Fuck Qualcomm right in their Qualcomm-hole


#99

Bluetooth is really a pretty bad protocol for audio. :/ I hope the newer standards for it are better. I’ll check out BT5


#100

#101

Wumpus and Qualcomm, sitting in a tree…


#102

The main difference is that one of those is reduced in significance over time as you become more adept with navigating the strengths and weaknesses of the UI; the other one is like a rock in your shoe that just won’t come out.

That said, I blame the software stack (and vendor-specific optimization) almost as much as I blame the hardware. There’s a reason that Google is exploring an entirely different software architecture with Fuchsia, aimed at much smoother UI performance from the ground up, and it’s not just about changing the UI paradigm.


#103

Methinks he doth protest too much!


#104

This is the thread for linking to your own twitter/stream/blog:


#105

I WANT TO LIKE SO HARD


#106

Soooo basically a brand new $479 OnePlus 5 is the same speed as a used $200 iPhone 6 from September 2014, as far as browsing the modern web goes…

iPhone results

  • 5s → 48
  • 6 → 71
  • 6s → 119
  • 7 → 175

And the other shoe has not dropped in the form of A11 (iPhone 8) yet.

I guess the “good” news is that the very latest Android devices can claim iPhone 6 speeds. 🎉


#107

Apparently the OnePlus 5 is cheating at benchmarks, so that score might actually be lower.


#108

I don’t trust anything from those idiots at Ars anymore. They include these totally worthless graphs too!



#109

Multi-core is irrelevant for 95% of daily use. Eight slow cores isn’t gonna deliver a good experience on the web; quite the opposite. “But this one goes to 11!” Now, @lantz, if you regularly encode videos on your phone, then yes.

However, the GPU definitely looks competitive, which is good news for gaming. If the PS4 and Xbox have taught us anything, it is that CPU isn’t nearly as important as GPU for games… very true.


#110

Single-core is irrelevant for 100% of the time your battery is dead. So that should be the only thing that matters on a phone. Or perhaps multiple things can be important and there are different tradeoffs that appeal to different people?


#111

How are the graphs worthless?


#112

Sure for stuff like “does it have a SD card slot” and “do I like the aesthetics of the case design” and “is the screen OLED” and “how big is the screen / battery” I agree with you.

Or if you’re in the “640kb is enough for everyone” camp where you literally don’t care, and your daily driver is an ancient Android 1.5 crapbox or iPhone 4, fine. I met someone like that a week or two ago. (But I hope you don’t get p0wned with the zillions of security exploits on an old device)

I just wish you could buy any – literally ANY – Android hardware that was within 50% of where an iPhone 7 is in day to day performance. You know, launching apps, rendering web pages, switching apps, the basics.

https://www.youtube.com/watch?v=-pFpv0Im1UY (Snapdragon 820, Galaxy 7)

https://www.youtube.com/watch?v=k_PK_6F_Bhk (Snapdragon 835, Galaxy 8)

At least it is not a complete ass whipping any more. Until the iPhone 8 and A11 are released in a few months, I guess.


#113

Goddammit Lantz don’t encourage him!


#114

Single-core (past a modest baseline) is irrelevant for 95% of daily use. Cutting page rendering time in half sounds amazing until you realize that a page that used to take 3.5 seconds to load now takes 3.25 seconds because the main bottleneck is still the 3 seconds to suck down all the data over LTE. And that once the page is loaded, there’s no perceptible difference in user experience. But for that quarter of a second, you’re truly a god among men.

I agree! That would be nice. I’d probably even be willing to pay an extra $50 for my next phone if that were an optional upgrade.

Wrong! The basics are using apps and reading webpages. Startup times amount to a rounding error across the total amount of time a device is used.

You’ve trotted this quote out several times, but seem blissfully unaware that there’s only one person in this discussion obsessing over ever-larger numbers divorced from any rational understanding of their real-world value.


#115

Sorry, poorly phrased sarcasm. It’s a $479 phone using a well known and endlessly talked about (in this thread) chipset. Of course it is going to perform exactly as it will in the one the one test he linked. No one in the world thought it was going to magically outperform a $650 phone with the chipset it has. Everyone interested in buying one cares about other things like camera and battery and not having to run iOS.

Sorry, I had so much free time because it takes so long for my Galaxy S6 to load anything because for me to be able to check my email I have to launch, wait for the menu to load, and then kill 25 apps and 10 games to get to the gmail part of my lap. That’s how everyone uses their phones right? Launching and killing apps all day in a loop right? I wish I could just load the 6ish apps that I use during the day and then multitask between them with a single button push.


#116

Nope! JavaScript rendering time is more on the order of seconds on less modern Android devices:

Nexus 6p (high-end Android)
1248ms avg (before)
248ms avg (after)

Nexus 7 (2013 mid-range Android)
4078ms avg (before)
636ms avg (after)

One of the reasons this is so irritating is that the velocity of the Qualcomm CPU perf change is terribad, so it will not only take 5 full years just for all the ancient very slow Android 2013-2014-2015 devices to filter out of the user base through attrition, but on top of that, 5 more years just to get parity with where the iPhone 7 has been for a year.

Unless you religiously Adblock with the fervor of a zealot, you are loading several JS frameworks with every web page. Hell just load up http://theverge.com, it’s basically a web benchmark all in itself. I hate to be the one to break this to you, but as much JS as there is on the web now already (it has not been possible to plausibly browse the web with JS off for years) there will be an order of magnitude more in 3 years.


#117

Unoptimized JavaScript based on a framework unsuitable for the use case, you mean. If your app takes 4 seconds to render a forum thread because it’s individually processing 150 components per post, that’s on you. As the article shows, you were able to get the same functionality down to 638ms on 2013 hardware after optimization.

I just did, and clocked 2.25 seconds on iPhone 6s vs. 3.0 seconds on Note 5. That difference isn’t nothing, but it’s small potatoes in the context of a several minute browsing session (there was no speed difference for opening an article, or scrolling around within it).

Plus which, it’s not 1997, so I don’t remember the last time I randomly loaded a front page like an animal. We have RSS readers for that shit now. And the sorts of actions I actually do dozens of times a day on my phone (opening a news story from RSS, browsing to a new thread on Qt3 or in Tapatalk, opening an email, changing tracks in a music player, switching channels in Slack) all have either absolutely no perceptible performance difference or a slight edge to the Note in how long they take between the phones.


#118

Perfectly suitable for use, provided mobile devices continue their meteoric yearly improvements in performance to close the gap with traditional notebook computers. Very true on the iOS side, but sadly… quite untrue on the Android side.

Yes, and it’s true this optimization helps all platforms. However, even after that “optimization” there is a 10x performance difference between iOS and Android. It’s just a smaller number (500ms vs 50ms), but the proportionality of the difference remains. For example, here’s the Google Pixel versus the iPhone 7:

Yes, that’s 310ms versus 27ms. One whole order of magnitude, 10x!

That is fucking nuts.

If you don’t care, then you don’t care. But let’s look at WebXPRT:

WebXPRT 2015 uses scenarios created to mirror the tasks you do every day to compare the performance of almost any Web-enabled device. It contains six HTML5- and JavaScript-based workloads: Photo Enhancement, Organize Album, Stock Option Pricing, Local Notes, Sales Graphs, and Explore DNA Sequencing.

(WebXPRT is not a great benchmark… but it’s another “general web” benchmark if you don’t like JetStream or Speedometer or Ember. I just ran it on my 2017 iPad Pro and got 265.)

So you’re telling me, your Note 5 … which scores 93 versus the 194 of an iPhone 6s in the above web based test, has

Look, if you prefer Android, that’s fine. My specific beef is that Qualcomm SoCs are holding the whole mobile industry back by failing to deliver anything remotely comparable to the performance of Apple SoCs. I don’t mean 20%, 50% slower. That wouldn’t be great, but it’d be tolerable, if short term. I mean 2x-3x slower and in modern JavaScript benchmarks, 5x or more slower. The bad / good news is, the web is lousy with JavaScript and it’s getting more lousy with the stuff by the second. “JavaScript performance” increasingly means “web performance”.

And worst of all, this performance slump seems to be long term, if not permanent. That is the absolute 🍒 on top of this 💩 sandwich.