iMac 24" out with just an M2 but during the same event we get M3 Pro and M3 Max for the MBP though.

 

[nquinn]

Looking forward to Apple using benchmarks against 3 year old cpus again.