The big improvements are in the Neural engine. Apple talked about Metal upscaling which I assume is something like DLSS in Nvidia. Today this is hardly a feature but in the future this might give Apple big gaming advantage.
@@SzBenedek2006 Qualcomm has GSR but I'm not sure if it uses the NPU. So it might be a different way of upscaling. The A17 uses its NPU and that's what makes it similar to Nvidia in my view because bot use AI for the job.
Not just in the future, 4 AAA console games are already announced on iPhone. It's actually just the bait for the developers to show the capability, but since Unity just made the developers to upset, this is a good oportunity for Apple, since their Pro iPhones and also all of their Mac are capable to run those games (imagine Apple selling 240million phones in a year, if only half of that is a Pro model that is 120million, if you add about 30million Mac for this, that is 150million gaming capable device in EACH years, even Steam 'only' has 120 million registered users)
About the small performance change - the move from 5nm to 3nm is probably more about energy efficiency - I mean the current iPhones are easily fast enough for the purposes people use them for. I'd rather have an extra few hours of battery life than rendering an extra 10fps playing Sudoku (literally the only game I've ever played on my iPhone 14)😁 The raytracing leads me to suspect it'll be used to drive a version of their headset - I doubt anyone really cares about ultra graphic quality for mobile games. If they do, they buy a Switch or Steam Deck.
@@DrAdityaReddy When you're blending CGI into the real world it's better to have higher quality so it's not so jarring - for instance using the view of the real world through the cameras to use as reflection textures on a virtual object that's floating in front of you. It's more than just a standard AR headset, as it can become totally VR too (you can't see directly through the goggles like with other AR headsets).
Don't count on having more battery time on this one. The GPU architecture might be less efficient, maybe have with teething issues, and that's what's really running your shiny, bouncy high resolution UI.
I also think that they are mainly focusing on efficiency this year, since their SoC is strong enough, if they can lower the power consumption for the same performancw that will a big help in gaming (since 4 AAA console games are already announced)
Very good and detailed breakdown of A17 Pro. A performance comparison between the A13 Bionic and A17 Pro will be interesting to investigate, 7nm to 3nm, over double the transistor count. Apple is having to design a new GPU mostly because of scaling it in the desktop class M series Silicon. The new GPU is more efficient the more power and cores are increased.
It's actually not a given that the new 6-core GPU would perform 20% better than the previous 5-core GPU, and I'm not even talking about bottlenecking. When you're adding more cores to a processing unit there's a chance you might not be able to run those cores at the same frequency as the previous generation. With the A14 for example the Neural Engine had 16 cores as opposed to the 8 cores of the A13, but it didn't perform 2x faster, but 80% faster, even in the theoretical max speed claim, because the cores were presumably clocked lower. Them claiming the new GPU is 20% faster could mean that they either were able to keep the clocks the same, or compensate for lower clocks with instructions per clock improvement in the individual GPU cores (which would be kind of chicken or egg game since higher IPC could also result in lower clocks). Don't get me wrong I don't need to tell you any of these things, but just wanted to add to the conversation.
Maybe there will be something like performance mode for certain games. This will utilize the full performance of its GPU. Just like on PS Vita and Nintendo 3DS
We'll, actual Adreno destroyed the A16 GPU so now Apple just catches, probably won't be any faster may be on par or if faster then it won't be greater gap than 5% , and sustainable performance on all iPhones has like 60% stability? They can throw even M3 chip in to phone, it won't beat Android gaming phones even Xiaomi has 80% stability with Gen2, and that's not a gaming phone, people are freaking out about numbers but in reality iPhones are useless and I'm using iPhone 14 pro, can't even play for longer than hour because phone overheats like any other and to compensate lack of heatsink they trying different tricks screen dining below minimal manual brightness, after heavy gaming session even keyboard can lag lol, even cloud gaming would be better on Android.
It would still beat next generation SD for performance and as smartphone SoCs gets better and better, manufacturers are focusing more on efficiency and better battery life.
Relax. For those people who do not cut and edit videos using their iPhone. A15 or A14 are fast enough. However, you need some feature to justify your premium price. SoC is the easiest solution. Do not care users. Apple needs the SoC update to sell for premium price.
@@Tech-is1xy I am talking about the SoC. You are talking about innovation. Every year comes out with a SoC with a slight improvement is not innovation. TSMC has improvement on their process technology every year. Apple can just take that uses 95% of the design file and send it for to TSMC for production. Slight improvement is attained. So it is not major deal. Qualcomm does minor improvement over their SoC more or less the same. I think most Apple users want major innovation and do not like them to squeeze tooth paste.
Maybe it should be no brainer to bring ALL full Unreal Engine 5.x games with full console graphics features from temporal super resolution via Metal 3 and MetalFX, to Nanite variable geometry and Lumen ray traced lighting via hardware ray tracing instead of software solution, as well as those like Witcher fourth one, Witcher 3 complete edition, even Cyberpunk 2077 complete edition with Phantom Liberty expansion, and Project Orion, the sequel to Cyberpunk 2077 as well :)
It’s humbling to think that I’m currently holding one of the most technologically advanced and powerful system on a chips in the world, in the palm of my hand. I just wish it wasn’t paired to T-Mobile.
I`m guessing huge performance steps that happened in early days are getting much harder with the smaller and denser the transistors are packed in there , no doubt apple and other companies could ramp them up more , but as always , whether it be cars , jets , rockets or computer chips ............. in all forms speed costs money . Maybe we are getting close to the end of anything but small increases in performance ?
One comment on the increase in number of transistors: To compare the improvement of A17 vs. previous generations, we should look at the relative increase in number of transistors, not the absolute: • A17 - 19bn transistors: 3bn increase - 19% increase • A16 - 16bn transistors: 1bn increase - 7% increase • A15 - 15bn transistors: 3.2bn increase - 27% increase • A14 - 11.8bn transistors: 3.3bn increase - 39% increase • A13 - 8.5bn transistors Looking at the percentages, we see that the A14 and A15 were bigger generational leaps compared to the A17 on a relative basis. Just one comment I think is helpful, and not a critique - Thank you for a good video! 😊
Looking at that performance core frequency hitting 3.7GHz, up from 3.46GHz in A16, the microarchitectural and design changes might not give any meaningful gain. The lower process node just allows this high frequency, which is almost 10% on its own.
The details of the new Huawei are very few but there is lots of speculation and rumor. Huawei needs to up its game in terms of launch information. It is also a China only device.
Knowing apple if they didn’t gloat about how much powerful the new chip is, it’s because this may be a similar situation to last year🤔 but we have to wait and see what’s going on.
@GaryExplains Alright 👍 I guess we'll wait for more info and maybe some official statements from Huawei themselves. Anyway your videos are amazing, keep it up 👌
I know other TH-camrs can highlight other videos in a video so it would be great if you could do the same to make it easier for people to go to them when you do talk about them in the video. I believe it shows up in some info logo at the top right corner and it will also be shown with thumbnail in the description (or rather shown at the bottom after the text description). Edit: It might also be a good idea use them at the end of the video instead of the latest videos.
We are going to hit the wall of what is possible so we can’t expect huge leaps with each iteration. Plus, how much performance do we need in a cellphone?
Seems like a very meh upgrade again! Just as a wild guess, I think this could be down to cache sizes not increasing any more due to it not scaling down well,on smaller nodes and Apple ,when compared to others, has always boasted huge cache sizes to boost performance in the past.
Well they introduced hardware-accelerated ray tracing which is nice for future games. They also intoduced in-game upscaling which I assume is something like DLSS in Nvidia which again is a feature that might prove to be good for future games.
I mean, these chips are already so fast. I feel like the industry is coming to a hault in terms of of improving these chips. Apple is still comparatively ahead of the competition, so maybe they are not as motivated to push their chips further.
Apple is lacking behind compietion in Ray Tracing (this camed 2yrs before iphone), SD 8 Gen 2 is better than A16 imagine how SD 8 gen 3' GPU will perform, Snapdragon are ahead in Storage speed, AI, High res Camera support upto 300MP, USB 3.2 Support, Better Modem, Ultrasonic Fingerprint Sensor Support, better display support 4k 120HZ(sony xperia), Superfast Charging upto 240 wattts (Xiaomi) & much more.........and i'm iPhone 13 user also i regret to own an apple!
@@mightyhuman5518ray tracing does a performance hit and no one is going to do that on a mobile phone also all those other features you said are not by Qualcomm but features sold by the phone manufacturer.
Sir, Apple didn't want to compare software ray tracing with hardware ray tracing, but wanted to compare the A17 Pro ray tracing performance to the one of the A16, please don't fool yourself.
@@GaryExplains You said basically that Apple took us by fools for comparing software ray tracing with hardware ray tracing, but they were actually just comparing the performance in ray tracing between the A16 Bionic and the A17 Pro. They, as usual, compared to themselves, and there is nothing disingenuous about comparing a GPU with hardware acceleration for ray tracing with one without one, or at least, is basically as disingenuous as comparing the multi thread performance of the i9 13.900K against the i5 5.700...
If you don't think it is disingenuous to compare software ray tracing to hardware raying tracing and point out the speed increase as some kind of meaningful number, then I can't help you. But thanks for watching.
@@GaryExplains How is disingenuous? The previous iPhone is incapable to do hardware ray tracing, so you say that Apple can't compare the ray tracing capabilieties of the iPhone 15 Pro against the iPhone 14 Pro? It would be clearly disingenuous if the previous iPhone had hardware acelerated ray tracing but they didn't use it in the comparation with the new iPhone 15 Pro using hardware ray tracing.
I've heard - not sure of the source - that the SoC can be tuned towards efficiency _or_ performance. Apple's been experimenting with hardware ray tracing, but at 4nm couldn't get it inside their energy budget. Now at 3nm, they've finally shoehorned it in and I bet they're running the SoC towards the efficiency side to allow for hardware ray tracing inside the iPhone's thermal envelope. If this is true, I'm sure they'll be running the M3 microarchitecture hotter for things like Macs with fans and heatsinks where the thermal environment isn't so constrained - which gives me hope for the all the processors and IP blocks in the Macintosh line, and may bring Apple Silicon a tad closer to being competitive against discrete GPUs - though Apple Silicon will _never_ run with the energy and heat budgets of those discrete components in PCs.
apple iphones are really good and optimised in gaming even though i believe android is more powerfull in raw gpu performance. my only wish is that iphones will have bypass charging like android gaming phones.
Wait 2 years after iPhone 15 pro and you'll get it at least 500 dollars cheaper. To be honest is not worth it paying so much money for a phone. They want to sell you that you can Play games like resident Evil 4 Remake. But c'mon guys. Get a ps5 and 4k tv and still cheaper than that phone 😂
I am disappointed in many "Tech TH-camrs" who say "all apple did was add a 6th gpu core for 20% more performance".....That's literally how GPUs improve, Nvidia, AMD, Qualcomm, Intel....Literally everyone....A "core" is just apple's current measure of shader grouping......A14 Bionic had 4 cores, each with 128 graphic shaders for a total of 512 shaders @ 1GHz.....for 1 teraflop......The A15 Bionic doubled the shader count per core to 256 and gave it up to 5 cores.....which is 256 * 5 = 1280 shaders @ 600 mhz....for 1.5 Teraflops......The A16 Bionic had the same 1280 shaders, but clocks were bumped up to 700Mhz for 1.8 Tflops. Finally the A17 Pro now has 6 cores, (256 * 6) = 1536 shaders @ 700Mhz which is 2.15 Teraflops....roughly the same as the snapdragon 8 gen 2/8+ gen 2 which have 1536 shaders @ 680/719 Mhz, The Snapdragon 8 Gen 1/8+ Gen 1 had 1024 shaders @ 818/900Mhz. Nvidia groups 128 Shaders into "SMs", AMD groups 64 shaders into "CUs", Intel groups 8 shaders into "EUs".......Qualcomm just gives their gpu a number....Adreno XXX......with shader and/or clock tweaks every generation.....but because apple talks about cores, it's somehow different. Cores are only a relative measure WITHIN a generation/device group........Even the apple M2 still had 128 shaders per core.....so it's 10 cores still had the same number of shaders as the 5 Core A15/A16 Chips which have 256 per core.....the 10 core M2 just ran at 1.4Ghz.......vs the 5 cores of the iphones at 700Mhz. It's literally how GPUs improve, minor architectural changes, but the main way is using newer nodes to increase shader counts and/or clock speeds. AMDs 6nm RDNA 2, Ryzen 6000 APUs had 8 CUs (64 * 8 = 512 shaders) same number as the steam deck @ 2.2-2.4Ghz. Then AMDs released their 4nm RDNA 3, Ryzen 7000 APUs with up to 12 CUs (64 * 12 = 768 shaders) same as the Z1 Extreme @ 2.7 to 3Ghz Can you say, it was not an improvement because all they did was "add more CUs"....they did, but in the roughly the same power/thermal/price/form factor as before...so it was an improvement. Apple could have increased the number of shaders per core, but they've decided to settle at 256 for now Apple could have increased the performance by 20% in one of three ways. 1) Keep the cores the same but increase the clockspeeds by 20%.....so 5 cores @ 700 Mhz......Vs 5 cores @ 840Mhz....a less efficient approach. 2) They could have kept the cores at the same 5 cores but bumped up the "shaders per core" count by 25%....20% isn't possible as 256 * 1.2 = 307.2.......so maybe bump it up by 25%......256 * 1.25 = 320.....but since each core now has 320 shaders....the gpu would have 320 * 5 = 1600 shaders.....@ 672Mhz....which would give exactly 2.15 Tflops, the same as the 6 core A17 Pro.......and youtube "techies" would be praising apple for having "more powerful GPU cores". 3) What they actually did.....Kept their core structure and clock speeds the same, 256 per core....but added another core for 20% more shaders....but apparently this approach is disappointing. For the A17 Pro, Apple increased the GPU shader count from 1280 to 1536 within the same power/thermal/price point.....added ray tracing, added mesh shading, doubled the neural engine from 17tops to 35tops for better GPU upscaling....with a relatively minor 4nm to 3nm move....all in 1 or 2 years...and you claim they didn't do anything. The only thing that's disappointing is that logic.
@@GaryExplains Again, that's literally the main way all of GPUs get upgraded. This is going to be a bit long but I need to touch all the bases so you understand what I'm saying with examples from Nvidia, Snapdragon, Playstation etc...........Here we go........You cannot have significant performance gains gen on gen while maintaining the same clock speeds and shader count. The GTX 960 on TSMCs 28nm process had 1024 shaders.......the GTX 1060 had 1280 shaders on TSMCs 16nm process......The 2060 had 1920 shaders on TSMCs 12nm process..........Imagine if they were cores, 128 shaders per core........the 960 would have 8 cores.......the 1060 would have 10 cores.......the 2060 would have 15 cores. By your logic, from the original 28nm process, they just used the power savings of TSMCs 16nm process to add 2 more cores and raised the clocks from the 960 to 1060. By your logic, from the 16nm process, they just used the power savings of TSMCs 12nm process to add 5 more cores from the 1060 to 2060. Same with 970 to 1070 to 2070........1664 to 1920 to 2304 shaders........13 cores to 15 cores to 18 cores. Same with 980 to 1080 to 2080........2048 to 2560 to 2944 shaders..........16 cores to 20 cores to 23 cores. The RTX 2060 was roughly on par with the 1070.......both had 1920 shaders or 15 cores....at around the same clock speed. That's the point, that's literally how GPUs work, more shaders, more memory bandwidth and/or higher clockspeeds = more GPU performance....and newer nodes allow them to fit/allocate more transistors and thus more shaders into the same power/thermal budget as before.....thus gains. There are minor architectural improvements and/or IPC gains here and there.......as well as newer features like raytracing cores, tensor cores, frame generation pipeline being added/improved. But overall pure IPC gains on GPUs hover around 5% at best without any changes to the shader count/clock speed/process node. The Official Teraflop count for GPU performance literally involves multiplying (Shader Count * 2 * Clock Speed) AMD refers to every 64 shaders as a "Compute Unit" or CU. The PS4 @ 28nm......had 18 CUs or (18 * 64) = 1152 Shaders The PS5 @ 7nm...... has 36 CUs or (36 * 64) = 2304 (Same as RTX 2070) PS4 is (1152 shaders * 2 * 800 Mhz) = 1843200 Flops or 1.84 Teraflops. PS5 is (2304 shaders * 2 * 2230Mhz) = 10275840 or 10.28 Teraflops. Would you say there was no change....they just used the power savings from 28nm to 7nm to just double the core count from 18 to 36 and almost triple the clock speed? Compared to Snapdragon, if we go by apples current standard of 256 cores per shader. SD 888 SAMS 5nm (512 Shaders) 2 Cores SD 8G1 SAMS 4nm (768 Shaders) 3 Cores SD 8G2 TSMC 4nm (1280 Shaders) 5 Cores Apparently, it would have been better if apple didn't state anything about cores......and just called their GPUs........A14 GPU, A15 GPU, A16 GPU.......and thus the A17 GPU is 20% faster. Such simple minded presentations would apparently impress the masses. I love your videos and have a lot of respect for you.....but you and all the others are dead wrong with this.....But as human beings we never stop learning new information to improve ourselves. Take care.
Funny that Arm's Mail GPUs have been able to improve the performance by changing the design of the GPU, not just by adding another core. Adding another core isn't the only way to improve performance (which I think is what you are saying). So that isn't literally the way all GPUs get upgraded. But you are only human.
@@GaryExplains It seems you deleted my last comment/reply that proved you wrong with the fact that Arm Mali GPUs have increased their shader core count from 256 to 1536 (same as a17 pro)...to 2048.....hilarious.
I can assure you that I have not deleted any of your comments. However Google does have a rigorous SPAM policy and maybe Google removed it. Did it have any external links in it?
surely the point of drawing distinctions from using Hardware over software, whilst obvious and maybe some marketing fluff is to draw the attention to other manufacturers that may use software instead? This is typical I would think of older slower platforms and PCs that may not have the latest HW.
Because the high performance cores are big and use a lot of power. As a result they have great performance but if there were 4 it would kill the battery.
Even Android is not using actually 4 performance cores, they also use 1 or 2 high performance cores and 3 or 2 mid performance cores what lot of peopla also call to performance cores, but not as strong as the main core. (You probably saw the 1/3/4 or 2/2/4 core settings already)
@@leledumbo one would hope Apple is not running a skeleton team: Apple Silicon uArches go into the Watch, iPad, iMac, Mac Studio, MacBooks, Mac Pros, Apple TVs, etc. Clock speeds are an important knob, sure, but IPC & uArch is what everyone needs to target. Apple Silicon isn't slow-far from it. But if Apple's designs lag in IPC growth (and indications point to that), it'll begin to hurt their entire product line. TSMC N3 was already delayed +1 product cycle and that is liable to keep happening, so uArch is the hard, but necessary task. I'd hope Apple was hiring CPU architects like crazy, if the talent loss to NUVIA, Rivos, etc. was really that terrible.
@@ikjadoon may be, but with their attitude towards those ex-engineers, it might be tough to hire good ones. there will be connections among engineers working on the same fields, and news are spreading rapidly like a pandemic among them. I have doubts they can get at least equally capable ones. One thing for sure, whatever left so far has only been able to tweak, no more significant generation over generation breakthrough for at least 2-3 years already.
I went from the A13 Bionic chip in iPhone 11 (8.9 billion transistors) to the new iPhone 15 Pro Max (19 billion transitions). For me, waiting so long for an upgrade has meant I REALLY feel the difference with more than x2 the transistors and things like 5G, faster refresh rate Pro screen, brightness, USB-C, and more. I'm a big mobile gamer so that part is amazing for me!
Third year in a row with a very underwhelming CPU upgrade. Probably the worst considering likely no efficiency improvement. Looks like Apple's CPU team still haven't recovered from the HUGE loss of talent to NUVIA. Even the GPU upgrade is underwhelming and just about on par with Qualcomm's 1 year old 8g2 (the 8g3 is a few months away)
So.... cooling isn't a problem in the smaller dimensions of the iPhone 15............ Imagine two...four......3nm "M" SoCs in the roomier and cooler MacBook Pro ? The relatively HUGE case and cooling of the Mac Studios should allow Apple to crank up the clock rate and even add MANY more CPU / GPU cores without heat dissipation issues. I think Apple hobbled the iPhone 15 Pro. It's no slouch....but with the exception of the A17...meh. They really just glossed over 3nm and didn't hammer it as a marketing tool. I bet the first "big show" for 3nm will be the next iPads .... 3nm chips in iPads will start to make laptops redundant for a growing demographic. Apple has already convinced a majority of Under-25 Consumers to stare at their phones...all day. In a few short years your HOME computer will be an iPad...that "talks" to your GIANT screen TV.
I don't think that is true. a) Arm's chips have shown double-digit performance increases for several years and continue to do so. b) Qualcomm is releasing its new Nuvia chip soon. c) Intel and AMD continue to release new processors regularly and each generation is better than the previous. d) Lots of GPU development. e) NVIDIA's Superchips (Grace Hopper etc).
HW based Raytracing of A17 is just three times faster, but four times as fast as SW based raytracing. „4 times faster“ is wrong Apple‘s marketing speech. As recognized, some experts, in the event, have learned and correctly stating the performance advance this way, although text says different… Apple does not want to lie anymore, maybe.
@@oo--7714 heat dissipation is important, iPhone don't have it at all and that's why when you run benchmark 2 time's IPhone's have actually the worst stability of all mobile devices, Gen2 for example in Xiaomi has 80~ stability that means better sustainable performance, I'm using iPhone 14 pro so I know that iPhone is just trash, people can say whatever they want, I switched from Android and I regret it, that's because of the marketing and people praising apple, next phone will be Android phone, Sony, Samsung or maybe Xiaomi, eventually gaming phone.
What I believe is this chip could potentially do much better but apple is intentionally down clocking it to restrict the performance a17 pro theoretically could run AAA games…so because of thermal envelope they might be limiting its performance and we might see its full potential when this comes in an ipad…we know m1 was also just tweeked version from a14 and with thermal envelop and active cooling it produced really good performance.
What evidence do you have that Apple is intentionally down clocking it to restrict the performance? As far as I know this is the highest clock mobile CPU that Apple has ever made.
@@GaryExplains after the launch 15 pro had overheating issue that other bionic series chips haven't had in past couple of years...surely the engineers at Apple must have noticed this before the launch also..so its just a rough speculation that seeking the overheating problems they could have down clocked it....a better way to test this theory could be to monitor clock cycles and thermal throttling when running benchmarks...hope to see new video demonstrating a17 pro benchmarks😉
All processors are "intentionally down clocked" because of thermal restraints. That includes Apple and Qualcomm and MediaTek, etc. That is just a fact of life when creating passively cooled smartphones.
To me, it seems the A17Pro design isn't finalised yet, and what we see in the iPhone 15 Pro is only incremental improvements; with the fully developed A17Pro released in the next versions of Apple devices. As an example, the Snapdragon 888 and the 888Gen1 were precursors to the true fully developed SoC 888Gen2, which is a very respectable piece of silicon - with the accelerators working as intended, and with on-chip data transfers optimised to the top. I would prefer the SoC design and fab to release a new chip every three years, instead of this forced release of a new SoC every year for commercial reasons. Dancing the waltzer, you do two step forward, and one back. Dancing the SoC, you do two steps around, and one forward. Thank you for the very illuminating video Mr. Gary - both in raster and raytrace modalities...
I want to see AMD enter the ARM soc market now that they own Xilinx they can provide competition and choice for phone manufacturers. I am interested in seeing triple A games like Resident Evil 4 running on the 15 pro.
3nm is a big efficiency upgrade we can expect much much better battery than the 14 pro because 14 pro said it had a 4nm chip but it was a tsmc node that was actually 5nm but with a deifferent name
@paoxiong6146 This is a direct quote from TSMC's website, "In addition, TSMC plans to launch 4nm (N4) technology, an enhanced version of N5 technology." The N4 process is listed under the N5 section of its website.
I'm not following battery tests anymore but isn't that the iPhone 13 max ultra or whatever is the name has still better battery than any iPhone 14 variant? I have iPhone 14 pro and indeed battery is good, but that comes with some costs, but no one talks about background activities as no one downloads 1k songs on Spotify at once, on Android I could do that watch something on YT or just play game and it carries on, but on iPhone it doesn't work lol I'm disappointed with iPhone since day one they updating it a lot and while they trying to fix some bugs they add new one's, and that lagging keyboard, I installed Google g, and while switching between keyboards I had massive input lagging, before last updated my phone dropped screen refresh rate to less than 30fps with stutter and lag without any reason, latest update fixed it
Thx for the report Gary! Do you think Apple will eventually hit with or without adopting latest ARM architecture limitation on optimizing their CPU core stack which will be significant enough over 2 digits? It's not that common with all processors in the industry to see big over 2 digits gen 2 gen increase in performance with more efficient and optimized architecture. Shrinking transistors and increasing their density per mm2 is slowing down so it looks smart and well optimized architecture design will crucial and most important besides advanced chip packaging techniques (3D packaging etc.).
What kind if further price increase? Did you aware of that Apple not increased the price since 2017..? Even this $100 on the iPhone 15 Pro Max is not a price increase, they just simply removed the 128GB version, and they replaced it with the 256GB, what was also +$100 even in last year.. (so the iPhone 14 Pro Max 256GB and the iPhone 15 Pro Max 256GB has the same price, technicaly not a price increase..)
@@U_H89 did you already forget that when they did this the first time..? Even the regular iPhone 14 got the iPhone 13 Pro chip instead of the regular iPhone 13 chip.. (the Pro chip also had one more GPU core), and since they not increased their price in the past 7 years, why do you expect this price increase just for that chip..?
A small ARM core is less than half the size of a piece of rice. A USB controller should be even smaller. I think a tenth of a piece of rice is enough. Apple's USB controller is a cheap one thus it will not be large. USB 3 nowadays are cheap stuff. However the coming USB 5 controller will be a different story.
Is it honestly better than a15 bionic? to be honest I think a15 and newer soc are almost the same in performance and also in real life most won’t see a big difference maybe in graphics but in speed it’s going to be about the same
Does apple take the armv9 on A17 Pro? Arm announces armv9 way back in 2021. A15 Bionic is said to come with that but it didn't happened it remains at the same armv8.5-A from the A14 Bionic and A16 Bionic from last year comes with slightly upgraded armv8.6-A.
ARM v9 has been out for quite some time now. It's surprising that they didn't use it. Though it would probably have been hard to figure out performance improvements even if they did, because Apple heavily customizes the ARM designs.
@@DK-ox7ze Because Apple' processors need to be 100% compatible and have to pass compatibility tests. At the moment Apple made Armv8.x compatible processors.
@@GaryExplains If they have their own design then why is the compatibility required? Is it because they are using ARM's instruction set? Or they use something else from ARM as well?
technically not 4nm on the A16 Bionic, its 5nm process using their 4NP process or something like that, so yeah a marketing term, but I think this is actually now 3nm, right? TSMC themselves quoted an increase of 15% going from 5N to 3N, I think they were somewhere in the middle so 10% kind of makes sense?
Using your analysis, 10% would make sense if this was an A16 made on 3nm, but it isn't, Apple said there were also microarchitecture design changes. I was expecting performance increases with the move to 3 nanometres and performance increases because it is a better CPU. 🤷♂️
The big improvements are in the Neural engine. Apple talked about Metal upscaling which I assume is something like DLSS in Nvidia. Today this is hardly a feature but in the future this might give Apple big gaming advantage.
True.
I think Qualcomm is working or does have a similar feature.
@@SzBenedek2006 Qualcomm has GSR but I'm not sure if it uses the NPU. So it might be a different way of upscaling. The A17 uses its NPU and that's what makes it similar to Nvidia in my view because bot use AI for the job.
The A17 Pro can make 35 Trillion Operations per second, 2 times more than using the new A16 Bionic 😂
Not just in the future, 4 AAA console games are already announced on iPhone. It's actually just the bait for the developers to show the capability, but since Unity just made the developers to upset, this is a good oportunity for Apple, since their Pro iPhones and also all of their Mac are capable to run those games (imagine Apple selling 240million phones in a year, if only half of that is a Pro model that is 120million, if you add about 30million Mac for this, that is 150million gaming capable device in EACH years, even Steam 'only' has 120 million registered users)
About the small performance change - the move from 5nm to 3nm is probably more about energy efficiency - I mean the current iPhones are easily fast enough for the purposes people use them for. I'd rather have an extra few hours of battery life than rendering an extra 10fps playing Sudoku (literally the only game I've ever played on my iPhone 14)😁 The raytracing leads me to suspect it'll be used to drive a version of their headset - I doubt anyone really cares about ultra graphic quality for mobile games. If they do, they buy a Switch or Steam Deck.
Why would ray tracing be needed for ar headset? 🤔
@@DrAdityaReddy When you're blending CGI into the real world it's better to have higher quality so it's not so jarring - for instance using the view of the real world through the cameras to use as reflection textures on a virtual object that's floating in front of you. It's more than just a standard AR headset, as it can become totally VR too (you can't see directly through the goggles like with other AR headsets).
Don't count on having more battery time on this one. The GPU architecture might be less efficient, maybe have with teething issues, and that's what's really running your shiny, bouncy high resolution UI.
Apple says the battery life will remain the same.
I also think that they are mainly focusing on efficiency this year, since their SoC is strong enough, if they can lower the power consumption for the same performancw that will a big help in gaming (since 4 AAA console games are already announced)
Coming from iPhone X, it’s a huge upgrade for me lol 😂
Damn. I don't know how and why you even kept that phone so long. It's still on 4G the whole time.
@asianstud7 LOL, I don't even have 5G coverage in my area, the iPhone X works just as well here as any other pho9ne even those with 5G modems!!! 😂
@@GaryExplainsi have 5g coverage and im sure 5g is amazing but Ive just had a 4g phone this whole time (11 pro) and been just fine
Very good and detailed breakdown of A17 Pro.
A performance comparison between the A13 Bionic and A17 Pro will be interesting to investigate, 7nm to 3nm, over double the transistor count.
Apple is having to design a new GPU mostly because of scaling it in the desktop class M series Silicon. The new GPU is more efficient the more power and cores are increased.
It's actually not a given that the new 6-core GPU would perform 20% better than the previous 5-core GPU, and I'm not even talking about bottlenecking. When you're adding more cores to a processing unit there's a chance you might not be able to run those cores at the same frequency as the previous generation. With the A14 for example the Neural Engine had 16 cores as opposed to the 8 cores of the A13, but it didn't perform 2x faster, but 80% faster, even in the theoretical max speed claim, because the cores were presumably clocked lower. Them claiming the new GPU is 20% faster could mean that they either were able to keep the clocks the same, or compensate for lower clocks with instructions per clock improvement in the individual GPU cores (which would be kind of chicken or egg game since higher IPC could also result in lower clocks). Don't get me wrong I don't need to tell you any of these things, but just wanted to add to the conversation.
Maybe there will be something like performance mode for certain games. This will utilize the full performance of its GPU.
Just like on PS Vita and Nintendo 3DS
That’s ~6W TDP limit. More cores and smaller node = same TDP but more cores
We'll, actual Adreno destroyed the A16 GPU so now Apple just catches, probably won't be any faster may be on par or if faster then it won't be greater gap than 5% , and sustainable performance on all iPhones has like 60% stability? They can throw even M3 chip in to phone, it won't beat Android gaming phones even Xiaomi has 80% stability with Gen2, and that's not a gaming phone, people are freaking out about numbers but in reality iPhones are useless and I'm using iPhone 14 pro, can't even play for longer than hour because phone overheats like any other and to compensate lack of heatsink they trying different tricks screen dining below minimal manual brightness, after heavy gaming session even keyboard can lag lol, even cloud gaming would be better on Android.
@@ThereIsOnlyOneRandom then don’t play mobile games they are such a waste of time 😂
@@ijackie3420 By that logic the A14 also switched from 7nm to 5nm, but it still didn't perform 2x better with 2x the cores.
"Up to 10%" is really disappointing.
But they make up for it by calling it pro 😂
It would still beat next generation SD for performance and as smartphone SoCs gets better and better, manufacturers are focusing more on efficiency and better battery life.
Up to 33% disappointing
Not to mention it’s only on the 2 performance cores.
At least would like to see a 20% higher battery life
Always look forward to your videos go job on explaining the new Chip.
Glad you like them!
A17 is a must for TikTok, FB and the Calculator...
You forgot calendar
@@bbcringegaming Sorry... my bad...
Relax. For those people who do not cut and edit videos using their iPhone. A15 or A14 are fast enough. However, you need some feature to justify your premium price. SoC is the easiest solution. Do not care users. Apple needs the SoC update to sell for premium price.
@@catchnkillso you would rather apple not innovate. And give you a 3 year old processor in a 2023 flagship.
@@Tech-is1xy I am talking about the SoC. You are talking about innovation. Every year comes out with a SoC with a slight improvement is not innovation. TSMC has improvement on their process technology every year. Apple can just take that uses 95% of the design file and send it for to TSMC for production. Slight improvement is attained. So it is not major deal. Qualcomm does minor improvement over their SoC more or less the same. I think most Apple users want major innovation and do not like them to squeeze tooth paste.
Can't wait for geekerwan to test this
I was hoping for GPU ray tracing! Wasn’t disappointed! Will be fun to see the M3 desktop class versions next! 🤯🤯🤯😎😎😎👍👍👍
And unlike RTX laptops, you can have all those RT performance on the go!!!
Maybe it should be no brainer to bring ALL full Unreal Engine 5.x games with full console graphics features from temporal super resolution via Metal 3 and MetalFX, to Nanite variable geometry and Lumen ray traced lighting via hardware ray tracing instead of software solution, as well as those like Witcher fourth one, Witcher 3 complete edition, even Cyberpunk 2077 complete edition with Phantom Liberty expansion, and Project Orion, the sequel to Cyberpunk 2077 as well :)
I need subtitles😅😅I always find it difficult to understand your accent
It’s humbling to think that I’m currently holding one of the most technologically advanced and powerful system on a chips in the world, in the palm of my hand. I just wish it wasn’t paired to T-Mobile.
I`m guessing huge performance steps that happened in early days are getting much harder with the smaller and denser the transistors are packed in there , no doubt apple and other companies could ramp them up more , but as always , whether it be cars , jets , rockets or computer chips ............. in all forms speed costs money .
Maybe we are getting close to the end of anything but small increases in performance ?
At least the A17 finally comes with 8gb of ram.
One comment on the increase in number of transistors: To compare the improvement of A17 vs. previous generations, we should look at the relative increase in number of transistors, not the absolute:
• A17 - 19bn transistors: 3bn increase - 19% increase
• A16 - 16bn transistors: 1bn increase - 7% increase
• A15 - 15bn transistors: 3.2bn increase - 27% increase
• A14 - 11.8bn transistors: 3.3bn increase - 39% increase
• A13 - 8.5bn transistors
Looking at the percentages, we see that the A14 and A15 were bigger generational leaps compared to the A17 on a relative basis.
Just one comment I think is helpful, and not a critique - Thank you for a good video! 😊
Very good point, I will make sure to include that going forward where appropriate. Thanks 👍.
All that incredible power, just a damn shame that Apple can't program or do adequate quality control. Even wordscapes is a jittery mess.
Looking at that performance core frequency hitting 3.7GHz, up from 3.46GHz in A16, the microarchitectural and design changes might not give any meaningful gain. The lower process node just allows this high frequency, which is almost 10% on its own.
Gary is one cool uncle everyone likes
lol , just as it came out but p60 is out how long but almost every phone youtuber is silent...smh you ppl.
The details of the new Huawei are very few but there is lots of speculation and rumor. Huawei needs to up its game in terms of launch information. It is also a China only device.
Bottom line, if you have an iPhone 13 or 14 don’t upgrade
Knowing apple if they didn’t gloat about how much powerful the new chip is, it’s because this may be a similar situation to last year🤔 but we have to wait and see what’s going on.
Gary explains how there's nothing new about the iPhone 15 😪
Is there any info about Huawei's Kirin 9000s? Would be good to see a review on that chip.
There is little information at the moment just lots of rumors and speculation.
@GaryExplains Alright 👍 I guess we'll wait for more info and maybe some official statements from Huawei themselves. Anyway your videos are amazing, keep it up 👌
I know other TH-camrs can highlight other videos in a video so it would be great if you could do the same to make it easier for people to go to them when you do talk about them in the video.
I believe it shows up in some info logo at the top right corner and it will also be shown with thumbnail in the description (or rather shown at the bottom after the text description).
Edit: It might also be a good idea use them at the end of the video instead of the latest videos.
We are going to hit the wall of what is possible so we can’t expect huge leaps with each iteration. Plus, how much performance do we need in a cellphone?
FINALLY... thank you Gary, no one but you talks about real tech details 😅
Seems like a very meh upgrade again! Just as a wild guess, I think this could be down to cache sizes not increasing any more due to it not scaling down well,on smaller nodes and Apple ,when compared to others, has always boasted huge cache sizes to boost performance in the past.
Well they introduced hardware-accelerated ray tracing which is nice for future games. They also intoduced in-game upscaling which I assume is something like DLSS in Nvidia which again is a feature that might prove to be good for future games.
I think you're spot on! Memory silicon density hit a brick wall when TSMC got to their 5nm node. CPU cache has stopped shrinking.
You need bigger caches to avoid memory bottlenecks.
@@ThePowerLoverThat's why A16 went to 24MB system cache vs 32MB on A15. A16 introduced LPDDR5.
This IS Qualcomm's chance to leap frog Apple. Lets goooooooooooooo!!
We need you to revive anandtech's SoC review deep dive
You forget the most important thing, 😂 The A17 Pro can make 35 Trillion Operations per second, 2 times more than using the new A16 Bionic
That number is meaningless, which is why I didn't mention it. You felt that the NPU was slow on the previous iPhone? 🤷♂️
I mean, these chips are already so fast. I feel like the industry is coming to a hault in terms of of improving these chips. Apple is still comparatively ahead of the competition, so maybe they are not as motivated to push their chips further.
Apple is lacking behind compietion in Ray Tracing (this camed 2yrs before iphone), SD 8 Gen 2 is better than A16 imagine how SD 8 gen 3' GPU will perform, Snapdragon are ahead in Storage speed, AI, High res Camera support upto 300MP, USB 3.2 Support, Better Modem, Ultrasonic Fingerprint Sensor Support, better display support 4k 120HZ(sony xperia), Superfast Charging upto 240 wattts (Xiaomi) & much more.........and i'm iPhone 13 user also i regret to own an apple!
@@mightyhuman5518ray tracing does a performance hit and no one is going to do that on a mobile phone also all those other features you said are not by Qualcomm but features sold by the phone manufacturer.
Sir, Apple didn't want to compare software ray tracing with hardware ray tracing, but wanted to compare the A17 Pro ray tracing performance to the one of the A16, please don't fool yourself.
Eh?
@@GaryExplains You said basically that Apple took us by fools for comparing software ray tracing with hardware ray tracing, but they were actually just comparing the performance in ray tracing between the A16 Bionic and the A17 Pro. They, as usual, compared to themselves, and there is nothing disingenuous about comparing a GPU with hardware acceleration for ray tracing with one without one, or at least, is basically as disingenuous as comparing the multi thread performance of the i9 13.900K against the i5 5.700...
If you don't think it is disingenuous to compare software ray tracing to hardware raying tracing and point out the speed increase as some kind of meaningful number, then I can't help you. But thanks for watching.
@@GaryExplains How is disingenuous? The previous iPhone is incapable to do hardware ray tracing, so you say that Apple can't compare the ray tracing capabilieties of the iPhone 15 Pro against the iPhone 14 Pro? It would be clearly disingenuous if the previous iPhone had hardware acelerated ray tracing but they didn't use it in the comparation with the new iPhone 15 Pro using hardware ray tracing.
@@ThePowerLover As I said, I can't help you.
Please bring back speedtest G
I've heard - not sure of the source - that the SoC can be tuned towards efficiency _or_ performance.
Apple's been experimenting with hardware ray tracing, but at 4nm couldn't get it inside their energy budget. Now at 3nm, they've finally shoehorned it in and I bet they're running the SoC towards the efficiency side to allow for hardware ray tracing inside the iPhone's thermal envelope.
If this is true, I'm sure they'll be running the M3 microarchitecture hotter for things like Macs with fans and heatsinks where the thermal environment isn't so constrained - which gives me hope for the all the processors and IP blocks in the Macintosh line, and may bring Apple Silicon a tad closer to being competitive against discrete GPUs - though Apple Silicon will _never_ run with the energy and heat budgets of those discrete components in PCs.
apple iphones are really good and optimised in gaming even though i believe android is more powerfull in raw gpu performance. my only wish is that iphones will have bypass charging like android gaming phones.
4:48 👍🏾 Apple knows there’s a demographic that hangs to their devices forever.
Significant improvement in gpu from 5 core to 6 core gpu with support for hardware based Raytracing a 20 percent increase. Lets see how it goes...
Wait 2 years after iPhone 15 pro and you'll get it at least 500 dollars cheaper.
To be honest is not worth it paying so much money for a phone.
They want to sell you that you can Play games like resident Evil 4 Remake.
But c'mon guys.
Get a ps5 and 4k tv and still cheaper than that phone 😂
I am disappointed in many "Tech TH-camrs" who say "all apple did was add a 6th gpu core for 20% more performance".....That's literally how GPUs improve, Nvidia, AMD, Qualcomm, Intel....Literally everyone....A "core" is just apple's current measure of shader grouping......A14 Bionic had 4 cores, each with 128 graphic shaders for a total of 512 shaders @ 1GHz.....for 1 teraflop......The A15 Bionic doubled the shader count per core to 256 and gave it up to 5 cores.....which is 256 * 5 = 1280 shaders @ 600 mhz....for 1.5 Teraflops......The A16 Bionic had the same 1280 shaders, but clocks were bumped up to 700Mhz for 1.8 Tflops.
Finally the A17 Pro now has 6 cores, (256 * 6) = 1536 shaders @ 700Mhz which is 2.15 Teraflops....roughly the same as the snapdragon 8 gen 2/8+ gen 2 which have 1536 shaders @ 680/719 Mhz, The Snapdragon 8 Gen 1/8+ Gen 1 had 1024 shaders @ 818/900Mhz.
Nvidia groups 128 Shaders into "SMs", AMD groups 64 shaders into "CUs", Intel groups 8 shaders into "EUs".......Qualcomm just gives their gpu a number....Adreno XXX......with shader and/or clock tweaks every generation.....but because apple talks about cores, it's somehow different. Cores are only a relative measure WITHIN a generation/device group........Even the apple M2 still had 128 shaders per core.....so it's 10 cores still had the same number of shaders as the 5 Core A15/A16 Chips which have 256 per core.....the 10 core M2 just ran at 1.4Ghz.......vs the 5 cores of the iphones at 700Mhz.
It's literally how GPUs improve, minor architectural changes, but the main way is using newer nodes to increase shader counts and/or clock speeds.
AMDs 6nm RDNA 2, Ryzen 6000 APUs had 8 CUs (64 * 8 = 512 shaders) same number as the steam deck @ 2.2-2.4Ghz.
Then AMDs released their 4nm RDNA 3, Ryzen 7000 APUs with up to 12 CUs (64 * 12 = 768 shaders) same as the Z1 Extreme @ 2.7 to 3Ghz
Can you say, it was not an improvement because all they did was "add more CUs"....they did, but in the roughly the same power/thermal/price/form factor as before...so it was an improvement.
Apple could have increased the number of shaders per core, but they've decided to settle at 256 for now
Apple could have increased the performance by 20% in one of three ways.
1) Keep the cores the same but increase the clockspeeds by 20%.....so 5 cores @ 700 Mhz......Vs 5 cores @ 840Mhz....a less efficient approach.
2) They could have kept the cores at the same 5 cores but bumped up the "shaders per core" count by 25%....20% isn't possible as 256 * 1.2 = 307.2.......so maybe bump it up by 25%......256 * 1.25 = 320.....but since each core now has 320 shaders....the gpu would have 320 * 5 = 1600 shaders.....@ 672Mhz....which would give exactly 2.15 Tflops, the same as the 6 core A17 Pro.......and youtube "techies" would be praising apple for having "more powerful GPU cores".
3) What they actually did.....Kept their core structure and clock speeds the same, 256 per core....but added another core for 20% more shaders....but apparently this approach is disappointing.
For the A17 Pro, Apple increased the GPU shader count from 1280 to 1536 within the same power/thermal/price point.....added ray tracing, added mesh shading, doubled the neural engine from 17tops to 35tops for better GPU upscaling....with a relatively minor 4nm to 3nm move....all in 1 or 2 years...and you claim they didn't do anything. The only thing that's disappointing is that logic.
They added another core while moving to 3nm... not exactly amazing, Apple just used the power savings on 3nm to add the extra core. 🤷♂️
@@GaryExplains Again, that's literally the main way all of GPUs get upgraded. This is going to be a bit long but I need to touch all the bases so you understand what I'm saying with examples from Nvidia, Snapdragon, Playstation etc...........Here we go........You cannot have significant performance gains gen on gen while maintaining the same clock speeds and shader count. The GTX 960 on TSMCs 28nm process had 1024 shaders.......the GTX 1060 had 1280 shaders on TSMCs 16nm process......The 2060 had 1920 shaders on TSMCs 12nm process..........Imagine if they were cores, 128 shaders per core........the 960 would have 8 cores.......the 1060 would have 10 cores.......the 2060 would have 15 cores.
By your logic, from the original 28nm process, they just used the power savings of TSMCs 16nm process to add 2 more cores and raised the clocks from the 960 to 1060.
By your logic, from the 16nm process, they just used the power savings of TSMCs 12nm process to add 5 more cores from the 1060 to 2060.
Same with 970 to 1070 to 2070........1664 to 1920 to 2304 shaders........13 cores to 15 cores to 18 cores.
Same with 980 to 1080 to 2080........2048 to 2560 to 2944 shaders..........16 cores to 20 cores to 23 cores.
The RTX 2060 was roughly on par with the 1070.......both had 1920 shaders or 15 cores....at around the same clock speed.
That's the point, that's literally how GPUs work, more shaders, more memory bandwidth and/or higher clockspeeds = more GPU performance....and newer nodes allow them to fit/allocate more transistors and thus more shaders into the same power/thermal budget as before.....thus gains.
There are minor architectural improvements and/or IPC gains here and there.......as well as newer features like raytracing cores, tensor cores, frame generation pipeline being added/improved.
But overall pure IPC gains on GPUs hover around 5% at best without any changes to the shader count/clock speed/process node.
The Official Teraflop count for GPU performance literally involves multiplying (Shader Count * 2 * Clock Speed)
AMD refers to every 64 shaders as a "Compute Unit" or CU.
The PS4 @ 28nm......had 18 CUs or (18 * 64) = 1152 Shaders
The PS5 @ 7nm...... has 36 CUs or (36 * 64) = 2304 (Same as RTX 2070)
PS4 is (1152 shaders * 2 * 800 Mhz) = 1843200 Flops or 1.84 Teraflops.
PS5 is (2304 shaders * 2 * 2230Mhz) = 10275840 or 10.28 Teraflops.
Would you say there was no change....they just used the power savings from 28nm to 7nm to just double the core count from 18 to 36 and almost triple the clock speed?
Compared to Snapdragon, if we go by apples current standard of 256 cores per shader.
SD 888 SAMS 5nm (512 Shaders) 2 Cores
SD 8G1 SAMS 4nm (768 Shaders) 3 Cores
SD 8G2 TSMC 4nm (1280 Shaders) 5 Cores
Apparently, it would have been better if apple didn't state anything about cores......and just called their GPUs........A14 GPU, A15 GPU, A16 GPU.......and thus the A17 GPU is 20% faster.
Such simple minded presentations would apparently impress the masses.
I love your videos and have a lot of respect for you.....but you and all the others are dead wrong with this.....But as human beings we never stop learning new information to improve ourselves. Take care.
Funny that Arm's Mail GPUs have been able to improve the performance by changing the design of the GPU, not just by adding another core. Adding another core isn't the only way to improve performance (which I think is what you are saying). So that isn't literally the way all GPUs get upgraded. But you are only human.
@@GaryExplains It seems you deleted my last comment/reply that proved you wrong with the fact that Arm Mali GPUs have increased their shader core count from 256 to 1536 (same as a17 pro)...to 2048.....hilarious.
I can assure you that I have not deleted any of your comments. However Google does have a rigorous SPAM policy and maybe Google removed it. Did it have any external links in it?
surely the point of drawing distinctions from using Hardware over software, whilst obvious and maybe some marketing fluff is to draw the attention to other manufacturers that may use software instead? This is typical I would think of older slower platforms and PCs that may not have the latest HW.
I care for the hardware i wont be spending that much on a iphone , ill keep my 12 mini & my galaxy 24 ultra
Why apple did not use 4x performance core??
Because the high performance cores are big and use a lot of power. As a result they have great performance but if there were 4 it would kill the battery.
Even Android is not using actually 4 performance cores, they also use 1 or 2 high performance cores and 3 or 2 mid performance cores what lot of peopla also call to performance cores, but not as strong as the main core. (You probably saw the 1/3/4 or 2/2/4 core settings already)
@@GaryExplainsbut wouldn’t the extra power draw be mitigated because it’s now using a 3nm process ?
@@Tech-is1xy I doubt it and I guess the same argument could have been used with 7nm to 5nm, but 6 cores gas been the setup for many generations.
Apple: Never mind the speed, feel the quantity of the transistors.
Probably more feel the efficiency. In terms of cpu a17 will be the champ still.
This launce was really disappointing
Gary! Love your detail videos like this.
Glad you like them!
Apple silicon = new intel.
Barely any IPC improvements, just riding the wave of process shrinks.
Well, they lose all their best people who have been advancing previous A series. What can they do?
@@leledumbo one would hope Apple is not running a skeleton team: Apple Silicon uArches go into the Watch, iPad, iMac, Mac Studio, MacBooks, Mac Pros, Apple TVs, etc. Clock speeds are an important knob, sure, but IPC & uArch is what everyone needs to target. Apple Silicon isn't slow-far from it. But if Apple's designs lag in IPC growth (and indications point to that), it'll begin to hurt their entire product line. TSMC N3 was already delayed +1 product cycle and that is liable to keep happening, so uArch is the hard, but necessary task.
I'd hope Apple was hiring CPU architects like crazy, if the talent loss to NUVIA, Rivos, etc. was really that terrible.
@@ikjadoon may be, but with their attitude towards those ex-engineers, it might be tough to hire good ones. there will be connections among engineers working on the same fields, and news are spreading rapidly like a pandemic among them. I have doubts they can get at least equally capable ones. One thing for sure, whatever left so far has only been able to tweak, no more significant generation over generation breakthrough for at least 2-3 years already.
Wait some days...
It is a joy that the United States has abandoned the obsolete English system and is using the international system of units.
I went from the A13 Bionic chip in iPhone 11 (8.9 billion transistors) to the new iPhone 15 Pro Max (19 billion transitions). For me, waiting so long for an upgrade has meant I REALLY feel the difference with more than x2 the transistors and things like 5G, faster refresh rate Pro screen, brightness, USB-C, and more. I'm a big mobile gamer so that part is amazing for me!
Bionic sounds way better than pro!!!!
Armv9???
Nobody can tell.
We need speedtest G
Innovation is over. Apple can't do it anymore
power efficiency is too bad in Apple chips 👎🏼
19 b transistors on a mobile soc is wild.
They will keep the next 10% for next year.
Third year in a row with a very underwhelming CPU upgrade. Probably the worst considering likely no efficiency improvement. Looks like Apple's CPU team still haven't recovered from the HUGE loss of talent to NUVIA. Even the GPU upgrade is underwhelming and just about on par with Qualcomm's 1 year old 8g2 (the 8g3 is a few months away)
So.... cooling isn't a problem in the smaller dimensions of the iPhone 15............ Imagine two...four......3nm "M" SoCs in the roomier and cooler MacBook Pro ? The relatively HUGE case and cooling of the Mac Studios should allow Apple to crank up the clock rate and even add MANY more CPU / GPU cores without heat dissipation issues. I think Apple hobbled the iPhone 15 Pro. It's no slouch....but with the exception of the A17...meh. They really just glossed over 3nm and didn't hammer it as a marketing tool. I bet the first "big show" for 3nm will be the next iPads .... 3nm chips in iPads will start to make laptops redundant for a growing demographic. Apple has already convinced a majority of Under-25 Consumers to stare at their phones...all day. In a few short years your HOME computer will be an iPad...that "talks" to your GIANT screen TV.
There’s stagnation in the west regarding chip designs despite the enormous amount of taxpayer money thrown for academic research.
I don't think that is true. a) Arm's chips have shown double-digit performance increases for several years and continue to do so. b) Qualcomm is releasing its new Nuvia chip soon. c) Intel and AMD continue to release new processors regularly and each generation is better than the previous. d) Lots of GPU development. e) NVIDIA's Superchips (Grace Hopper etc).
HW based Raytracing of A17 is just three times faster, but four times as fast as SW based raytracing. „4 times faster“ is wrong Apple‘s marketing speech. As recognized, some experts, in the event, have learned and correctly stating the performance advance this way, although text says different… Apple does not want to lie anymore, maybe.
When should we expect 2nm and 1nm mobile CPU's ? And what goes beyond 1nm technology if technically possible?
2nm in 2026
@@DrAdityaReddyby the time lpddr6 comes, by then phones even mid range ones will have ps4 like performance.
Lpddr6 is just a year or two away. @@oo--7714
@@oo--7714 heat dissipation is important, iPhone don't have it at all and that's why when you run benchmark 2 time's IPhone's have actually the worst stability of all mobile devices, Gen2 for example in Xiaomi has 80~ stability that means better sustainable performance, I'm using iPhone 14 pro so I know that iPhone is just trash, people can say whatever they want, I switched from Android and I regret it, that's because of the marketing and people praising apple, next phone will be Android phone, Sony, Samsung or maybe Xiaomi, eventually gaming phone.
we need more speed test g
Is there an architectural change??// or it's just more GPU core and higher CPU frequency?
CPU no change, 1 additional core to GPU, new USB C controller instead of lightning.
And going from 4nm to 3nm
There are.
yeah.. rumors say, TSMC, Samsung both 3nm are around same 50ish yield level.@@steveseidel9967
chinas pissed
That is actually wild, it has more transistors than the m1 chip
M1 is based on a14 aka a14x
M2 is based on a15 aka a15x
M3 is based on a16 aka a16x
I cant wait for Qualcomm to release Nuvia SOC's and crush Apple once and for all.
It will certainly be interesting to see the real world performance numbers.
@@GaryExplains Truly exciting times ahead
what's the range of USB 4? how far can you send 40 Gbps?
Without Ray tracing lighting does not have to be baked in. I guess it might be on some mobile games.
What I believe is this chip could potentially do much better but apple is intentionally down clocking it to restrict the performance a17 pro theoretically could run AAA games…so because of thermal envelope they might be limiting its performance and we might see its full potential when this comes in an ipad…we know m1 was also just tweeked version from a14 and with thermal envelop and active cooling it produced really good performance.
What evidence do you have that Apple is intentionally down clocking it to restrict the performance? As far as I know this is the highest clock mobile CPU that Apple has ever made.
@@GaryExplains after the launch 15 pro had overheating issue that other bionic series chips haven't had in past couple of years...surely the engineers at Apple must have noticed this before the launch also..so its just a rough speculation that seeking the overheating problems they could have down clocked it....a better way to test this theory could be to monitor clock cycles and thermal throttling when running benchmarks...hope to see new video demonstrating a17 pro benchmarks😉
All processors are "intentionally down clocked" because of thermal restraints. That includes Apple and Qualcomm and MediaTek, etc. That is just a fact of life when creating passively cooled smartphones.
Waiting for the day when chipset will be made based on 0.000001 nm.
To me, it seems the A17Pro design isn't finalised yet, and what we see in the iPhone 15 Pro is only incremental improvements; with the fully developed A17Pro released in the next versions of Apple devices.
As an example, the Snapdragon 888 and the 888Gen1 were precursors to the true fully developed SoC 888Gen2, which is a very respectable piece of silicon - with the accelerators working as intended, and with on-chip data transfers optimised to the top.
I would prefer the SoC design and fab to release a new chip every three years, instead of this forced release of a new SoC every year for commercial reasons.
Dancing the waltzer, you do two step forward, and one back. Dancing the SoC, you do two steps around, and one forward. Thank you for the very illuminating video Mr. Gary - both in raster and raytrace modalities...
I want to see AMD enter the ARM soc market now that they own Xilinx they can provide competition and choice for phone manufacturers. I am interested in seeing triple A games like Resident Evil 4 running on the 15 pro.
One step closer to the nd of Lagsung 😂
😊that name relatable 😢
until today i never really knew how bad the USB naming actually was!!..
3nm is a big efficiency upgrade we can expect much much better battery than the 14 pro because 14 pro said it had a 4nm chip but it was a tsmc node that was actually 5nm but with a deifferent name
It was only Samsung 5nm++ that they name it 4nm
There are so many people with misinformation.
@paoxiong6146 This is a direct quote from TSMC's website, "In addition, TSMC plans to launch 4nm (N4) technology, an enhanced version of N5 technology." The N4 process is listed under the N5 section of its website.
@@GaryExplains thanks, Gary 🙂 the more you know 😉🤗
I'm not following battery tests anymore but isn't that the iPhone 13 max ultra or whatever is the name has still better battery than any iPhone 14 variant? I have iPhone 14 pro and indeed battery is good, but that comes with some costs, but no one talks about background activities as no one downloads 1k songs on Spotify at once, on Android I could do that watch something on YT or just play game and it carries on, but on iPhone it doesn't work lol I'm disappointed with iPhone since day one they updating it a lot and while they trying to fix some bugs they add new one's, and that lagging keyboard, I installed Google g, and while switching between keyboards I had massive input lagging, before last updated my phone dropped screen refresh rate to less than 30fps with stutter and lag without any reason, latest update fixed it
@@ThereIsOnlyOneRandom bruh can you reply in a way which i can understand isntead of writing a essay ?
I need more processing power to run ML/AI on my phone.
I am disappointed with no of transistors 😂
Can you possibly go deeper into Apple’s Neural Engine ?
Compared to what.
Eh?
I must admit, I was shocked at only a 10% gain. It makes me think we’re dealing with a rebranded a 16 chip.😢
Thx for the report Gary! Do you think Apple will eventually hit with or without adopting latest ARM architecture limitation on optimizing their CPU core stack which will be significant enough over 2 digits? It's not that common with all processors in the industry to see big over 2 digits gen 2 gen increase in performance with more efficient and optimized architecture. Shrinking transistors and increasing their density per mm2 is slowing down so it looks smart and well optimized architecture design will crucial and most important besides advanced chip packaging techniques (3D packaging etc.).
Next year A17 'non pro' for iPhone 16 of course - further price increase expected..
Why not, people still buy iPhone no matter what. In fact 53% US market share is iPhone and growing slowly over the year
What kind if further price increase? Did you aware of that Apple not increased the price since 2017..? Even this $100 on the iPhone 15 Pro Max is not a price increase, they just simply removed the 128GB version, and they replaced it with the 256GB, what was also +$100 even in last year.. (so the iPhone 14 Pro Max 256GB and the iPhone 15 Pro Max 256GB has the same price, technicaly not a price increase..)
@@TamasKiss-yk4st this year you get A16 in iPhone 15, next year you get A17 'amateur' in iPhone 16 => price increased
@@U_H89 did you already forget that when they did this the first time..? Even the regular iPhone 14 got the iPhone 13 Pro chip instead of the regular iPhone 13 chip.. (the Pro chip also had one more GPU core), and since they not increased their price in the past 7 years, why do you expect this price increase just for that chip..?
A17 is not happening. Its expensive N3B node which TSMC want to move away from. A18 on N3E and A18 Pro on N3E next year. Both codenamed Tahiti.
I did not know USB had renamed to the FAR simpler speed-based naming. That is amazing.
I'll give the video 2 Arms Up! Just so long as M3 is using Armv9.
Thanks for that overview. It is cool when looking at it that way
Please do M2 vs A17 Pro chip
Why? One is a phone processor the other a laptop processor.
@@steveseidel9967 It is some I can do, but the effort vs reward means it is often not worth it.
It seems diminishing return!
Video done very well. Thanks
Does the USB controller really take that much space? Wonder how big it will actually be when the die analysis came out
A small ARM core is less than half the size of a piece of rice. A USB controller should be even smaller. I think a tenth of a piece of rice is enough. Apple's USB controller is a cheap one thus it will not be large. USB 3 nowadays are cheap stuff. However the coming USB 5 controller will be a different story.
Is it honestly better than a15 bionic? to be honest I think a15 and newer soc are almost the same in performance and also in real life most won’t see a big difference maybe in graphics but in speed it’s going to be about the same
Does apple take the armv9 on A17 Pro?
Arm announces armv9 way back in 2021.
A15 Bionic is said to come with that but it didn't happened it remains at the same armv8.5-A from the A14 Bionic and A16 Bionic from last year comes with slightly upgraded armv8.6-A.
This is that information what nobody can answer yet. We will see it in a few days.
I was expecting better to be honest
Just because the A17 Pro has jumped up by 3 Billion transistors on a chip, the overall percentage of increase is getting less and less.
Prob 2 of those are just in the GPU core. Same was with A15 cuz 5th core over 4 in A14
ARM v9 has been out for quite some time now. It's surprising that they didn't use it. Though it would probably have been hard to figure out performance improvements even if they did, because Apple heavily customizes the ARM designs.
Apple doesn't customize Arm's designs at all, because it doesn't use Arm's designs, it uses its own.
@@GaryExplains In that case how will they shift to ARM V9? I mean how does ARM version even matter then for Apple CPUs?
@@DK-ox7ze Because Apple' processors need to be 100% compatible and have to pass compatibility tests. At the moment Apple made Armv8.x compatible processors.
@@GaryExplains If they have their own design then why is the compatibility required? Is it because they are using ARM's instruction set? Or they use something else from ARM as well?
It is because they are using Arm's instruction set.
I can't remember when I met a person doing something with an iPhone that couldn't have been done with a phone for 200 bucks.
i’ve met hundreds
What exactly did they do?
Geekbench 7 has entered the chat 😂😂😂😂😂
Wow I remember watching his A chip breakdowns years ago😮
Welcome back. I have made one every year.
Thanks professor. Apple being Apple
Here we goooooo!!! I wait all year for this video
I missed this series damn its been a while for me.
How could they do 5nm to 5nm 3 billion transistors number up??
I don't understand. What does the process node (5nm) have to do with the transistor count?
technically not 4nm on the A16 Bionic, its 5nm process using their 4NP process or something like that, so yeah a marketing term, but I think this is actually now 3nm, right? TSMC themselves quoted an increase of 15% going from 5N to 3N, I think they were somewhere in the middle so 10% kind of makes sense?
Using your analysis, 10% would make sense if this was an A16 made on 3nm, but it isn't, Apple said there were also microarchitecture design changes. I was expecting performance increases with the move to 3 nanometres and performance increases because it is a better CPU. 🤷♂️
@@GaryExplains did they say they improved the cpu cores?
@@MaximoJoshua Yes, I even show the quote in the video.
Its still optical shrink which sits between 5 and 3.
@@steveseidel9967 Apple has bought 100% of their capacity, what other vendors?