A17 Pro is HIDING a dirty secret?

[@peraman2022]

I believe that Apple is going to stretch out the benefits of the 3nm process over the next few years. The CPU performance doesn't really matter a whole lot, but having more specialized features will further impact the performance of a particular system.

[@lukemiani]

I think that's exactly what they're doing. Roll out a new GPU and advertise 3nm now, then a new CPU with big gains next year, then a minor refresh to both, etc etc

[@mikelay5360]

​@@lukemianiwho is apple in competition with? Qualcomm isn't even trying.

[@chidorirasenganz]

yeah most of the transistor density gains were in the neural engine, hardware raytracing, extra gpu core and the AV1 encoder. All of that without blowing the power budget

[@United_Wings]

Duh 🙄

[@DowneyMax]

​@@mikelay5360Mediatek had a great comeback though

[@johnarcarese172]

Typically when moving to a new process, you don’t usually see a new cpu design. The next cpu, A18, will show major improvements.

[@Errcyco]

I hope so, having a17 performance with say double the efficiency would be enough for most people for the next couple years at-least. Would also like to see them focus on battery life, like almost exclusively-that’s what we all need. They are plenty fast.

[@bulletz4life]

this is literally what everyone said the a17 would do......

[@lvrt8371]

Maybe a19

[@utubekullanicisi]

The TSMC claims for speed and power efficiency improvements of N3 are compared to the original N5, not N5P or N4. Apple has already increased frequency on the performance cores by 15.3% since N5, this 9% increase is on top of the 15.3% that we've already got. I wasn't really expecting any improvement on N3 knowing this, or if we got microarchitectural improvements I thought the prices would surely go up. I'm surprised to find out that the iPhone 15 Pro price stayed the same. This is probably partly due to the rumored 'sweetheart' deal between TSMC and Apple for 3nm. Here's the secret about TSMC 3nm: A lot of people have been hyping up 3nm as this supercycle when it comes to process node advancements but it's actually the smallest node improvement that's referred to as a 'full-node' and not half node improvement, in the history of semiconductors. Don't get me wrong, the shrink factor and power efficiency/frequency improvements are not that bad compared to earlier, but what's not good about it is the transistors/$ performance. 3nm is 30% denser than 5nm, but for chip designers like Apple to be able to utilize those savings on die area and have the luxury to keep the die size the same as (if not make it bigger than) the previous gen, despite the shrinkage of the transistors (in other words, for them to add more features to the die instead of just shrinking the same die), the new node needs to have better price/performance. But while the chip can be shrunk by 30%, it's also 25% more expensive for the same die area. So if they don't want to increase the cost of the chip, they have to add few more if no more transistors and thus they have very little wiggle room to add more features to the chip, instead of just shrinking it thanks to 3nm. Keeping in mind all of this, I was surprised that they even were able to add 18.75% more transistors without increasing the MSRP of the new phone, because that's not the wiggle room they had. And this is why the improvement is smaller than in previous supercycle years despite 3nm being a "full-node" improvement in terms of die shrink and power efficiency. The significantly better transistors/$ performance despite the technological superiority of previous supercycle nodes allowed chip designers to amplify the benefits of these supercycle nodes, but now we're in a situation where we can shrink the chip fine, but we can't add more transistors to the chip because the cost is prohibitive. So in the future we might watch these chips get ever smaller and get frequency and power efficiency improvements that are enabled just by the new node, instead of the chips being roughly the same size as last year because of added features and beefier cores that amplify the benefits. Or we might have to accept increasing prices.

[@MaximoJoshua]

Thank you for writing this, great details! I thought it was just because it was comparing N5 to N3 in the TSMC paper (15% gain) but comparing N3 to N4P in the performance benchmarks (10% gain)

[@dustinj7028]

Well said.

[@fixups6536]

Last night I commented on your previous video about this rather disappointing improvement in CPU speed, and pointed out that it was bad news for us, who are waiting for the M3. Now thinking about it, I believe the answer is: ...Vision Pro! At this point it is obvious that the A17 or a close derivative will be the heart of the Vision Pro. For this device, what Apple needs most is a significant improvement in GPU performance (for obvious reasons), and probably also in the neural engine (for the gestures controls). It looks like this work has stretched their Apple Silicon engineering team thin, and they realized they could not work on everything at once, so they made a conscious decision of leaving the CPU section in its current state this year.

[@thebuddercweeper]

They also need a bump in GPU performance on the M series to be properly competitive with AMD and NVIDIA's high end, and for their current focus on Mac, iPhone and iPad gaming.

[@fixups6536]

@@thebuddercweeper Yes, that's right. They are basically competing with every major player in a number of IT fields, taking on each of them on their strongest point, which is crazy when you think about it. And this competition is very good, as we get better Macs, better PCs, better iPhones, better Android phones, better GPUs and more as a result.

[@5urg3x]

@@thebuddercweeper Apple’s GPU performance in the M series lineup is embarrassing compared to NVidia or AMD and even Intel’s higher end dGPUs. It wouldn’t be an issue if they allowed customers to choose and use dGPUs still in combination with their own, like they used to on the Intel macs…Apple is the only tech company in the world that forces pro customers to use integrated GPUs. They give you no choice and you can’t even get an eGPU anymore either.

[@PanosPitsi]

@@5urg3x they don't have a choice no gpu supports the arm architecture. Arm windows laptops don't support external GPUs either.

[@5urg3x]

@@PanosPitsi You have no idea what you’re talking about. Ampere Altra supports dGPUs, plenty of people use them for insane AI cloud workloads. Even Intel has ARM Linux drivers for their Arc dGPUs! They don’t support Windows because of Microsoft’s whole licensing thing with Qualcomm… it has nothing to do with any architectural limitation.

[@AlfredoGutierrezNavarro]

I think the key here is the neural chip, and it is so because one main reason: AI. Apple is a software company, as you pointed in your example with ray tracing. I do think that they will be moving many iOS tasks that now are being done by the CPU to the GPU and the neural chip.

[@miguelash886]

That makes more sense as performance gains would be much higher than any performance gain on the CPU

[@ghfsd786fa]

Yeah, more and more problems are being solved with ML, CPU is kinda pointless from now on...

[@MarkEmeryPhotography]

Agreed, the CPU becomes more of a task scheduler and data marshaller, letting dedicated hardware do the heavy lifting. Given it won’t be doing things it had been relied on before, the 10% will appear much higher.

[@ekinteko]

I think I may have figured it out. TSMC has a big standing order with Apple. The current TSMC-3nm has low yields for what was promised +30% improvement. So instead of throwing them in the bin, or making TSMC pay Apple money, they are doing the next best thing. Firstly, using the new lithography to make the chips much larger than before. This is focused primarily on adding new features around ML, AI, DL, etc etc. There's some improvement to the other factors like GPU, CPU, and Efficiency but they aren't the main focus. Next year it will be around another property, like the CPU performance, with minor improvements to everything else. And the following year it will be a different property, like the GPU performance, again with minor improvements to everything else. As for M3, let's hope it is different. That the CPU built for the M3 is the next-gen core, and this time they will build it first for the laptop and then introduce it/scale it down to the phone. The last two times it was sort of the opposite, where A13-A14 development happened alongside and adapted to the M1. Then the A15 development occurred, again adapted to the M2 as a minor refresh. Now perhaps the A17 is built off the A16, whilst M3 releases first which gets adapted for the A18. It would be improbable, but not unusual. We will have to wait and compare the A17, M3, and A18 to be able to understand the strategy they were thinking for this migration/jump to the new Node platform.

[@AzizRahman1]

To be fair, though, the CPUs in these chips are already pretty fast anyway. Better now to focus on things like the GPU and the Neural Engine to round out the package.

[@maxffarrell]

Important to remember M2 based on A15, so M3 gen could have ~30% gains over M2. AV1 and ray tracing are pretty sorely needed for futureproofing so it’ll be a well rounded package

[@chidorirasenganz]

Plus 2x on the neural engine

[@thebuddercweeper]

applying the percentage performance increase in A16 over A15 to M2 and then the leaked A17 Pro increase to the result, I got 3339 single core and 12935 multicore, which is 32% faster single core and 37% faster multicore.

[@chidorirasenganz]

@@thebuddercweeper Sounds about right 😮‍💨

[@G43X]

Hardware AV1 decode will help me out greatly.

[@madriditunes7021]

​​​@@thebuddercweeperIs more simple to think that A17 Pro can make 35 trillion operations per second, the new A16 Bionic only 17 trillion operations per second, the A17 Pro 2x more fast than the new A16 Bionic, the A17 Pro define can run games of new generation of consoles like PlayStation 4 or 5

[@jorgeomarcosmerivera4871]

I don’t care, I’m moving from an iPhone 8 Plus to the 15 pro max. 😂

[@dm01417]

Same i moved from iPhone 7 Plus

[@eman20911]

I moved from iPhone 7 Plus too 😂

[@meislit9217]

Somewhat disappointing gains indeed, but on the other hand it makes me even happier and more confident that I got an M1Max as soon as they were released and didn't wait. Seeing the changes that have been made to M2 and what's to expect for M3 it definitely seems like it was a good choice to go for it then and not wait.

[@reda497]

Even TSMC mentioned that 17 pro chip has a problem. And actually the A18 pro will be the real 3 nano chip.

[@MacGuyliver]

I rather think that the CPU has become smaller compared to its predecessor due to the new architecture. This creates space for the other components in the chip. This explains the small increase in performance. Opening the chip could solve the mystery.

[@thebuddercweeper]

That's certainly possible, considering the addition hardware of ray tracing, and the rumoured performance core increases in the M3 series.

[@Wokiis]

The CPU-portion is probably the hardest part to improve at this point. You can only make them so wide before running into diminishing returns. But an improved neural engine and revamped graphics sound pretty much like what they **should** focus on anyway. Really excited to see the graphics performance of the M3-line. I don't expect the M3 to be **completely** based on the A17, though. Namely the process node, TSMC's troubled N3B, seem to be a bad choice for.. well.. anything. Might be a one-and-done for the A17 Pro and then everything else will come once N3E (which seems to be a bit worse from a performance perspective) is ready. Else, M3 could end up costing too much for the kind of devices it is supposed to go into. Also N3B likely why the A17 got a "Pro"-moniker. They don't want to commit having this hard-to-produce chip in anything come next year.

[@joelv4495]

Yup. Many people have no idea how complex writing software that fully takes advantage of multi-core hardware. So even "making it wider" has limited benefits if the software isn't there.

[@HairyHands]

I'm actually more excited for the standard iPhone 15 basically matching features of the 12 Pro Max and being USB-C huzzah

[@intranix]

If anyone here remembers the Amiga, it had a mind-range cpu but coupled with a bunch of custom chips to do cool stuff in hardware. That's what we need now, improvements in graphics, machine learning and related stuff. as well as video en/decoding. This is what it'll be about going forward. small improvements overall, while adding extra hardware for custom stuff.

[@StrtSlct]

4:35 It's worth noting... the A16 had 5 GPU cores, the A17 Pro has 6 CPU cores. That alone is the 20% improvement...

[@The.JZA.]

They are pushing for gaming performance. What do we need when gaming? Consistency, a phone that doesn’t overheat and slow down after 5 mins. Their CPU is already the fastest in any phone and even rivals high end desktop CPUs in single core performance, so they spent the extra 3 billion transistors on an extra GPU core, ray tracing, way faster AI/ML cores, hardware AV1 decoder, display controller, USB controller etc. If they want to play console games on the iPhone, they need to be able to crank out heavy graphics for long periods of time without overheating. THAT is why they didn’t push for crazy performance. Seems pretty obvious to me.

[@teg24601]

Using the same core on the new process makes a lot of sense. They know how it works, it is tried and true, and they need to make sure it works correctly, especially with the new process. The A18 will likely be the redesign you are looking for, with more performance gains than you would be expecting. This cycle, they were focussed on the GPU, as that seems to be what everyone keeps griping about when it comes to the M-series. So get that up to speed, with RT, and better performance this cycle.

[@imelliam]

I don’t see why the cpu even needs to be faster. I want GPU power.

Rumors for years were that apple cannot figure raytracing in ios chip. So probably they were fully focused on that and left CPU as it is. Especially because it will be very useful in VR headset

[@Ssin1235]

On top of the mediocore improvement on the CPU side, A17 pro GPU without ray tracing is just the same core as A16 and A15. They went from 5 cores to 6 cores so basically they gain that 20% speed increase by adding another same core, no IPC boost this time.

[@PureGlide]

Gotta remember, they only have limited engineering resources (people), so maybe the chip designers are working on other projects (e.g. modem chips). Also it's not stupid to just focus on one major change at a time. So maybe the change to the 3mm process is change enough. I know they work many years out in advance, but they still have to ensure things are completed each year. Changing process AND redesigning the chips at the same time just sounds way more risky.

[@jasonmajere2165]

New design means new software to go with it.

[@pingpong1727]

I think it has to do with the fact that iPhone sales are dwindling, they need to ensure that they have more to update next year. I’m sure they will do much better with the macs.

[@SunsetNova]

iPhones are still selling well above competitor's

[@pingpong1727]

@@SunsetNova Regardless, it is not good for their shareholders or the company itself to have shrinking iPhone sales.

[@thorium9190]

The A18 Pro is out! Can’t wait for your analysis

[@sharonb.9128]

Great video. Apple has shown us time and again they spend years of R&D plotting the course for their devices. They told us at the developer’s conference that has been creating Spatial Computing. ANY review of Apple’s strategy that does not at least ask the question of how it relates to, improves or creates new 3D, AR/VR and AI capabilities is blind. Now Spatial Audio makes sense

[@sharonb.9128]

@@wesayhowdyhere9589thank you.

[@j340_official]

1:11, that quote is for 3nm vs 5nm, but a16 is 4nm. So the performance increase from the jump from 4 nm to 3 nm is even less.

[@christopher3d475]

Those of us who do raytracing and 3D work see this as a big deal. Apple hasn't been able to compete with Nvidia/Optix until now and I'm anxious to see how this plays out on the M3 chips. Even if it's not as fast as a 4090 for raytracing operations, still if it's 4x as fast as GPU driven raytracing, it'll be huge.

[@lukemiani]

The GPU is absolutely a big deal

[@Egor9090]

This doesn't explain where did much expected power efficiency go

[@5urg3x]

Android phones have had USB3 controllers for the better part of a decade now

[@carlosnumbertwo]

Still prefer iPhone. And I had the T-Mobile g1 :D

[@Mohamedahmed-jz7xd]

Saving resources, adding gimmicky feature and innovating nothing = Typical Apple event.

[@penguinlust6749]

When you transition to a new process you typically take a conservative approach to the chip design. For a brand new chip-making process, this all makes sense. As is, there's a lot of additions, and this is just the beginning of exploiting this process. Give it time.

[@borislavrashev]

I guess they didn't switch to arm v9 again. Reason is simple - android chips are still behind raw power. Next year the node will still be 3nm, but they can boost from the new instruction set like 30 % easily. No need to boost much this year.

[@DiegoJose-gx8bs]

I find this chip somewhat underwhelming. The issue lies in the fact that none of the improvements will truly impact my daily experience. I don't use my phone for gaming; I have a PS5 and a PC for that purpose. The 10GB/s transfer speed isn't necessary for my needs; Airdrop is already fast enough. While the neural engine is impressive, I rarely find myself needing it. That's why improvements in CPU performance, especially in single-core scores, are crucial. Enhancing CPU gains will benefit everyone in their daily tasks, from social media and web browsing to video streaming and app downloads. That's just my perspective.

[@SweetSweetCandyBoyz]

to be fair, they are still miles ahead of anything Qualcomm and MediaTek have in their flagship mobile processors.

[@ChiquitaSpeaks]

Kind of surprised nobody has mentioned this seeing as it makes so much sense as on Apple's website they say "Using an industry-first thermo-mechanical process, the titanium bands encase a new substructure made from 100 percent recycled aluminum, bonding these two metals with incredible strength through solid-state diffusion. The aluminum frame helps with thermal dissipation and allows the back glass to be easily replaced." Which means a large part of the benefit might be that they are getting better cooling on the phone and therefore just able to clock it to sustain more/higher. The ideal would be that they have genuine architectural advantages combined also from the 3 nm process yielding the seen increase even though they put more into efficiency and are just able to sustain higher clocks longer in certain scenarios so they probably just gave it the same battery rating because it will drain a lot more when running its real performance that much harder. This would actually make a lot of sense as to why they called it 'Pro' as if it was the first chip more akin to use similar to that in desktop scenarios and why they advertise gaming so much (makes sense because a 20% GPU increase is plenty within what we've seen before) since those require that sustained performance…in addition to all of the specialized hardware added

[@Knevitzdoug]

remember this A17 with GPU and ray tracing was scheduled to be the A16, and that's what Apple delivered. At the time, TSMC couldn't start the 3nm process either, so the A16 was a placeholder. This chip isn't even ARM V9 yet, the big leap in performance will probably be next year, and even then, nothing exceeding 30%, adding CPU and GPU. The regular A17 of the iPhone 16 will be the same A17 Pro in colors, but using the conventional 3nm lithography process, not the one Apple developed for the A17 Pro and not the N3E of the A18 Pro.

[@anelhovermale2317]

I am positive it has to do with power consumption.

[@tonyburzio4107]

Yeah, this has been covered on other channels.

[@mylittlepimo736]

But there was no battery increase

[@6XCcustom]

there is no difference in battery life between iPhone 14 and iPhone 15, although according to TSMC there should be at least a 30% difference, so Apple has not done a good job Steve Jobs had fired Tim Cook because he doesn't understand what Apple's core value is, or should I write where Tim Cook's core value is to make as much money as possible and quality, innovation is not something Apple should care about anymore, therefore Steve Jobs had given Tim Cook a kick in the ass with every right

[@mylittlepimo736]

@@6XCcustom TSMC said 15% performance increase OR 30% power difference with no other changes. Apple increased the CPU performance by 10% and GPU performance by even more, so that's why there's not a noticeable reduction in power consumption. I'm more disappointed that there was no architectural improvements on the CPU.

[@tipoomaster]

A16 was supposed to be for N3B, but they had to revert it to N4 and revert the GPU to the older one. This seems like the A16 we were supposed to get. I hope M3 on N3E debuts an actual new CPU architecure .

[@N3xlow]

To be clear 10% in a top tier processor is a lot in therms of temperature

[@Alt-Key_Here]

I remember hearing that the yields of the 3 nm chips were not great, so I wonder if the 4nm CPU was needed to meet the launch day inventory. I wonder if the base 14s had the previous chip due to 4nm yield issues, and for the 15, it was the same issue. Probably greed though.

[@盾神京矢]

1:17 The 15% improvement is from 5 nm to to 3 nm. But the A16 to A17 is a jump from 4 nm to 3 nm. So we expect less than 15%. 10% sounds just right.

[@theotherphil]

Here we are talking about an underwhelming CPU performance gain when we now have an iPhone with equivalent performance of a desktop M1 Mac. M1: 2,384 / 8,642 A17P: 2,908/ 7,238 In the iPhone, there is very real energy and thermal considerations and Apple has clearly used some of their thermal/ battery overhead on GPU performance, ML performance and RAM. Extra RAM needs extra power. Just last year they had to push back their GPU upgrade due to thermal/ power considerations in this form factor, and it's now possible thanks to a die shrink. I think they've done an amazing job to give this level of performance with the same battery life given it's form factor, to be rivalling their M1 desktop machines which just 18m ago was revolutionising the PC industry. I don't think the desktop versions of these SoC's will be limited in this way, especially when M2 is based on A15, so M3 gen could have ~30% gains over M2.

[@Seanconnerlaw]

Who needs that much faster of a CPU in the iPhone? Genuine question? Because from everything I know, software still hasn’t caught up with the performance potential of the A16, much less the A17 Pro. With the improved graphics performance isn’t that still a massive improvement?

[@regalblack1503]

Others are so behind apple in mobile CPU performance that they have become lazy. Snapdragon 8 Gen 3 is still behind A17 Pro In Cpu performance

[@Jjarret]

A bit of an Iphone User here. We already knew this. Apple doesn't redo their architecture until next year. They use a 5 year cycle. Name's and numbers on processors mean literally nothing and has always been so.

[@MaximoJoshua]

I just think I would never switch to a new process and make a bunch of other changes at the same time, typically you want to change one thing at a time to minimize errors

[@Dygear]

It’s all about the GPU. Meaning the M3 chips should be good at gaming. An area that Apple seems to be really focused on recently.

[@th0m]

2:17 yeah I guess speed doesn't matter all that much if you're upgrading every year or every other year. I don't think most of us are doing that.

[@Adgale]

They don't have the need to increase the performance that much because they are a head of the competition 2 o 3 years.

[@GalaxyFur]

Actually, they are behind. That's why Apple had to get hardware-based Raytracing out this year. Because the competition from Android phones like the Galaxy phones already had hardware-based Raytracing. And the A17 Pro still doesn't support Variable Rate Shading like Galaxy phones or other Snapdragon phones. This is why Android phones often outperform iPhones in real-world performance, depending on the applications. Unfortunately, synthetic benchmarks are notorious for being unreliable, and people don't know this, causing people to often make the mistake of believing this tall tale of Apple being a few years ahead of the competition. And Apple is manipulating benchmarks sometimes as well. Plus, when they often quote things like a 30% performance increase or a 20% increase, they are often found not to be referencing last year's model but rather from two years ago or more. For example, they were caught being very disingenuous with their laptops and skewing numbers against Intel-based chips. Yet, the Intel chips performed better when the tests were done correctly. To be fair, Intel also did this practice against AMD in some marketing material as well. Chinese phone manufacturers are also notorious for fudging benchmarks. In fact, they're the worst offenders by far. But this shows how much benchmarks are often a poor indication of performance. Not in all situations. But in many. The reality is that the latest Snapdragon and A-series chips are neck and neck.

[@Adgale]

@@GalaxyFur What Apple do better is offer the same or more performance consuming less power. That's a win for mi. And also i think you're wrong when you say that in terms of graphic performance Snapdragon + Android are ahead. We all know that Iphones always run games smother that the competition. But yeah is true that this gap is less pronounce now, but it remains the same. And in regards A series CPUs they perform better and and even the ones that came out two years ago or even three still compite with newrs Androids CPUs. Another thing that Apple do pretty well is to do things better than competition when they release a product or a technology and they don`t care if it is released after them because they care about user experience more than anything and also they pretend to earn money so they are not going to put something in the market that they think is not going to make profits.

[@GalaxyFur]

@FallingESP iPhone doesn't run better. They have less powerful GPU's. I can run programs often faster on an Android compared to an iPhone. The iPhone also has thermal throttling issues with the A16 series, for example. Apple always slows down iPhones after two years. This is why they have been sued multiple times. The iPhone has many limitations since it runs IOS and is another reason why the EU is stepping on Apples throat now. I can't run an iPhone, for example, personally because It is incapable of doing what I need my phone to do. I have family that loves their iPhone's. But they aren't power user's either. iPhone is efficient since everything is made in-house. But also because IOS is a smaller operating system and coupled with the fact that the iPhone runs a slightly lower screen resolution compared to most top-level flagship Android phones. Personally, I'm not a fan of iPhones because of the way IOS is. It's too restrictive for my uses. But it's still a great phone for many people. So I do always recommend iPhones to people who I think may like one. 👍

[@francisdelacruz6439]

The explanation could be as simple as losing their cpu designer architect to a start up and then floundering to keep the design going. Risk V would be an interesting choice for cutting edge efficiency per watt designs that could catch up or even pass ARM designs. It’s the people that designs chips.

[@polypolyman]

...so they essentially took some of their power budget from the CPU cores and fed the more powerful GPU and accelerators - neat.

[@frankinla91602]

When they originally developed Ray Tracing (A15 or A16), they had to leave it out of shipping chips because of energy consumption and thermal issues. Perhaps these A Series chips are slightly throttled so as to get RTing in this time. Perhaps the M Series chips, with higher thermal and power factors, will not be so throttled.

[@DowneyMax]

I'd buy the Air if they had hardware RT capability. I'm not sure what's going on with Asahi linux. Either way, I'd install it with the laptop

[@piotrd.4850]

@@DowneyMax Surely Apple is going cutrail it on non-Pro lineup. Anyway, with their upgrade cost, Air line basically exist only to upsell Pro at this point.

[@magicmanchloe]

3:45 that is just not how that works. A CPU Architectural needs to be designed for the note that it’s running on. Yes, there are some basic functions that can be somewhat node agnostic. But at the end of the day when you’re getting ready to tape out the chips you were going to have to make changes and design for the new node. This is why AMD is chiplet architecture allows them to innovate much faster. Because they can shrink the compute dies and leave the exact same IO die, this allows them to not have to redesign the extremely complicated IO die every time they release a new CPU.

[@Jo21]

yeah they keep cpu at the same size and use teh space at gpu so the 10% gain is just from 3nm boost, to be fair it will prob save battery life

[@NinjaKiller1022]

Yea, I’m not upset with the smaller than usual upgrades, but I’m worried what this might mean for other chips moving forward. I was at least expecting more efficiency.

[@cynicalguy]

I think... they moved to a new process that would give them 15% more performance at the same power... but instead of taking the 15%, they lowered power use to grab 10% more performance, and re-allocated the power that would have used to GPU and AI. Makes sense, the previous gen phone was already a performant beast, so the balance between performance and battery life wound up being more power for new features, less required for the CPU, and the amount of time you can use it staying relatively equal.

[@boomers_pb]

Why boost CPU performance all that much when neither Samsung nor Qualcomm can even catch up to iPhone 12?

[@brianstory]

They appear to not be looking to do much this gen of iphone except keep selling them. Next year we might see a much better upgrade. Not much reason to update the CPU on these phones for processing when they are already plenty fast.

[@youknowwithMartyKauffman]

Yeah, I’m gonna keep my iPhone 13 mini

[@walterwhite1]

Smart move

[@youknowwithMartyKauffman]

@@walterwhite1 yeah, I wanted to buy another one, but Apple isn’t selling them anymore

[@MrGwizyadig]

I absolutely love that phone, but the battery life was terrible. I miss it though

[@youknowwithMartyKauffman]

@@MrGwizyadighey iPhone mini for life.

[@PCJesus0]

Even if the cpu is essentially the same as the a16, the efficiency and new gpu are still huge upgrades for iPhone and (possibly in the future) iPad. Batttery will most likely be fantastic on the 15 pro and pro max because now they’re using less power for a minor bump in power

[@thebuddercweeper]

I do wonder why they didn't advertise a battery life increase though.

[@PCJesus0]

@@thebuddercweeper beats me

[@yordi3945]

on the website it says the same amount as iPhone 14 pro

[@PCJesus0]

@@yordi3945 thats only in 3 scenarios, the actual battery life will probably be an hour or two better than the 14 pro and Pro Max

[@locust76]

4:12 … except that the process has been shrinking and 3nm is absurdly advanced… how dare they shrink the process to insane levels and *only* increase performance by a _measly_ 10% year over year?!

[@blakespot]

Are you under the impression that Apple should entirely rehash the architecture every year to somehow get 30% CPU gains yearly? Each iteration of the A chips is a tweak of the previous, basically. M1, a larger A14, made huge news because it landed in Macs, leaping beyond their previous Intel performance at a small fraction of the power. Some were perplexed that the M2 wasn’t an equal jump over M1 as M1 was over Intel. No one that has a solid notion of how these things work expected that. 10% CPU gains here don’t seem shocking, but expected, to my eye.

[@shueibdahir]

Let's be honest. Cpu is not the achilles heel of the aplle silicon. It's the GPU. It so damn underpowered it bottlenecks everything. I remember getting my macbook air m1 and being so disappointed by the gpu. Such speed when it came to anything outside of graphically intensive workloads but the moment i needed to use plugins that harness the gpu it made everything worse

[@francoisbouchet9907]

Luke Miani: "Why would Apple do that?" Apple: Hold my Maximizing ROI Beer.

[@MrXxshayaxx502]

Does this mean that the iPhone 15 Pro will have heat problems and battery loss? The battery life is the same as the iPhone 14 Pro, and could it be worse since it is the same chip and they increased its performance to 10 and 20 percent?

[@miks564]

The improvement over the A16 might be unexpectedly small for the main cores and it is indeed a surprise. But we can't forget that the A16 to which we're comparing it to, is still a top performer in the business. So adding a bit more performance, plus RT with a better and faster GPU is not that bad I would say.

[@derfatalerror1541]

I just watch the 15 Pro review from Geekerwan and he showed that the CPU is definitely not the exact same and that there are actually some small IPC and performance improvements depending on the workload. Maybe a more in depth video would be nice now that we know more about it.

[@Longuro]

You have no idea how the M3 processors cpus will be based on the a17. It’s just baseless speculation right now.

[@Barney_Greenway]

I just hope they underpromised and will overdeliver on the battery... 🔋

[@Zantsak]

I wonder why nobody is talking about Apple going after consoles and the gamer PC market?

[@z1tterbewegung]

There was a rumor that the a16 was supposed to have a Raytracing module in the iPhone 14 pro and pro max but Apple couldn’t finish it in time so they had to do a redesign for the a16 to have something in that year for the iPhone . If that is true it makes sense that this chip focuses on the GPU to have that single feature on the chip in general.

[@FouDeFafa]

About this 20% improvement on the GPU side, you should mention the fact that the A17 Pro has 6 cores vs. 5 for the A16. This would imply the same performance per core YoY 🤔 Also worth mentioning that those GPU cores are now custom designs instead of simply licensing the ones from PowerVR. A plausible interpretation of all this is that this first custom design performs very similarly to last year's PowerVR offering and they're trying to hide it by adding an extra core to the SoC

[@i103772]

I really hope that's not the case

[@gehadsheha4775]

They have been using custom designed cores since the A11 Bionic, so your point doesn't make sense.

[@FouDeFafa]

Well well well :( Results are now in thanks to Geekerwan's in-depth analysis and it seems like this 3nm process node jump isn't really living up to the expectations with yet another YoY peak + average power increase despite a not-so impressive performance jump. Not a bad chip by any stretch of the imagination of course. Just a little disappointed, especially in the GPU department 😢

[@Localtravel3r]

My man wants the Soc to blow through the case.

[@AyaWetts]

why would they do it all at once...? They hold back so they call sell new ones every year.

[@MilushevGeorgi]

My boss told me my paper was underwhelming, hearts when you say it

[@ConsistentlyAwkward]

i was hoping for better battery life that’s all i wanted

[@taidee]

My suspicion that led to these mediocre improvements is extremely low yields at 3nm. Even with them just deciding to dedicate the SoC to the iPhone Pro lineup, I still think the numbers still couldn’t get there and what we are seeing might be binned chip.

[@Anders01]

I was actually impressed that Apple managed to release a 3nm chip. Samsung has been struggling with their 3nm process and even TSMC seems to have had problems. So my guess is that it's the traditional silicon microchip technology itself that is flattening out in terms of price/performance. Maybe forcing Apple and other companies to artificially lower the performance initially so that it can be increased in future chips even with a technology that is running out of steam.

[@milkedition]

I was also wondering if the unreliability of the node is forcing them to clock conservatvely.

[@psycl0ptic]

There is so much more than just a CPU in the performance of a whole system, esp when it's a SOC: GPU; storage/IO; Memory/IO; NE; AV1; network; etc... And lets not forget the phone CPU alone is still scoring higher than some intel desktop CPUs. and a GPU perf boost will deff be welcome gain for the M3 SOCs....

[@arturoehr]

Yearly exponential gains aren’t realistically possible nowadays

[@greghudson3434]

I think it is possible that Apple has plenty of performance in iPhone and rather than needlessly increasing CPU power Apple is looking to run the chip cooler. They did mention that they were looking for more SUSTAINED performance for games. So it would make sense if you can get 10% more performance and run much cooler and longer and no throttling concerns to just do that. When this comes to the desktop Macs where Apple has much beefier cooling they can run at higher clock speeds and run hotter and draw a bit more power. This makes sense as we are seeing Apple differentiate chip performance more across its range as they will have more CPU cores (rumored) on the Max lvl M3. So while an M1 was and M1 and scaled linearly from the A series it was derived from Apple might be more focused on power consumption/heat more in iPhone/iPad/Macbook Air this time and push the envelope more on performance/power/heat in desktop Macs this time around. After all eventually a Mac Studio might have need of that big cooler unit maybe that takes up half the enclosure? Just a thought

[@sbstndltn]

They are using early TSMC 3nm. It is not as efficient as the later revisions that aren’t yet ready for mass production. N3X is significantly better and is not ready. That’s why it’s not showing up well in benchmarks.

[@Pipsispite]

Whilst I think it's fair to say they want people to upgrade. TSMC largely oversold their 3nm. MediaTek announced the worlds first 3nm chip the friday before the Apple event. So much for exclusivity for 3nm btw. Anyway Chinese phonemakers are complaining they're having issues keeping it cooled. Which is strange considering the efficiency improvement. Yields are also only at ~55% and TSMC is requesting suppliers to boost supply or they'll have delays in production. It ain't great.

[@OverclockerX]

I hope performance of M3 chip won’t increase also by just 10%.

[@JonSchwark]

I think we are headed for the day when you come home at night with your phone, and you can plug it into a dock and use it with a monitor, keyboard, and trackpad. There won't me enough of a difference to justify getting a mac mini for most people.

[@TheCentristChad]

Interesting. Maybe Apple is focusing on the GPU because they plan to add it on to the Vision Pro to focus more on gaming, their main weak point

[@waltonks8172]

A16 uses N4 and the A17 uses N3. That is just a nanometer smaller fab. The TSMC's claim was referred to N5 to N3. It's makes totally sense. I’m sure we will get a slightly bigger jump with the M3 series.

[@johnatkinson1111]

I’m thinking it has to do with where developers are most likely to be hitting roadblocks. I don’t think many developers are getting bottlenecked by the cpu but games and heavy workloads are more likely to be getting bottlenecked by the GPU and ML cores. Apple likely took most of their chip development team and told them to work on the GPU and ML cores and maybe had a few on the CPU to make sure it’d work with the process node shrink.

[@thebuddercweeper]

That's a really good point, especially when it comes to phones - the CPU hasn't been the bottleneck for anything on an iPhone for years. Even the CPU in A14 is very powerful and no one will notice a difference between that and A17 Pro, but they might notice a difference in the GPU.

[@agentofenhanced2428]

Who cares about gaming tho? Theres barely any games to play and its a waste of time when you can litetally get a xbox series s for 250 dollars. In day to day perfomance ram and cpu are more important, i got no idea what apple is doing but it looks stupid

[@johnatkinson1111]

@@agentofenhanced2428 even for non gaming workloads the GPU is important such as video encoding/decoding and rendering, or real time simulations of data models. ML cores are important for more machine learning tasks coming to iPhone and having the ability to improve the already existing processes, Siri and dictation went on device a few years ago but for it to improve it needs more specialized cores. ML cores also power Face ID so the iPhone 15 pro may have faster Face ID.

[@agentofenhanced2428]

@johnatkinson1111 so basically you mentioned face id and thats it, ive been an iphone user for 3 years i pretty much run my entire buisness on my phone whilst having a m1 pro that i barely use. So totally useless argument nobody can convince me people even use phones for gaming

[@TheAuthoritarian_Band]

Current OS is a tweak of NEXT, they tweak, thats what they do. Plus, holding things back - pre tweakers From a business perspective, its incredibly clever.

[@Alex-vp4ky]

This matches the late rumors that said apple wasnt building the architecture on 3nm, just building the same 5nm structure but on the smaller die

[@TimurTripp2]

As you yourself said they’ve been plenty fast for years. Clearly the R&D is going into things like the GPU and camera instead of the CPU, things Apple deem as bigger priorities for improvements.

[@aaronthedj12]

Should have been an M3 chip , instead of a revamped A16. What an opportunity wasted....

[@siaseeds8950]

They are like "we gave you USB C, so just be happy with that"

[@StrangerNoises]

a little while ago there were a bunch of stories about Apple Silicon engineers, CPU designers, leaving the company to work for other companies, or start up their own. Could this be the downstream result of that? They've just lost the talent.... Alternatively, turn it around, were they unhappy with the business decisions and priorities that meant CPU improvements were going to languish, and then left because of that? These are the peope who'd have the pick of places to go that would let them do what they have fun doing.

[@AppleKid]

The fact that they went from 5 nm to 3 nm should mean that they should be able to fit more cores on the dye and give it to us for a good price?

[@United_Wings]

3nm is amazing

[@Garrettdx1988]

I hope they did this to get the cooling under control. In extended scenarios the A16 would throttle to be slower than the A15 when it was throttled. I remember seeing a rumor a while ago that Apple was going to add a vapor chamber cooling solution. I'm looking forward to the teardowns.

[@michealadebz]

Not true

[@piotrd.4850]

In A20....maybe..

[@booneh]

10% gain vs. 15% with much lower yields? If it’s your company which would you choose?

[@BigSharkFreak]

Pretty clear to me that they are managing to a reasonable YoY performance increase instead of Max gains in one big step. This allows them to have steady progress over time while managing the power consumption in the iPhone platform. We don’t know if they’ve actually maintained or even dropped the CPU frequency, which is a big missing piece of this discussion.

[@Gabriel-iq6ug]

Apple to focuses on GPU over CPU. It’s make sense to me, especially after the boost of performance of the efficiency core from M2

[@greenmedic88]

The only thing that comes to mind is: A) Apple is using the same or almost the same core designs, only with a 3nm die process, instead of 5nm. B) Clock speeds are throttled below capacity, so they can be throttled up in future iterations. I have no explanation as to why the more thermally efficient 3nm process is yielding the same battery life as the 5nm, unless that power consumption difference is going into the neural engine, video encoders, etc.