Only thing that would make extra performance matter are video games and I don't play those on phone.
Theoretically, I would care. But in practice, I just took out an old budget android phone, gave it a go at browsing the web (news, images, and videos), and it seemed not to be noticeably slower than my much newer and pricier iPhone.
Benchmarks would probably show a great difference between the two, but my eyes can't tell.
My interest in this topic is mostly the same as the author of the post I linked, namely making sure that developers test and measure on lower-end devices to make sure they're not building sites which exclude key demographics. For example, a government site really needs to work well on the kinds of phones seniors get under the FCC Lifeline program.
But that is different from the observation of the beginning of this thread, which is that Geekbench scores don’t really reflect practical limitation of devices. A big difference in the score does not mean a noticeable difference in performance when doing everyday web browsing on the phone. An observation that my own experience echoes.
Even if I’m extra-tolerant (I don’t think I am), it would surprise me that a performance difference I can’t even notice is going to be so intolerable to another human being that they are excluded from using the website.
When it comes to governmental services on the internet, I think really the focus is on whether the site _works_ on all kinds of devices, which means whether the rendering and interactions are correct. Some can fail to render on a low budget phone, or if some button doesn’t work, etc. these are real problems but that is quite irrelevant to benchmarks.
The broader point wasn’t that you should buy an iPhone but simply that the reason to care about single core benchmarks is that browsers fit that profile. It’s still perfectly reasonable to conclude that’s good enough for the sites you use.
I also feel like there’s an interesting angle about ad blocking here since that probably matters more than multiple processor generations.
One factor which used to be under appreciated was battery degradation prior to that whole “Batterygate” flap back in 2016 where a lot of people learned that iOS throttled processor performance when the battery could no longer supply enough voltage for peak performance. That’s a lot more visible now so it’s less of a surprise than it used to be.
Most of the problem is that developers aren’t focused only on performance, so as the baseline hardware capacity increases the apps will slowly start to use more since very few people are going to spend time on something which seems fast enough.
That doesn’t happen all at once but it adds up over the half decade a phone will last. If you bought a phone and never updated anything, its performance would seem far more consistent … at least if you could avoid getting malware installed.
That third point is why I shared the link above: a lot of developers upgrade more frequently than average people and that means that our instincts for what seems fast enough might be missing things with our apps. That usually doesn’t mean things are unusable but it’s still polite to use your work on, say, the phone a senior citizen gets subsidized to make sure that you’re comfortable with that being the public face of your work.
(Bandwidth usage is at least as important here, too: use your website on 3G or ask how much it’d cost to use on a metered plan)
As a web browser has become a general purpose VM, it has capabilities far surpassing what is needed at least 50% the time, I'd go as high as 90% of the time personally. Which is great that we have a magical run anything anywhere VM, but it's a double edge sword of the performance, security and privacy issues that come along with it.
If there was a well defined subset of functionality that browsers presented as the default "web enclave" and then opening up all the bells and whistles further was an opt in choice, a large chunk of those issues would go away. But that requires a bunch of people to agree on what that subset should include and the horror of trying to keep it updated over time. And that's _everyone_, including users putting up with "opting in" rather than everything just working.
I'm not sure I can see a subset of features becoming the default approach, at least until there is some horrible self propagating browser worm that impacts a large percentage of the world (or maybe by the 2nd or 3rd time, we will have had enough).
What about buying a ticket and it failing at the last step because your browser determined that js was unneeded?
Basically have to keep giant lists of what js actually enhances the people experience vs just being useful to the publisher.
It’s cool and all for Apple to have an edge in single core performance but it doesn’t matter all that much in the real world and the s23 is objectively the faster phone.
Do you have any reproducible test results showing that or are you just asserting tribal loyalty? I jest, we all know the answer.
You might want to read up on how iOS and Android differ in key ways: native code versus JIT, garbage collection versus ARC, and all of the system level differences for how many things can run in the background as well as quickly apps can save and restore state - faster CPUs and SSDs allows iOS to be more aggressive there, too, and developers can tune more aggressively because they have fewer, more consistent configurations to support.
Also consider it from the perspective of the engineering trade offs that the manufacturers make. More RAM means lower battery life and a higher price, and consumers are also sensitive to both of those. Apple controls their stack so tightly that they designed a custom SoC and market-leading CPU, and they measure the user experience very closely since that’s a lot of what they’re selling — if there really was a gap, adding more RAM would be one of the easiest ways to close it. Similarly, Samsung wouldn’t make every phone they make have lower battery life and cost more if that only benefited a small percentage of users. They’re making different choices because they have different systems but both of them watch this stuff closely.
Apple hardware is just that much better, it is not about the software stack.
(And, yes, I should have said “Java's GC vs. ARC” since that tends to have higher peaks)
Reference counting is a GC algorithm.
Thanks, I knew that Android had native code support but I've always heard that described as something mostly used by games.
> Reference counting is a GC algorithm.
Yes, I should have been pedantic and written “Java's GC” since the point was that historically it's tended to trade RAM for performance. The classic complaint I've heard from our mobile developers was that Objective-C forced you to focus more on memory management.
This is important for developers to know about because if you use a recent iPhone your experience is not just fundamentally better than any Android user's but it's especially significant compared to the kinds of cheaper phones many people buy. Look at the chart here: a Galaxy S22 is comparable to the iPhone XS which shipped 4 years earlier but even the S22 towers over the Moto E30 & other budget phones.
https://infrequently.org/2022/12/performance-baseline-2023/#...