Google Tensor G4 fails to outscore Snapdragon 8 Gen 1 and barely beats Tensor G3 in 3DMark Wild Life Extreme
Posted by Qsand0@reddit | hardware | View on Reddit | 121 comments
ComeGateMeBro@reddit
How long before google decides this isn't for them and it goes to the google grave?
chronocapybara@reddit
And yet, the Snapdragon and Apple chips actually perform much better on AI benchmarks as well.
Not-the-best-name@reddit
Lol. Who gives a shit about AI performance?
FinBenton@reddit
I mean it's getting more and more important every day.
BWCDD4@reddit
Is it? I mean I work in IT as a Network engineer, never used it or plan to, has no use or practicality for me.
PunjabKLs@reddit
Online AI/ML "experts" are the C or even D team of folks who didn't catch the first wave of tech dev with silicon valley.
It's obvious because they are the only one who can't tell that the service has just put lipstick on a pig and is trying to sell it as something else.
Anyone skeptical on AI speaks from using these LLMs enough to understand they have hit a wall. I'll wait for the next nerd to tell me wait for GPT-5 bro or wait till we have more compute bro
MarcusOrlyius@reddit
Pretending that AI is useless and has no future is simply delusional and ignores all the evidence of many people using it to accomlish all kinds of goals and ignores all the money that has been invested in it.
Why? You'll just say the exact same thing again as it's not based on facts but your feelings.
Look at this.
https://gemini.google.com/app/d6424832f48b716c
If I wanted those pictures 10 years ago, I would have had to find, hire and pay an artist to produce them. Today I can have AI produce them in seconds.
How about stuff like maths and science?
"I gave the new model a challenging complex analysis problem (which I had previously asked GPT4 to assist in writing up a proof of in https://chatgpt.com/share/63c5774a-d58a-47c2-9149-362b05e268b4 ). Here the results were better than previous models, but still slightly disappointing: the new model could work its way to a correct (and well-written) solution if provided a lot of hints and prodding, but did not generate the key conceptual ideas on its own, and did make some non-trivial mistakes. The experience seemed roughly on par with trying to advise a mediocre, but not completely incompetent, graduate student. However, this was an improvement over previous models, whose capability was closer to an actually incompetent graduate student. It may only take one or two further iterations of improved capability (and integration with other tools, such as computer algebra packages and proof assistants) until the level of "competent graduate student" is reached, at which point I could see this tool being of significant use in research level tasks."
https://mathstodon.xyz/@tao/113132502735585408
Strazdas1@reddit
AI uis se very useful and has a future. These LLM chatbots on steroids though? not very much.
MarcusOrlyius@reddit
They're very useful for various tasks as demonstarted in my previous post. I couldn't of drew those pictures myself so I would have had to pay someone else to draw them for me.
MarcusOrlyius@reddit
If you worked for me, I'd be looking to replace you as soon as possible if I heard you make such an ignorant and short sighted comment.
Not-the-best-name@reddit
Have you tried using ChatGPT to configure NGINX recently?
94746382926@reddit
If you've never used it how would you know that it doesn't offer you a practical use case?
dnkndnts@reddit
This logic applies to literally everything.
How do you know buttplugs have no application for you, Mr Netadmin, hmm? Have you ever tried them? Many users say it increases comfort and reduces job stress, how is that not applicable to network administration?
94746382926@reddit
Yeah I suppose you're right, fair enough!
Gao_Dan@reddit
I have used buttplugs and can wholeheartly say that it improved my comfort and enjoyment at work, recommended.
namemcname02@reddit
I heard it also helps while playing in chess tournaments
FinBenton@reddit
If you are using your phone you are using ai even if you don't know about it.
WingedGundark@reddit
I’m with you on this. I hope they find some other practical use for these matrix calculations before the AI fad runs out or I really can’t get excited about these. There are of course practical scientific and engineering uses for them, but you probably don’t want to run those in your mobile device in any case.
benjiro3000@reddit
I fear you will be wrong on this part, because its not going to go away.
To be honest, i was like you also. Its the like steps of denial
The same also with a lot of the code development. You need to learn where to flaws are, where it get stuck, how to formulate sentences. But it can really help to quickly prototype functions without you spending a ton of time in the documentation, or finding libs that you never find with googling (what has become frankly a lot worse).
In the end, it saves me a ton of time. GPT 3.5/4 Models are darn good IF you know how to deal with it. And you NEED knowledge of the languages you work in, to spot the issues that it sometimes (often) produces. Especially on more complex code.
Again i am on the fence for mobile devices? What is the use of AI on mobile devices, ... But then i think back to the same issue i had for ChatGPT and CoPilot as development tools. And its more of a "good integration + people learning how to use it, will be the motivator for people not wanting to be without AI assistance".
Let be honest, Mobile devices interfaces kind of suck for complex things, with limited screen sizes. Well, this is a potential avenue for such things. How about local translation models that are more accurate/make less mistakes. How about local sentence improvement without you sending your data to whatever megacorps for their data collection (they are still going to collect it, be sure of it!).
Ironically, this is one field where Linux / privacy focused OS may benefit more of faster local AI. But most people do not care anyway, so ...
The thing is, i often see people thinking too small. Whenever a new tech becomes available, everybody is testing the ground what sticks and what does not. When the first cars came out, i am sure plenty of people looked down on them because horses are better. Try finding that same argument 20 years later. Things need time and there is always a lot of negativity on new stuff. I remember the first days of the internet, and most people did not see a need for it. Slowly things changed because of X, or Y and ... and now everybody freaks out the moment they lose internet connections for 5 seconds.
Cryptic0677@reddit
There are way more use cases of AI than just generative AI
Strazdas1@reddit
arguably generative AI is the least impressive part.
Strazdas1@reddit
As a VBA/Python programmer sometimes i feel like im permanently in this stage for 2 years now. You just cant trust what the model gives you to work and troubleshooting it takes too long. sometimes it gives you ideas though.
theQuandary@reddit
What kinds of frontend jobs are you doing where the code is the hard part? In all my jobs, finding out what you need to do is massively harder than writing the actual code.
Strazdas1@reddit
I hope they will find some other practical uses for all these fiber optics before the internet fad runs out.
QuinQuix@reddit
It's not a fad, I was on the fence until o1 dropped.
The only discussion left now is when
Creative_Purpose6138@reddit
Does o1 run locally?
QuinQuix@reddit
Not yet it doesn't.
No product by OpenAI runs locally as far as I know.
But even if you're personally committed to local solutions Skynet can come over and kill us all running a proprietary model.
Committing to local solutions is a personal choice that I emphatize with but it is a distinction that has no bearing on the speed of societal upheaval.
work-school-account@reddit
I think their point is since the good models like 4o can't run on your phone, there isn't much point in including AI hardware on phones, not that AI in general is pointless.
QuinQuix@reddit
AI hardware is useful for much more than LLM's.
It is useful for voice recognition, noise canceling, speech synthesis, computational photography, image editing applications, video editing applications, masking in apps like tiktok and snapchat, translation services and potentially in some games as well.
It will basically run functions that already exist at less than half of the current power usage or strongly boost the 'smart' capabilities of such applications at iso power. But sharply reduced power usage and more responsive performance in already existing apps is the biggest immediate benefit.
You should think about this much more like an ordinary accelerator like h.264 or AV1, or the MMX extensions that helped make the pentium a succes. The benefit is pretty generic.
It really isn't only about chatgpt.
Cryptic0677@reddit
AI has already been doing a lot of things for years before generative AI became hot news recently. One major thing you interact with every day is photos
Not-the-best-name@reddit
Yea, the only thing that makes sense now is improving your camera, AI chips do really well on graphics cards for improving gaming frame rates. Not something that my phone needs, but that is where real time tensor cores work.
I guess auto complete using a locally running model on your phone could be useful but Iam fine now.
gokarrt@reddit
companies who wanna sidegrade consumer devices while avoiding hard gains in conventional performance.
pianobench007@reddit
Gamers do. DLSS performance matters. If XeSS delivers better visual, less artifacts, and at a lower cost per node for Intel, then they will win the price to performance and take marketshare.
They don't want the older gamer. Market share is always growing in the young markets.
Likewise, for the Ai phone? It's not you they are targeting. You and I are old? We are busy with life, cars, fixing a house, planting a garden, running a business, visiting our kids, seeing our kid's kids, and seeing our friend's kids. Etc....
We are busy. But the kids? They grow up not knowing Google. They will grow up talking to this Ai. They will grow up using the Ai to make their Art, tell them pickup lines to send to the boy/girl they like, and more.
They will find ways to game the Ai to play pranks on each other. Photoshop on the fly after snapping a photo of their friends and other things.
The Ai is there and it is really useful. Maybe you take a photo of our new car. Or you photograph you and your friend but someone photo bombs you.
No sweat. You can edit them on the fly with the first Ai application. But the next one and the next one? Maybe it's already here?
Photos sent via text get compressed to hell. Maybe next Ai is to upscale those images. Basic shit. Next is video. And now we can send 4K video texts.
The shit will scale. Nvidia threw us a bone. YouTube Ai video compression. Saves on marginal data compression but it's there. It works.
1930s depression era guy? Hears about 2020s mining $64,000 on a coin from a box?
Mind blown. He doesn't compute. It's incomprehensible to him. This idea. Same thing going on here right now.
trololololo2137@reddit
Me when waiting for photos to process on pixel 7 pro
TheLegendOfMart@reddit
Is Snapdragon 8 Gen 1 considered dog slow or anything?
This constant race to be faster and hotter is dumb.
iDontSeedMyTorrents@reddit
Google always claimed they tailored the NPU to their own algorithms.
Geekbench AI scores taken from here:
PMARC14@reddit
That is actually an impressively huge gap. I do wonder what Qualcomm's next smartphone NPU will look like now that AI stuff is in full swing.
conquer69@reddit
Is it though? I can't think of any use case for that theoretical 6x advantage.
TheAgentOfTheNine@reddit
6 timea faster cat filters for your teams calls.
kenyard@reddit
I mean the pixel does some pretty good stuff with photos for one. I only have the 7pro. But it heats up pretty crazy while doing photos even. A faster more efficient chip for all of that will be a great improvement. AI annoys me as it was a term for something which learns and improves. Now its a term for stuff based on machine learning even those which doesn't improve on that database through user input
conquer69@reddit
Sure but how much better is it than the competition? Is it worth having a much slower and less efficient soc in every other aspect on the off chance the AI improves the photos?
8g4 is coming out soon as well.
kenyard@reddit
Oh no, snapdragon is probably the better option overall.
QuinQuix@reddit
The swing hasn't hit the ball yet, but I fear for the ball.
It definitely is in full swing.
basedIITian@reddit
None of the vendor NPI drivers are mature right now. GeekBench AI numbers will remain unusable for quite some time.
Vince789@reddit
Yep, NNAPI has been depreciated, so GeekBench AI scores using the NNAPI Backend are completely useless
E.g. Google's Tensor G2 actually gets higher NNAPI scores than the newer G3/G4
Qualcomm, MediaTek, and Samsung have released their own custom TensorFlow Lite delegates, hence have some more representative scores
But GeekBench AI is very new, so no support for MediaTek yet
And scores will probably be unusable since GeekBench AI will be updated with fix/changes. Similarly those custom TensorFlow Lite delegates are also quite new and getting frequent updates/fixes too
iDontSeedMyTorrents@reddit
Thanks for the education.
PMARC14@reddit
Thanks for sharing I went looking after my comment and couldn't figure out the difference in reporting. What I get for taking Geekbench numbers at face value.
ExtendedDeadline@reddit
I hate this timeline
Vince789@reddit
Google hasn't updated their NNAPI drivers for ages (NNAPI has been depreciated)
Hence the Tensor G2 has the fastest NPU according to NNAPI benchmarks:
chasteeny@reddit
deprecated
College_Prestige@reddit
What ai features do the pixels run locally though?
mikethespike056@reddit
why are they using NNAPI lmao 💀
dj_antares@reddit
That score is a lie.
Single Precision Score | 456
Half Precision Score | 16150
Quantized Score | 38846
https://browser.geekbench.com/ai/v1/55021
no_salty_no_jealousy@reddit
Google pixel is trash. Glad i'm not buying those garbage overhyped google phone.
Lalaland94292425@reddit
Google is an evil corporation
winner00@reddit
No surprise as G4 is pretty much G3 but with new CPU cores and a slightly higher clocked GPU.
I expect G5 to be a big improvement as it'll be made on TSMC and also made by pretty much an entirely new team.
Here's a spreadsheet of who they've hired over the past few years to work on their Next Generation CPU project. A good majority of them were hired in 2022 and beyond.
The main people were all hired from IBM and were responsible for creating multiple generations of IBM's CPUs, which are fully custom. As you can see in the spreadsheet they all worked there for 20+ years. Some even 30+ years. Lots of patents between them also. Now you might be asking "well wouldn't these people be for Google's server team?". It's a good question but no they are for Tensor because they've said so. The Chief Architect has said "We're the team behind Google Tensor...".
Now that we’ve established this project is for Tensor I also believe a future Tensor will have fully custom CPU cores. I think this because the Director of Next Generation CPU Design has said “Is CPU design in your DNA? Do you want to take the challenge of building one from scratch?”. Also because the CPU Performance MicroArchitecture Lead has said “together we will build new & awesome CPU cores for our new Android friend.” I think the second comment is the best evidence we have of custom cores. I don’t know if we’ll see these custom cores in G5 or off the shelf ARM cores because 2 years to create custom cores could be a little too short of a timeline.
As for the rest of the team you can see they’ve hired a huge amount of talent the past few years from everywhere in the industry. Places like Qualcomm, Intel, Apple, TSMC, IBM, and many others. A lot of these people probably haven’t had a big effect on Tensor yet because G3 would’ve been pretty much done when they got hired and G4 is G3 with new CPU cores. My guess is the TSMC G4 was supposed to be the first thing from this team but got delayed so G5 will be the first thing we see from them.
Overall though i'm pretty optimistic about the future of Tensor.
Old-Benefit4441@reddit
It's quite a bit more efficient though. I'm sure the Snapdragon 8 Gen 1 still performs fine. I think Google is right to focus on AI stuff, that seems like the most practical use for more processing power at this point.
virtualmnemonic@reddit
My s22u is still performant as ever on the latest OneUI; I imagine a similar chip running vanilla android will have zero problems.
This isn't the issue it's made out to be for 95% of real-world consumers.
Anxious_Section_9026@reddit
Doesn't surprise me at all. Google has never positioned Tensor as a performance beast. It has always been all about power efficiency and AI.
That said, given the price new Pixels go for they should at least be on par with the latest gen chips.
noxx1234567@reddit
They don't even do efficiency right , tensor is infamous for heating up even for simple tasks
NeonBellyGlowngVomit@reddit
Heating up alone isn't a reliable indicator for efficiency. Thermal transfer as opposed to insulation, mate. Basic physics.
The Tensor G4 is the third most efficient soc, right behind the 8Gen3 and 8Gen2, in this very same test.
https://i.imgur.com/FkV66V8.png
damwookie@reddit
No it isn't. The reason you can compare 8gen1 efficiency is the scores are similar. There are nearly a dozen chips you cannot compare because you would have to limit their speeds heavily down to the G4 speed to be able to compare their efficiency. Most chips lose efficiency the higher they boost. A chip that cannot boost for shit doesn't magically make it more efficient than the ones that can.
NeonBellyGlowngVomit@reddit
Y'all love to play mental gymnastics with this.
You can have a chip that performs lower and has higher efficiency without relying on masturbatory handwaving like "cannot boost for shit."
Hey look, ANOTHER test where the G4 is in the top 3 for efficiency, not performance: https://i.imgur.com/viz5Vyf.jpg
bob-@reddit
Pixel loses in every battery life test to similarly priced phones, what exactly does this amazing efficiency translate into exactly? 🤡
NeonBellyGlowngVomit@reddit
Because SOC power consumption is the only thing that drains the battery?
Look everybody, Simple Simon over here is a tech expert!
dogsryummy1@reddit
Google is fucked. The only way they're managing to keep up with the competitors in efficiency is by limiting performance to ONE THIRD. If they try and clock their cores any higher, power consumption will skyrocket. Meanwhile if you bring down any Qualcomm/MediaTek/Apple GPU to Tensor's performance level, they can do it at a fraction of the power. These graphs tell the full story:
https://www.socpk.com/gpucurve/ https://www.socpk.com/cpucurve/
It's incredibly misleading to take one point along that power-performance curve and draw conclusions.
NeonBellyGlowngVomit@reddit
Google is fucked but not all those low end and midrange budget model manufacturers? Okay, bro.
Honestly, I think it's hilariously you guys are this triggered over a SOC. Something that isn't even 5% of Google's income yet somehow they are FUCKED. They made no promises to you guys and yet y'all are acting like they broke your hearts.
dogsryummy1@reddit
We speak in hyperbole here. First day on the internet?
Google is stuck in a corner where the only way they're managing to keep up in the efficiency department is by dialling back their clock speeds more and more. Funny you mention low end models, are you suggesting that Google's chips are comparable? Bit of an own goal there. May I remind you what price bracket Google's phones occupy? (hint: it's not the budget segment)
Put another way, give me a list of $800+ phones and find me one that performs worse than Tensor. For the prices that Google charge, it's simply unacceptable. This is the second year of price hikes, while the iPhone hasn't changed in price since 2017.
NeonBellyGlowngVomit@reddit
Yawn. Wake me up when the majority of phone sales aren't subsidized before anything you said has any real relevance. Phones bought at full price are a tiny fraction of actual sales. Verizon and AT&T, a good 67% of the US Market share, don't even allow outside devices to be brought in on their contract plans.
You might have heard of something called... discounts.
People are paying $25 a month for the Pixel, not $800 up front. When it comes to these payment plans, people aren't shopping by value or price, they're shopping by the whole package. These people, who make up the vast majority of the cellular users out there mind you, aren't masturbating to benchmarks and absolute figures like you guys are wont to do. They want a device that works well. Not cuts their bleeding edge like a depressed goth girl.
Your outrage is performative. And tiresome. You'd find a reason to shit on the Pixel even if it ticked all the boxes.
And yes, I care about efficiency more than I care about performance. That's why I use a custom kernel and downclock and undervolt my phone. Still won't catch me using a Pixel for other reasons.
bob-@reddit
LOL if you "downclock" then pixel loses in efficiency to a whole bunch of other phones
mockvalkyrie@reddit
Sir, this is a hardware subreddit...
damwookie@reddit
When other chips are outputting double FPS you cannot compare efficiency. You need to compare similar to similar. A chip able to output 109fps will likely do so less efficiently than when outputting 44fps. The clue is in the name "performance benchmarks". To compare efficiency you need to have them complete the same task. Not have one chip perform over double the tasks of the other.
noxx1234567@reddit
Anyone who is using pixel phones will attest that tensors are nowhere near efficient compared to TSMC made chips.
Even if Google did focus on efficiency , you can't overcome hardware inefficiency
bokaaaa-@reddit
I'd blame Samsung, at least for the efficiency
996forever@reddit
I’d blame them for using a cheap fab without passing savings onto the end consumer, actually.
noxx1234567@reddit
True that samsung fabs are not good , thankfully Google is moving to TSMC soon
ExtendedDeadline@reddit
Nah
Tasty-Traffic-680@reddit
Should they though? At what point is performance good enough? Just look at Huawei. They've been selling the shit out of phones that have worse performance than this. Phones have gotten really good in the last few years, even the cheap ones.
sabot00@reddit
Probably loses to what Huawei gets out of SMIC at this point 😂
Tasty-Traffic-680@reddit
Not really. Especially since they started using their own GPU cores.
https://www.notebookcheck.net/Kirin-9010-Huawei-Pura70-Ultra-s-chipset-fails-to-beat-mid-range-SoCs-in-benchmark-tests.829853.0.html
The upcoming 9100 has all kinds of big rumors about similar performance to snapdragon 8 gen 1 or higher but it's also delayed until at least November so we'll have to wait and see. Also that's just for cpu performance. It's likely their GPUs are still shit compared to ARM and everyone else's designs.
kyralfie@reddit
But extremely close. Kirin 9010 is the latest one, it's extremely close to Tensor G3 and 8 gen 1. https://socpk.com/allperf/
This site also has CPU, GPU and efficiency curves.
Tasty-Traffic-680@reddit
Cool list, thanks! I wonder how many samples they have for each. They have the dimensity 9300 just a hair over the snapdragon 8 gen 3 but I have a feeling it would be reversed if not for the 30% GPU weight. Then again from what I have seen they literally are neck and neck depending on phone config and tuning.
kyralfie@reddit
They have separate charts for CPU & GPU. No need to wonder. ;-) Dimensity 9300 is mighty impressive on all fronts.
Tasty-Traffic-680@reddit
Oh cool. I was checking it out on mobile before and didn't see the arrow to go back. Going to bookmark the site, thanks
kyralfie@reddit
Yeah, these guys are great. They also have a youtube channel called Geekerwan in chinese but it usually comes with english subs on day 1. Highly recommend it. The depth of testing is unrivaled.
_Lucille_@reddit
I am curious how the chip has such a high power efficiency score while still having similar battery life compared to phones that use rival SoCs.
I guess the other SoCs uses less power by getting stuff done faster or that they have efficiency modes?
chronoreverse@reddit
Power efficiency curves aren't a straight line so the faster you push the more power you use to do it. On the flipside, when you run something slower, you can often get higher efficiency.
There's a point, depending on the chip where slowing it doesn't doesn't really gain much. There's also a point where increasing the frequency (assuming it even can do it) where the additional cost of power is so great it doesn't make any sense.
This is why you can't compare efficiency when running at different power and performance. You need to make multiple comparison at iso-power and iso-performance to get a real picture.
Going back to your original question, I suspect the Tensor simply doesn't have a great curve (only some points in a narrow band are efficient) along with Pixel software being full of stuff waking up using the battery (for "quality of life" purposes). Many other phone manufacturers, particularly ones not available in the States, have aggressive background process management which gives better battery life (in exchange for the "quality of life" services being as available).
specter491@reddit
If you don't game on your phone, will you ever notice the difference? I run android auto, YouTube, take pics, etc and everything runs fine. I don't game so maybe that's why I don't notice any issues
clingbat@reddit
I'm enjoying my Pixel 9PXL a good bit more than I did my Pixel 8P, so there's that.
ExtendedDeadline@reddit
Pixel evangelists have been saying this y/y since pixel 6. I tried the 7 and 8 and they were both not great. Is the 9 actually resolving issues of battery, thermals and signal?
clingbat@reddit
In my experience, yes. At least on T-Mobile in the US service has been better. Battery lasts about a day and half now for me with the XL vs dead at end of day with the 8P and it charges faster. I've haven't had a big issue with thermals in the past but they added a vapor chamber in the pro, pro XL to help with that.
NeverDiddled@reddit
I'm still rocking a Pixel 3. Battery life is 1.5-2 days even after 6 years. Loads websites plenty fast, almost as fast as my gaming PCs.
The only complaints I have are: 1) It stopped getting security patches. Fortunately the new pixels get 7+ years. 2) Camera is worse than modern competition, particularly when zooming in.
I would not describe myself as a "Pixel evangelist", but I will say it's a 6yo phone that solved every problem I wanted it to. Only thing I wish is that the newer versions of it were cheaper.
clingbat@reddit
I've been a long term Googler phone wise:
Honestly the only reason I started upgrading faster after the Pixel 6 is Google keeps giving generous trade-in values in the US if you time it right, the key is not to force it. I paid like $30 to trade up from 6 to 7, about $150 up to the 8P and after the Google store credit, about $200 extra up to the 9PXL ($400 total after trade-in). We used the $200 store credit to subsidize my wife going from a 7 to 9 cheap with her trade-in.
NeverDiddled@reddit
Do you use the digital assistant to answer unknown callers? I have been pondering switching an iPhone. But that is one feature I would sorely miss. I can not remember the last time I got a spam call. It's been years. It was definitely before I switched on Google's AI to auto-answer and screen unknown callers.
Honestly when folks talk about no applications for AI on a phone, I shake my head a little bit. I think these folks have limited their definition of AI to LLM, and naturally don't see much use. But Google has been doing AI since long before LLMs, and at times finds terrific applications. No more spam calls? That's a great application of AI.
clingbat@reddit
Yea I never deal with spam, screen if necessary and use the assistant for waiting on hold too, that's all nice.
bob-@reddit
Better battery life than what phone? Because its loses there as well
Pixel 9 Pro battery comparison
kyralfie@reddit
How can one be so context unaware? He obviously compares it to his old Pixel 8 Pro.
bob-@reddit
So it has better battery life than the absolute bottom of the barrel phone out there? That's good I guess 😂
EastvsWest@reddit
That would be okay if they weren't asking for high end prices for their flagships. I really like the new pixel phones but I'll stick to Samsung that actually gives you high end hardware and software for the money they ask for.
zakats@reddit
Didn't they put all of their R&D effort into their AI bullshit for this gen?
Google wants to gain a major foothold with their AI nonsense so that they can continue to lead the data cartel.
ExtendedDeadline@reddit
More research into phone signal would have been appreciated
iDontSeedMyTorrents@reddit
No, G4 is a very minor update to G3, basically just an update of the CPU cores. G4 was supposed to be Google's first TSMC chip, but wasn't finished in time. So instead they just slightly modified the G3 and sent it out.
EnesEffUU@reddit
What do you mean it was supposed to be their first TSMC chip? It's been rumored for over a year that Pixel 10/G5 would be the first TSMC chip.
metal079@reddit
The ai performance on it is pretty bad as well.
kouyou@reddit
The tensor are still bad, that when I sent my Pixel8 to a service center, I used my old Huawei Mate 20 Pro, released in 2018 with a Kirin 980 SoC and it was marginally better than my P8 at running Genshin Impact.
Sudden_Mix9724@reddit
yea this was expected...it's not magic....
companies take decades to make the best chip...
even samsung who's making Exynos is struggling in high end while making for like 10yrs..
it's Google's 3 or 4 gen is it?? maybe give it another 3-5years and I m sure they will be good contender in flagship space...
in the end consumers need choices & competetion afterall..
BrideOfAutobahn@reddit
Google could have chosen to develop their chip internally until it was competitive with the market instead of releasing products with bottom of the market performance.
Exynos still sucks too?? Wtf is going wrong at Samsung.
Sudden_Mix9724@reddit
that's unrealistic.. microprocessor being one of the most complex things that humans ever made..it takes decades of hardware+software optimisatiopn...not to mention 3rd party supports which will require it to be in market...
just like how intel Arc GPU are at moment...now they are in bottom of the charts and in 2 generation....but it might years years before they can catch up to market dominators..
it's impossible to keep it inhouse for soo long and bring out a top tier processor just like that...
BrideOfAutobahn@reddit
Apple kept their ARM CPU project in-house for at least a decade before M1 launched, and it launched in a very competitive state.
fenrir245@reddit
???
Apple has been using complete in-house silicon in their phones since iPhone 5 at least. And their first attempts wasn’t getting up to par with the Android counterparts either.
PMARC14@reddit
They are supposed to be making a mostly in-house design next gen ( I assume they will still be getting bits like modem and WiFi from Samsung). I am interested in what that scores as I think they originally wanted to bring that out now with all the changes but the team couldn't manage. And the flagship costs for the phones alongside it make it pretty disappointing, especially when you have to eventually pay for the AI features.
Darkpurpleskies@reddit
Was surprised when my old S21 FE (SD 888) did object erase in photos faster than my Pixel 8.
tr2727@reddit
How does this compare to g2? Like the tensor are not good but the modem improvements and what else? Using pixel 7 and debating myself how long before I look to get something better
BrideOfAutobahn@reddit
Do the specs impact the usability of the phone at all? Are their chips good in any way? I’ve heard their image processing is good, but I’m not sure how much of that is the chip vs. proprietary software.
I don’t think I’ve ever seen a Pixel phone IRL, now that I think about it. Do many people buy these things or are they more of a niche item?
Google’s Pixel store page headline is hilarious too, “The only phone engineered by Google.” AKA “Please don’t blame us for all the bad Android phones everyone else makes” 🤣
Tomi97_origin@reddit
Still pretty niche as even in their best markets they only have like 5% market share.
Maleficent-Salad3197@reddit
Well that does it. I'm throwing my Pixel 3 out the Window.
gelade1@reddit
we will see how behind is their chip design department in tensor g4 when it's made by tsmc. no excuse then.
bn_gamechanger@reddit
Google can get away since their intended market is US and not many people here play games on their phones. To use social media apps and click pictures, it’s a good phone.
Too bad their USP of being cheaper is gone and they charge like Apple does.
iDontSeedMyTorrents@reddit
You mean to tell that the same GPU from the G3 but clocked ~6% higher scores about the same amount higher? No way, who could predicted this?
PolishedCheeto@reddit
Doesn't tensor3 beat spadragon 8.1?
Creative_Purpose6138@reddit
8 gen 1, nahh that's wild 😂😂