What a lot of phone manufacturers keep forgetting is, its not about raw power, its all about heat management. Its all well and good having a blistering fast SOC, but if after 10 mins of use its gets so hot that it throttles itself back so much, it becomes a lot slower. This is thecasewith the A17 pro, beats everyone on benchmarks, but play 10 mins of call of duty and watch the framerates drop like a rock.
That’s not the case at all. Even when it throttles, it’s still faster than previous gen chips. There have been a few videos showing long term heavy gaming with full brightness, and no noticeable frame drops or screen dimming after 30-60 minutes.
@@omCiMaybe not 3 or 4 but definitely 1 or 2. Apple's chips have been lightning fast with efficiency good enough to get great battery with less battery capacity.
@@pangolin83yep and now the a17 is nipping at the heels of the m1 in terms of benchmark results. I could imagine that by the time they get to the a18, and with the usb controllers, it may actually be suitable for a low cost laptop.
Do you happen to know the system level cache for SD in the comparison? MediaTek Dimensity 9200 comes 6MB and Apple also has tons. The MediaTek chip also has a slightly higher memory RAM clock for its LPDDR5X compared to SD. The new Apple is also using a new fabrication process. How much of a difference would these things make in single core or multicore performance? Do you think A710 is dragging down SD? How would you rate Samsung's cooling? Is cooling giving SD an edge? Samsung chip seems to be different from other SDs.
8 gen 2 has 8MB of system level cache. A710 core is not a huge downgrade compared to A715 and it redeems itself with 32 bit support. A715 is 64 bit only. Samsung's cooling is decent, but not the greatest. Also thermal threshold on Samsungs and Pixels are lower than Chinese competitors, so that's that. And the SoC itself in the Samsung phones are simply higher clocked, high binned 8 gen 2s.
They’re all fine. I’m actually more interested in the dedicated parts of the SOC that’s not CPU/GPU. The trend for some years has shown to be about the dedicated/specialized cores on these new SOCs that enable new features and capabilities. ML cores; Ray tracing; encoders; computational photography…
U do realise these are all bs marketing buzzwords right? Ray tracing on phones?? pffff..... Anyone having even the slightest idea of how gpu demanding raytracing is won't evn think twice about having raytracing on a phone in atleast next 5 years or so. Computational photography?? pffff......It's just a fancy way of saying, "we couldn't improve the sensors this year, therefore, we r throwing in some buzzwords so that our customers can have some placebo effects."😅 I could write more, but hope I made my point clear.
@@naqibfarhan4356It’s clearly you’re 5+ years behind or know nothing about what you’re talking about. I’d give you Zero in computer architecture if you were one of my students with that bold naive statement to cover up for not doing your research. Sheesh, getting old and tired of millennials getting their facts out of Gaming forums, Facebook and TikTok instead of scientific books, researches and science advancements.
@@fmax3000 bro don't project yourself onto others. Instead of saying dumb stuff (i.e. I have so much knowledge in terms of technology;source : trust me bro) why not present ur arguments based on actual facts like an actual professional, IF YOU CAN. Yes I double dare you Mr. Teacher. PS: Don't just assume everyone on the internet are milenials and/or they know nothing, and you are superior because you sigh all the time. Get down from your high horses and may be people will stop calling you a boomer.
@@naqibfarhan4356 Sorry, I don’t have time to educate you. Google is your friend. You first need to do your homework. Knowledge doesn’t come without efforts. You should support your claims in the first place when you try to contradict science and everybody else. You don’t even have the slightest knowledge how to argue. That was a feee advice but nobody had the time to teach you how to search, read and learn. Good luck.
There had been pre-release rumors that in addition to better thermals, that the P8Pro also differentiated the Tensor G3 by packaging technology-FoWLP vsersus FoPLP. Any thoughts on whether this rumor ended up being true or are the differences in G3 performance purely attributable to phone size and some copper (instead of graphite) heatsinking.
Great video. Are there standard ways to test the TPUs? Are the benchmarking tools that can test how different architectures differs in handling AI/ML type of performance?
I've came from a samsung qualcomm 8+ gen 1 device and assumed the pixel would perform at least as good. The samsung was essentially flawless in terms of performance but ive been pretty dissappointed with the tensor. Its fine i guess but little stutters throughout the OS. Things like twitter scrolling just have little stutters where the samsung would be butter smooth. Im stuck with it now but have a bit of buyers remorse. Maybe a few software updates will straighten it out.
@@lordvader3697 dunno what to tell you man, using it back to back with my old Samsung and it ain't as smooth. I can scroll through the comments on this thread and it just isn't a perfectly smooth experience. It doesn't track my finger flawlessly. I can only assume it's the chipset as my Samsung with the Qualcomm chip and it's better benchmarks did. I'm not saying it's a terrible experience but it's a newer processor and it's an objectively worse experience.
@@nikc1313 nope just you my friend just you no one else have reported this problem maybe since it just launched some optimisation problem or you got scammed with a faulty device
@@lordvader3697 I suspect it's more just coming from a more powerful device with over a year of software optimisations. I'm probably noticing tiny stutters that the average person wouldn't?
@@nikc1313I think it's app dependent. Reddit always sucks. I've heard Twitter can be bad too. It's a shame the developers on the Android side still haven't fixed this issue for high end phones.
Its a solid chip for what it is. Since the Pixel phones these days occupy mostly budget flugship and flagship space (they are still under 1K$) I think its reasonable to not have the #1 best chip. I don't like their mission statement however. "Traditional Computing plays second fiddle to A.I." is something I don't agree with. Both should get equal attention i.m.o..
It's a bullshit statement. Most tasks you do on your smartphone do not involve AI. AI are just special use cases for idiotic gimmicks like eraser... xD They didn't even gave good use cases to show so they stick to the same stuff as Pixel 7 Pro. They should show me how their AI makes me get more battery life.
I thought the whole point of the Tensor chip was to cut off the middle man (Qualcomm) and pass the saving on to the consumer. With Pixels costing as much as IPhones, why would I opt for inferior hardware for the same cost price?
Because it does a bunch of stuff iPhones can't do? Also, the phones do not cost the same - where are you getting that from? The 15 Pro Max vs. P8P probably differs by what, $200-$300?
The biggest disappointment is the slow UFS 3.1 storage. Google really should've gone with 4.0 and upped the storage speeds. This would make a much bigger difference than the SoC (which i think is plenty good)
Pixel phones are never about hardwares. Also they don't sell that much phones so it will not be as profitable like Samsung to be using the cutting edge technology for them. That's why pixel usually is 1-2 generations behind in terms of hardware. But once they captures enough markets (which will never happen because Google is very very good at keeping promises) they can have enough money to make their own hardware to have cutting edge technology.
@@chairpantsthey are already 2 gens behind the 8 gen 2 and the 8 gen 3 comes out this month. 7 years of updates my ass. It will be irrelevant after 4 years with that anemic SOC. 5 years max
@@MarvoloRiddle 7 years Google update is just plain lying. Google have a tendency to announce big stuffs then wrap it up within 2 years. So this one probably will be something like that too. I have 0 trust in Google claims.
Wildlife on iPhone is lower bcuz it seems like it gets a 60fps cap on the app for some reason, where galaxy is able to go past and push much higher frames resulting in higher scores, that’s why it turns around on the extreme version. I will say I wish Apple would upgrade storage speed and ram speed though, especially with them pushing for AAA console games, it would have been nice to have 5x ram instead of 5, and faster SSD. Maybe with iPhone 16 they will jump straight to ddr6 ram though, at that point even 5x might feel like a let down but if they keep ddr5 yet again it would sting
Impressive to see that Apple is so much ahead of its competition in terms of CPU performance. Although they have improved their GPU this time, they still need some catching up to do if they wish to target AAA gaming in mobile.
The Sony Xperia I V is the fastest Android I own. Extremely snappy and fast. I'd imagine if (Google) used flagship snapdragon cpus in their phones performance would not be an issue. My P6P and P7P have extremely lackluster wireless performance. In areas where it shows no service (TMO) I see UC on my S23 Ultra and get speeds over 100Mbps down and 5 up! THAT is a huge deal. And don't get me started on video. iPhone 13 Pro Max can capture UHD60 with HDR. The 8 Pro still cannot do this. UFS 3.1 instead of 4.0. Going with the flat screen is nice.
@@mistamaog They are completely different. Snapdragon is Qualcomm and Tensor G3 is made by Samsung. This is expected to change in 2025 where TSMC will be making them.
@@cpufrost They both use Arm Cortex X3, A715, and A510 cores for the CPU. Just in different quantities and different clocks, as well as Qualcomm also using A710 cores.
Where is the comparisons of the NPU or tensor sections. Google is touting that part of the chip but yet you show no comparisons. Moreover, even the A17s most impressive leap was with its neural engine being 2x as fast. This breakdown feels incomplete
Unfortunately testing the NPUs is hard as not all of the functionality or features are exposed to the user, they are mainly used by the OS. There are some benchmarks but it is still an immature area.
It also outperforms the A17 in Almost every GFX Bench benchmark including Aztec high tier. I think the wildlife extreme is just a weird outlier that maybe apple optimized for.
Because the first test is shorter, where Samsung still not start to throttling, but Apple is already throttling, but in the Extreme you can see even the throttling iPhone outperform the also throttling S23U after 8 cycles.
Yeah, this is becoming more and more important. The base core's seem to be already fast enough for some years now. It's about algorithms, smart things that makes the software more interesting nowadays. Google is walking the right path as far as I'm concerned.
But that is a huge problem if you planning to buy a Pixel 8 Pro just for the 7 years of updates, and you are lagging after the 3rd uprade, because even the 4 years old competitor phones are performing better..
The Cortex-A715 core, like all of ARM's intermediate cores, just isn't that efficient. ARM needs a far more efficient small core to replace all of its current small and intermediate cores - Apple's efficiency cores are embarrassingly superior (in terms of outright efficiency and in terms of the trade-off between efficiency and performance) to what ARM is offering. Google could have chosen lower clocked X3s, instead of the A715s, which would have provided better support for ML models and, at those lower clocks, could have avoided any heat or battery drain issues. But, Google always tries to save those pennies. I'll wait for the Pixel 9 or 10. By then we should have decent 3nm silicon.
Because, as Google says , the ML hardware mainly runs OS level models, not user level. This is the same for most platforms (Apple and Qualcomm). There are some APIs but the whole area of testing ML tasks is still quite chaotic and not yet stable.
Sorry @steveseidel9967 but you are wrong. I have no ulterior motive for showing one set of results or another. If the iPhone is better, great, I don't have a problem with that. Where I do have a problem is what I said, the features and performance of the NPUs are not fully exposed to user space. I don't mean users as in the consumers can't access features like face recognition that use the NPU, of course that do, that is obvious. I mean that app writers and therefor benchmark writers don't have full access to the NPUs, they are a black box used mainly buy the OS. Geekbench ML relies of the API functionality exposed to app writers and it isn't the same as the internal API used by the OS.
Yes, but you see the difference is that I have actually spoken to the engineers at Qualcomm and at Arm and others and I know that not all the features and functionality of the NPUs are exposed via the APIs (regardless if it is NNAPI, or CoreML etc). What these tests are really testing is how well the NPU handles the respective API and how much of it is optimized specifically for the NPU, and how much is handled by some software layer. While I agree that it is likely that the Hexagon SDK and CoreML are probably better than the more generic ones like NNAPI, there is still a lot of ambiguity. Plus the actual APIs have very different design concepts and so one API might be better in one workload compared to another, which means the API becomes the bottleneck, not the NPU. The fact that you had to list 6 different native approaches is the very proof of what I am saying.
@steveseidel9967 But the bottom line is this. Benchmarking of ML performance is still in its infancy. It is something that needs to be explored more and something that should be included in benchmarking in the future. Geekbench ML is a good starting point. My motivations for not using ML benchmarking at the moment has absolutely nothing to do with who would win or who would lose.
While tensor G3 may have decent performance for most uses, it's lousy for gaming and I'm a hardcore gamer, thus my preference for 8 gen 3 and A17 bionic over tensor G3 is a no brainer.
It looks like the 3d wildlife (non extreme) is not stressing the iPhone correctly thats why there is basically no degradation of performance and the score is so low.
People forget that 95% of the games in mobile land can easily be handled by many phones and 3D Wildlife represents how well the iPhone handles them. The Extreme targets the AAA mobile game performance, the other 5% that maybe many of us do not care about, or just use lower settings.
Tensor’s claims about being more optimised and being faster for AI tasks is COMPLETELY bullshit. One of my friends has a Pixel 7 and I had a Xiaomi 12 pro. I had installed a Pixel ROM and it had all the “Pixel AI” features. Both phones were practically identical in terms of speed including those AI tasks that Google boasts about. In some cases my phone was even faster. So just don’t give me that crap. You may be able to fool the tech illiterate or general public, you can’t fool me.
So why is that? The differences in raw power are not noticeable anymore in normal tasks. The A17 is just about geekbench and not so when comparing other benchmarks. Google is walking the right path, adding companion cores to do smart things with their software. Thumbs up for Google.
I wish you had included A15 and A14 benchmarks just to have a good visual representation of how much Google is concentrating on AI modeling. They can’t possibly add performance and AI modeling at the same time. They could if they wanted to. But they don’t want to. Android for life.
I mean for someone like me that would prefer not to have my data used for AI modelling and sent to the cloud, very little of that matters to me. I want performance vs efficiency, with my ML needs only going to local computational photography.
Apple has the best AI/ML versus the Pixel 8 Pro as well as faster GPU and CPU, so you may want to rethink that. The Tensor 3 is a modern chip on an old node and old ground architecture and that is why the GPU and CPU are years behind the monster A17Pro.
The most important tests are often never done which is signal tests and uptime tests. The Exynos modems for the Pixel 6 and 7 had major network issues. Had a Pixel 6 dozens of times a day no signal at all up to 20 minutes at a time on both ATT and T-Mobile even in populated cities around DFW. Same sim cards in a $50 UMX phone didn't have those issues. Got it RMA'ed same exact issue with a different Pixel 6 device inspected. Others had 3+ RMA's and just ended up giving up selling the phone for a loss or traded it in which is what I had to do too. Google only offers to keep doing RMA's.
Its sad that google can slap a thermometer in pixel 8 pro, which nobody wanted or asked for, but they can not acomodate a SD card slot. These companies stripping away popular features like headphone jack or sd card slot, makes me look toward sony xperia 1 v. If sony promised 5years of software n security, i would buy it at any price at all. Please compare google pixel 8 pro, vs sony experia 1 v, vs samsung s23 ulta, in terms of user experience, UI, in hand feel, features, battery drain time, thermal management, and value for money. I ll buy one of these 3. Waiting for response. Thanx
The main is media engines support, thermals and efficiency that help maximize power while keep being efficient and cool. Lets face it... The power has been plenty powerful the last few gens of chips. Try and get to much power and you end up like apple. Impressive for 10 minutes and then massive battery and heat output which leads to throttling.
@@omCi8 gen 2 has weeker cpu but better gpu than a16 bionic. a17 pro d equal the gpu of 8 gen 2. I think 8 gen 3 will be similar to a16 bionic in cpu and more powerful gpu than 17 pro.
it is what it is. if u looking for better raw performance look elsewhere than the Pixel, if u don't mind the useless benchmark.. rather than focus on the what the phone can do for you.. Pixel is a great device with focus on the AI performance
@@garrusvakarian8709 in no way does Pixel do it more? so what kind of AI that iphone offers as features ? inferior Siri & speech recognition? is there call screening? what AI features in camera?
tensor is a joke a.i is all about processing power they assume their customers are morons if samsung or apple choose to go a.i they will blow google out of the water instantly me i loathe a.i so i hope they don't
Even if Google's tensor g3 is superior to their first 2 tensor chips, the tensor 3 will still be easy prey to the far superior 8 gen 3, dimensity 9300 and Apple's A17 bionic which is the only 3nm process on the market.
Let's be honest the gen 3 and a17 are overkill. Even the gen 2 is. Tensor 3 is still a very good proccesor if you look at how far mobile chips have come. I have a gen 1 in my samsung tab s8 and that thing is a beast!
Just as Dodge challenger hellcat has proven horsepower overkill is a good thing, the samething applies to 8 gen 3 and A17 bionic being overkill is a good thing
@user-mn7yz8jj5s I'm not saying it's a bad thing. More power is always better but the tensor g3 is still plenty of horsepower for 99% of people . Even the s23 fe isn't a bad choice in 2023 and the tensor g3 is slightly better than the gen 1. I have all 3
From india i love this type of deep and technical info my request is please help me i am looking for smartphone my budget is (30,000/- indian rupees ) Not for gaming but for long term use like 4-5 years overall balance and efficient soc non heating i am professional photographer i love pixel in in android but pixel good at camera but lack in other departments like heating so no overall balance phone
I'm more concerned on the efficiency of the chipset rather than the raw performance. Nowadays, most of the devices could run Android smoothly (except the OneUI lag)
Looks like chip technology being used today has again reached point of diminishing returns with performance battery heat issues in such compact smartphone package, I have noticed bench mark tests don’t show true real world daily use experience, Apple devices still perform more consistently for daily use than Samsung & Google especially over years of use, don’t get me wrong they are getting there but they also have to compete with the Apple ecosystem, 👍👍video
I agree. With no active cooling, we cannot expect too much more as the heat saturation will be faster due to initial higher performance. The future is AI and ML along with tricks like frame interpolation.
@@dansourile5444 not saying other devices stop working or are bad they just don’t work as consistently reliable as apples ecosystem of devices, not just looking at one device from 1 person perspective but look at each companies lineup and their devices reliability working together overall for 5-7yrs
Honestly it’s all about profit making. Snapdragon chips are way superior in every aspects from this thing still they going with Samsung. Lost interest after seeing Samsung 4nm written. The fabrication which destroyed 2 generations of Snapdragon reputation.
So far Pixel is focused on their AI performance. Indeed Pixel is doing amazing in this area and their AI tools and features are exceptional to say the least. But in 2025 we will see a custom CPU design with in-house cores and TSMC manufacturing process. This should greatly increase the performance and efficiency. 2 generations from now.
That's what they market. Reality is that there isn't a single AI task that the Tensor 3 could do that other processors can't do at the same level. The only limitation is that Google doesnt provide the software.
@@robertmusil1107 Google focused greatly on AI in the recent years. They are aiming to become the best in the industry. Indeed I won't be surprised if A17 Pro has more powerful AI NPU than the Tensor. Both companies are using AI in different areas for the most part. It's just that Google is more focused on the volume of AI tools and features
@@ulikemyname6744how has Apple used that AI power? I've never seen anything AI related with Apple before. They always maximize the performance but limit your abilities (like the lack of true multitasking).
@@dansourile5444 Apple uses AI from basic stuff like understanding what material is transparent so it can when you put your wallpapers to using DLSS-like features in gaming. Apple is also extensively using AI for some camera workloads and for the FaceID system. These and more are some of the areas where Apple uses AI.
It’s kinda sad that even the Apple 15 Pros beat all the competitors in AI performance. Geekbench ML has a variety of AI & ML related benchmark tests and Apple consistently out scores the competition
@@GaryExplains i dont use number's. Just run more than two apps on iphone and it starts struggling whereas snapdragon keeps going well even with for apps. I don't trust geekbench anymore since apple cpus are designed to just perfectly fit on the tasks of the benchmark.
I am not sure what you mean by the iPhone starts to struggle with two apps. Could you describe that in detail. I have never seen that myself. Also, did you watch my interview with the boss of Geekbench?
apple has lost its lead against samsung now apple mustn't be so complacent then samsung will run over apple even more if apple wants they can change this but they have become too self-righteous wonder how much better the S24 is when samsung releases this then apple A17 pro will be even more behind
Yes, the A17 uses the next generation 3nm SoC as first company. Qualcomm is still using 4nm but in a few months other companies will also. So, then comparing makes more sense.
There are some reviews that show that power being used and it is much faster than the best of Android. Of course, you are talking about split-screen and stuff like that which 99% of the world doesn't care about!
People who like clean and smooth UI on Android? For good camera? For best in class AI that makes life easier? Not everyone need maxxed out raw performance.
@@Sarthak_Joshi , huh,..I kind of agree. Samsung just has too much bloatware. But Apple has both good ui and camera too but closed garden. I despise AI with a passion so that disqualifies me. Thanks!
Really great testing!!! Love this breakdown. Puts things in perspective for sure
What a lot of phone manufacturers keep forgetting is, its not about raw power, its all about heat management. Its all well and good having a blistering fast SOC, but if after 10 mins of use its gets so hot that it throttles itself back so much, it becomes a lot slower. This is thecasewith the A17 pro, beats everyone on benchmarks, but play 10 mins of call of duty and watch the framerates drop like a rock.
That’s not the case at all. Even when it throttles, it’s still faster than previous gen chips. There have been a few videos showing long term heavy gaming with full brightness, and no noticeable frame drops or screen dimming after 30-60 minutes.
lol.
The lengths of twisting yourself into pretzel you will go to explain why Apple is 3 maybe 4 generations ahead.
@@omCiMaybe not 3 or 4 but definitely 1 or 2. Apple's chips have been lightning fast with efficiency good enough to get great battery with less battery capacity.
@@pangolin83yep and now the a17 is nipping at the heels of the m1 in terms of benchmark results. I could imagine that by the time they get to the a18, and with the usb controllers, it may actually be suitable for a low cost laptop.
To be fair why people forget the Mediatek Dimensity9200/9200+ even it was the only sole rival of Qualcomm after Exynos ditched the Exynos2300
Honestly performed a lot better than I expected
Do you happen to know the system level cache for SD in the comparison? MediaTek Dimensity 9200 comes 6MB and Apple also has tons. The MediaTek chip also has a slightly higher memory RAM clock for its LPDDR5X compared to SD. The new Apple is also using a new fabrication process.
How much of a difference would these things make in single core or multicore performance?
Do you think A710 is dragging down SD?
How would you rate Samsung's cooling? Is cooling giving SD an edge? Samsung chip seems to be different from other SDs.
8 gen 2 has 8MB of system level cache.
A710 core is not a huge downgrade compared to A715 and it redeems itself with 32 bit support. A715 is 64 bit only.
Samsung's cooling is decent, but not the greatest. Also thermal threshold on Samsungs and Pixels are lower than Chinese competitors, so that's that.
And the SoC itself in the Samsung phones are simply higher clocked, high binned 8 gen 2s.
They’re all fine. I’m actually more interested in the dedicated parts of the SOC that’s not CPU/GPU. The trend for some years has shown to be about the dedicated/specialized cores on these new SOCs that enable new features and capabilities. ML cores; Ray tracing; encoders; computational photography…
Exactly. For speed and efficiency, specialized units are the only way forward. GPUs are already along that line.
U do realise these are all bs marketing buzzwords right?
Ray tracing on phones?? pffff..... Anyone having even the slightest idea of how gpu demanding raytracing is won't evn think twice about having raytracing on a phone in atleast next 5 years or so.
Computational photography?? pffff......It's just a fancy way of saying, "we couldn't improve the sensors this year, therefore, we r throwing in some buzzwords so that our customers can have some placebo effects."😅
I could write more, but hope I made my point clear.
@@naqibfarhan4356It’s clearly you’re 5+ years behind or know nothing about what you’re talking about. I’d give you Zero in computer architecture if you were one of my students with that bold naive statement to cover up for not doing your research. Sheesh, getting old and tired of millennials getting their facts out of Gaming forums, Facebook and TikTok instead of scientific books, researches and science advancements.
@@fmax3000 bro don't project yourself onto others.
Instead of saying dumb stuff (i.e. I have so much knowledge in terms of technology;source : trust me bro) why not present ur arguments based on actual facts like an actual professional, IF YOU CAN.
Yes I double dare you Mr. Teacher.
PS: Don't just assume everyone on the internet are milenials and/or they know nothing, and you are superior because you sigh all the time. Get down from your high horses and may be people will stop calling you a boomer.
@@naqibfarhan4356 Sorry, I don’t have time to educate you. Google is your friend. You first need to do your homework. Knowledge doesn’t come without efforts. You should support your claims in the first place when you try to contradict science and everybody else. You don’t even have the slightest knowledge how to argue. That was a feee advice but nobody had the time to teach you how to search, read and learn. Good luck.
There had been pre-release rumors that in addition to better thermals, that the P8Pro also differentiated the Tensor G3 by packaging technology-FoWLP vsersus FoPLP. Any thoughts on whether this rumor ended up being true or are the differences in G3 performance purely attributable to phone size and some copper (instead of graphite) heatsinking.
can you explain the technology difference? maybe more physical distance between cache and cpu cores?
@@hanspeter24no
@@Xevos701 then what can you explain or no knowledge?
Man.. the vs in the thumbnail have me some hope that speedtest g is back.
Great video. Are there standard ways to test the TPUs? Are the benchmarking tools that can test how different architectures differs in handling AI/ML type of performance?
Thank you For The Informative Graphic Analysis Between CPU' and GPU' Arrangements. Crucial and most Appreciated. Important
Interesting comparison, thanks for sharing!
Is there a modem upgrade?
(as tensor chipsets are hot potatos when it comes to mobile internet.)
The real world performance is fine. Now if they could just improve efficiency so that battery life doesn't suck.
I’m really looking forward to the transition to tsmc 3nm tensor g5
I've came from a samsung qualcomm 8+ gen 1 device and assumed the pixel would perform at least as good. The samsung was essentially flawless in terms of performance but ive been pretty dissappointed with the tensor. Its fine i guess but little stutters throughout the OS. Things like twitter scrolling just have little stutters where the samsung would be butter smooth. Im stuck with it now but have a bit of buyers remorse. Maybe a few software updates will straighten it out.
Stutter even my pixel 6 doesnt stutter
@@lordvader3697 dunno what to tell you man, using it back to back with my old Samsung and it ain't as smooth. I can scroll through the comments on this thread and it just isn't a perfectly smooth experience. It doesn't track my finger flawlessly. I can only assume it's the chipset as my Samsung with the Qualcomm chip and it's better benchmarks did. I'm not saying it's a terrible experience but it's a newer processor and it's an objectively worse experience.
@@nikc1313 nope just you my friend just you no one else have reported this problem maybe since it just launched some optimisation problem or you got scammed with a faulty device
@@lordvader3697 I suspect it's more just coming from a more powerful device with over a year of software optimisations. I'm probably noticing tiny stutters that the average person wouldn't?
@@nikc1313I think it's app dependent. Reddit always sucks. I've heard Twitter can be bad too. It's a shame the developers on the Android side still haven't fixed this issue for high end phones.
Its a solid chip for what it is. Since the Pixel phones these days occupy mostly budget flugship and flagship space (they are still under 1K$) I think its reasonable to not have the #1 best chip.
I don't like their mission statement however. "Traditional Computing plays second fiddle to A.I." is something I don't agree with. Both should get equal attention i.m.o..
$1700 in Australia for a 128GB Pixel 8 Pro
It's a bullshit statement. Most tasks you do on your smartphone do not involve AI. AI are just special use cases for idiotic gimmicks like eraser... xD They didn't even gave good use cases to show so they stick to the same stuff as Pixel 7 Pro. They should show me how their AI makes me get more battery life.
I thought the whole point of the Tensor chip was to cut off the middle man (Qualcomm) and pass the saving on to the consumer. With Pixels costing as much as IPhones, why would I opt for inferior hardware for the same cost price?
Why do you think the Tensor chip is about cost savings? Designing and manufacturing an SoC is a multi million dollar task.
Because it does a bunch of stuff iPhones can't do?
Also, the phones do not cost the same - where are you getting that from? The 15 Pro Max vs. P8P probably differs by what, $200-$300?
The biggest disappointment is the slow UFS 3.1 storage. Google really should've gone with 4.0 and upped the storage speeds. This would make a much bigger difference than the SoC (which i think is plenty good)
Pixel phones are never about hardwares. Also they don't sell that much phones so it will not be as profitable like Samsung to be using the cutting edge technology for them. That's why pixel usually is 1-2 generations behind in terms of hardware.
But once they captures enough markets (which will never happen because Google is very very good at keeping promises) they can have enough money to make their own hardware to have cutting edge technology.
@@chairpantsthey are already 2 gens behind the 8 gen 2 and the 8 gen 3 comes out this month. 7 years of updates my ass. It will be irrelevant after 4 years with that anemic SOC. 5 years max
@@MarvoloRiddle 7 years Google update is just plain lying. Google have a tendency to announce big stuffs then wrap it up within 2 years. So this one probably will be something like that too. I have 0 trust in Google claims.
@@MarvoloRiddleim still using a oneplus 7 pro from 2018 with a sd 855 and it's fast enough even for games
@@niveZz- when your Oneplus was released was the SOC 2 generations old? If it were it would feel like 9 years old after 7 years. Get my point?
Wildlife on iPhone is lower bcuz it seems like it gets a 60fps cap on the app for some reason, where galaxy is able to go past and push much higher frames resulting in higher scores, that’s why it turns around on the extreme version. I will say I wish Apple would upgrade storage speed and ram speed though, especially with them pushing for AAA console games, it would have been nice to have 5x ram instead of 5, and faster SSD. Maybe with iPhone 16 they will jump straight to ddr6 ram though, at that point even 5x might feel like a let down but if they keep ddr5 yet again it would sting
Impressive to see that Apple is so much ahead of its competition in terms of CPU performance. Although they have improved their GPU this time, they still need some catching up to do if they wish to target AAA gaming in mobile.
The Sony Xperia I V is the fastest Android I own. Extremely snappy and fast. I'd imagine if (Google) used flagship snapdragon cpus in their phones performance would not be an issue. My P6P and P7P have extremely lackluster wireless performance. In areas where it shows no service (TMO) I see UC on my S23 Ultra and get speeds over 100Mbps down and 5 up! THAT is a huge deal.
And don't get me started on video. iPhone 13 Pro Max can capture UHD60 with HDR. The 8 Pro still cannot do this. UFS 3.1 instead of 4.0.
Going with the flat screen is nice.
The CPU structure isn't much different from the Tensor G3 and the 8 Gen 2, they're both using the latest ARM CPUs
@@mistamaog They are completely different. Snapdragon is Qualcomm and Tensor G3 is made by Samsung. This is expected to change in 2025 where TSMC will be making them.
@@cpufrost They both use Arm Cortex X3, A715, and A510 cores for the CPU. Just in different quantities and different clocks, as well as Qualcomm also using A710 cores.
Why is Immortalis on the 9200 not present on the wildlife extreme test bar graph?
Oh so MP is the amount of graphics processors, kinda like AMD using CUs (compute units) for their GPUs?
Where is the comparisons of the NPU or tensor sections. Google is touting that part of the chip but yet you show no comparisons. Moreover, even the A17s most impressive leap was with its neural engine being 2x as fast. This breakdown feels incomplete
Unfortunately testing the NPUs is hard as not all of the functionality or features are exposed to the user, they are mainly used by the OS. There are some benchmarks but it is still an immature area.
haha, 'the last version of ARM CPUs' for the next 30 days!
Gary the legend 🤝
not the first but I see the 1 view counter still, after finisching the vid.
Very cool, interessting for a P6 user ;)
Thanks
Thank you
Is this speed testing coming back for major competitors🥺
why does the 8 gen 2 outperform the A17 in Wild Life test, but not in the Extreme version?
It also outperforms the A17 in Almost every GFX Bench benchmark including Aztec high tier. I think the wildlife extreme is just a weird outlier that maybe apple optimized for.
Because the first test is shorter, where Samsung still not start to throttling, but Apple is already throttling, but in the Extreme you can see even the throttling iPhone outperform the also throttling S23U after 8 cycles.
However, what will you measure TPU against ? 🤔
Yeah, this is becoming more and more important. The base core's seem to be already fast enough for some years now. It's about algorithms, smart things that makes the software more interesting nowadays. Google is walking the right path as far as I'm concerned.
The tensor cpu was never bad on paper. It just sucks at benchmarks and gaming.
If you are a performance-chasing smartphone user then the Pixel 8 Pro is not a phone that should even be on your shortlist.
But that is a huge problem if you planning to buy a Pixel 8 Pro just for the 7 years of updates, and you are lagging after the 3rd uprade, because even the 4 years old competitor phones are performing better..
It's time that artificial benchmarks start doing artificial intelligence core benchmarking.
why there's no speed test videos !!!!!
how to measure TPU performance improvement?
It isn't clear the best way to do that as the NPUs run mainly OS level things. This is true for most platforms (Apple, Qualcomm, etc).
The Cortex-A715 core, like all of ARM's intermediate cores, just isn't that efficient. ARM needs a far more efficient small core to replace all of its current small and intermediate cores - Apple's efficiency cores are embarrassingly superior (in terms of outright efficiency and in terms of the trade-off between efficiency and performance) to what ARM is offering.
Google could have chosen lower clocked X3s, instead of the A715s, which would have provided better support for ML models and, at those lower clocks, could have avoided any heat or battery drain issues. But, Google always tries to save those pennies.
I'll wait for the Pixel 9 or 10. By then we should have decent 3nm silicon.
Why there arent any ml tests?
Because, as Google says , the ML hardware mainly runs OS level models, not user level. This is the same for most platforms (Apple and Qualcomm). There are some APIs but the whole area of testing ML tasks is still quite chaotic and not yet stable.
Sorry @steveseidel9967 but you are wrong. I have no ulterior motive for showing one set of results or another. If the iPhone is better, great, I don't have a problem with that. Where I do have a problem is what I said, the features and performance of the NPUs are not fully exposed to user space. I don't mean users as in the consumers can't access features like face recognition that use the NPU, of course that do, that is obvious. I mean that app writers and therefor benchmark writers don't have full access to the NPUs, they are a black box used mainly buy the OS. Geekbench ML relies of the API functionality exposed to app writers and it isn't the same as the internal API used by the OS.
Yes, but you see the difference is that I have actually spoken to the engineers at Qualcomm and at Arm and others and I know that not all the features and functionality of the NPUs are exposed via the APIs (regardless if it is NNAPI, or CoreML etc). What these tests are really testing is how well the NPU handles the respective API and how much of it is optimized specifically for the NPU, and how much is handled by some software layer. While I agree that it is likely that the Hexagon SDK and CoreML are probably better than the more generic ones like NNAPI, there is still a lot of ambiguity. Plus the actual APIs have very different design concepts and so one API might be better in one workload compared to another, which means the API becomes the bottleneck, not the NPU. The fact that you had to list 6 different native approaches is the very proof of what I am saying.
@steveseidel9967 But the bottom line is this. Benchmarking of ML performance is still in its infancy. It is something that needs to be explored more and something that should be included in benchmarking in the future. Geekbench ML is a good starting point. My motivations for not using ML benchmarking at the moment has absolutely nothing to do with who would win or who would lose.
Please add 4k video edit output time!
That isn't really a realistic test, it just tests the speed of the video encoder hardware and not much else.
What about Mediatek 9200+ ?
Missed the battery life test...
While tensor G3 may have decent performance for most uses, it's lousy for gaming and I'm a hardcore gamer, thus my preference for 8 gen 3 and A17 bionic over tensor G3 is a no brainer.
It looks like the 3d wildlife (non extreme) is not stressing the iPhone correctly thats why there is basically no degradation of performance and the score is so low.
People forget that 95% of the games in mobile land can easily be handled by many phones and 3D Wildlife represents how well the iPhone handles them. The Extreme targets the AAA mobile game performance, the other 5% that maybe many of us do not care about, or just use lower settings.
Tensor’s claims about being more optimised and being faster for AI tasks is COMPLETELY bullshit. One of my friends has a Pixel 7 and I had a Xiaomi 12 pro. I had installed a Pixel ROM and it had all the “Pixel AI” features. Both phones were practically identical in terms of speed including those AI tasks that Google boasts about. In some cases my phone was even faster. So just don’t give me that crap. You may be able to fool the tech illiterate or general public, you can’t fool me.
How do you know that your phone was doing the processing locally and not using the cloud? Did you try all the AI related features when in flight mode?
Wow! The A17 Pro makes the Tensor G3 look laughable. Even the Dimensity 9200 was better. 🤭😆
So why is that? The differences in raw power are not noticeable anymore in normal tasks. The A17 is just about geekbench and not so when comparing other benchmarks.
Google is walking the right path, adding companion cores to do smart things with their software. Thumbs up for Google.
@@hermanstokbrood 😿😆 You crying isn't gonna change the fact that Google's Tensor G3 even got shamed by the 4-year-old iPhone 12's A14 Bionic 🤭🤣
I wish you had included A15 and A14 benchmarks just to have a good visual representation of how much Google is concentrating on AI modeling. They can’t possibly add performance and AI modeling at the same time. They could if they wanted to. But they don’t want to. Android for life.
I mean for someone like me that would prefer not to have my data used for AI modelling and sent to the cloud, very little of that matters to me. I want performance vs efficiency, with my ML needs only going to local computational photography.
But yet the iPhones still FAR outscore the competition in AI and ML related benchmarks on Geekbench ML. Lol
Apple has the best AI/ML versus the Pixel 8 Pro as well as faster GPU and CPU, so you may want to rethink that. The Tensor 3 is a modern chip on an old node and old ground architecture and that is why the GPU and CPU are years behind the monster A17Pro.
@@andyH_Englandyes and they use all that power for... Absolutely nothing. Still a waste.
Exynos 2400 ?
is it glorified exynos 2400? i wonder
or dimensity 9200+ glorified?
Glorified Exynos 2200 it looks like.
More like a gimpwd Exynos 2300
The most important tests are often never done which is signal tests and uptime tests. The Exynos modems for the Pixel 6 and 7 had major network issues. Had a Pixel 6 dozens of times a day no signal at all up to 20 minutes at a time on both ATT and T-Mobile even in populated cities around DFW. Same sim cards in a $50 UMX phone didn't have those issues. Got it RMA'ed same exact issue with a different Pixel 6 device inspected. Others had 3+ RMA's and just ended up giving up selling the phone for a loss or traded it in which is what I had to do too. Google only offers to keep doing RMA's.
Its sad that google can slap a thermometer in pixel 8 pro, which nobody wanted or asked for, but they can not acomodate a SD card slot. These companies stripping away popular features like headphone jack or sd card slot, makes me look toward sony xperia 1 v. If sony promised 5years of software n security, i would buy it at any price at all.
Please compare google pixel 8 pro, vs sony experia 1 v, vs samsung s23 ulta, in terms of user experience, UI, in hand feel, features, battery drain time, thermal management, and value for money. I ll buy one of these 3.
Waiting for response. Thanx
So now they're finally going to odd numbers
I wonder how the upcoming SnapDragon 8 Gen 3 is going to compare to those CPUs, especially to the A17 Pro.
Let’s see if it can get to A16 first. I also wonder how the upcoming A18 will compare to SD8 gen3 or gen4.
The main is media engines support, thermals and efficiency that help maximize power while keep being efficient and cool.
Lets face it... The power has been plenty powerful the last few gens of chips.
Try and get to much power and you end up like apple. Impressive for 10 minutes and then massive battery and heat output which leads to throttling.
I think it’ll be closer but it won’t beat Apple in either single or multi core performance. They just aren’t able.
@@omCi8 gen 2 has weeker cpu but better gpu than a16 bionic. a17 pro d equal the gpu of 8 gen 2. I think 8 gen 3 will be similar to a16 bionic in cpu and more powerful gpu than 17 pro.
@@soorajj9207No
8 gen 2 has better gpu than a17 pro. Tested and confirmed by many. A17 pro sucks big time in gpu
it is what it is. if u looking for better raw performance look elsewhere than the Pixel, if u don't mind the useless benchmark.. rather than focus on the what the phone can do for you.. Pixel is a great device with focus on the AI performance
@@garrusvakarian8709 in no way does Pixel do it more? so what kind of AI that iphone offers as features ? inferior Siri & speech recognition? is there call screening? what AI features in camera?
what about speed test g video huh
tensor is a joke a.i is all about processing power they assume their customers are morons if samsung or apple choose to go a.i they will blow google out of the water instantly me i loathe a.i so i hope they don't
Well to point to that. Under snapdragon we got unlimited photo storage. With tensor that's gone. So I keep my 4a 😂.
Even if Google's tensor g3 is superior to their first 2 tensor chips, the tensor 3 will still be easy prey to the far superior 8 gen 3, dimensity 9300 and Apple's A17 bionic which is the only 3nm process on the market.
Let's be honest the gen 3 and a17 are overkill. Even the gen 2 is. Tensor 3 is still a very good proccesor if you look at how far mobile chips have come. I have a gen 1 in my samsung tab s8 and that thing is a beast!
Just as Dodge challenger hellcat has proven horsepower overkill is a good thing, the samething applies to 8 gen 3 and A17 bionic being overkill is a good thing
@user-mn7yz8jj5s I'm not saying it's a bad thing. More power is always better but the tensor g3 is still plenty of horsepower for 99% of people . Even the s23 fe isn't a bad choice in 2023 and the tensor g3 is slightly better than the gen 1. I have all 3
From india i love this type of deep and technical info my request is please help me i am looking for smartphone my budget is (30,000/- indian rupees )
Not for gaming but for long term use like 4-5 years overall balance and efficient soc non heating i am professional photographer i love pixel in in android but pixel good at camera but lack in other departments like heating so no overall balance phone
This is cool more competition more products with more power . But apple stills winning . Give more google . Your are realy good Samsung
Wow, you make me feel less bad about paying loads more ^^"
I'm more concerned on the efficiency of the chipset rather than the raw performance. Nowadays, most of the devices could run Android smoothly (except the OneUI lag)
Looks like chip technology being used today has again reached point of diminishing returns with performance battery heat issues in such compact smartphone package, I have noticed bench mark tests don’t show true real world daily use experience, Apple devices still perform more consistently for daily use than Samsung & Google especially over years of use, don’t get me wrong they are getting there but they also have to compete with the Apple ecosystem, 👍👍video
I agree. With no active cooling, we cannot expect too much more as the heat saturation will be faster due to initial higher performance. The future is AI and ML along with tricks like frame interpolation.
This hasn't been true in years, regarding Samsung phones. The Note 20 Ultra is still a monster with no issues, and that's 3 years old.
@@dansourile5444 not saying other devices stop working or are bad they just don’t work as consistently reliable as apples ecosystem of devices, not just looking at one device from 1 person perspective but look at each companies lineup and their devices reliability working together overall for 5-7yrs
Honestly it’s all about profit making. Snapdragon chips are way superior in every aspects from this thing still they going with Samsung. Lost interest after seeing Samsung 4nm written. The fabrication which destroyed 2 generations of Snapdragon reputation.
Lets be honest who even has the time to play high demanding games on their phones that often??
And even if there are people; I'm sure they will have a descent game smartphone like the Asus ROG Phone 7.
It's not about speed performance battery or heat..... Ok Google lol 😂😂😂
More like "next gen CPU throttling"
Need speed test G back!
Looks like a huge performance increase from the pixel 6 Pro to the pixel 8 Pro.
Snapdragon 7+ gen 2 laughing over it's value for money advantage
So far Pixel is focused on their AI performance. Indeed Pixel is doing amazing in this area and their AI tools and features are exceptional to say the least. But in 2025 we will see a custom CPU design with in-house cores and TSMC manufacturing process. This should greatly increase the performance and efficiency. 2 generations from now.
That's what they market. Reality is that there isn't a single AI task that the Tensor 3 could do that other processors can't do at the same level. The only limitation is that Google doesnt provide the software.
@@robertmusil1107 Google focused greatly on AI in the recent years. They are aiming to become the best in the industry. Indeed I won't be surprised if A17 Pro has more powerful AI NPU than the Tensor. Both companies are using AI in different areas for the most part. It's just that Google is more focused on the volume of AI tools and features
@@ulikemyname6744how has Apple used that AI power? I've never seen anything AI related with Apple before. They always maximize the performance but limit your abilities (like the lack of true multitasking).
@@dansourile5444 Apple uses AI from basic stuff like understanding what material is transparent so it can when you put your wallpapers to using DLSS-like features in gaming. Apple is also extensively using AI for some camera workloads and for the FaceID system. These and more are some of the areas where Apple uses AI.
How about comparing AI power?
It’s kinda sad that even the Apple 15 Pros beat all the competitors in AI performance. Geekbench ML has a variety of AI & ML related benchmark tests and Apple consistently out scores the competition
very bad comparison. in fact sd8gen2 is far faster than a17 pro.
Really? How are you measuring that? What numbers do you have to show that?
@@GaryExplains i dont use number's. Just run more than two apps on iphone and it starts struggling whereas snapdragon keeps going well even with for apps.
I don't trust geekbench anymore since apple cpus are designed to just perfectly fit on the tasks of the benchmark.
I am not sure what you mean by the iPhone starts to struggle with two apps. Could you describe that in detail. I have never seen that myself. Also, did you watch my interview with the boss of Geekbench?
Interesting, hopefully as for decades apple will go ahead
No
apple has lost its lead against samsung now apple mustn't be so complacent then samsung will run over apple even more
if apple wants they can change this but they have become too self-righteous
wonder how much better the S24 is when samsung releases this then apple A17 pro will be even more behind
Yes, the A17 uses the next generation 3nm SoC as first company. Qualcomm is still using 4nm but in a few months other companies will also. So, then comparing makes more sense.
Apple is very funny, they have the power of the universe yet they can't even use it at full potential because of the software they use lmao
lol is that what you tell yourself to sleep at night? 🤦♂️. Keep finding explanations why you are using inferior hardware and software.
There are some reviews that show that power being used and it is much faster than the best of Android. Of course, you are talking about split-screen and stuff like that which 99% of the world doesn't care about!
Um remind me,...why should anyone buy a Google phone?
People who like clean and smooth UI on Android? For good camera? For best in class AI that makes life easier?
Not everyone need maxxed out raw performance.
@@Sarthak_Joshiyeah ok. lol. 😂
@@omCihe bodied that dude and his points are valid
@@Sarthak_Joshi , huh,..I kind of agree. Samsung just has too much bloatware. But Apple has both good ui and camera too but closed garden. I despise AI with a passion so that disqualifies me. Thanks!
@@Sarthak_Joshi In fact few people need maxed out raw performance.