a companion discussion area for blog.codinghorror.com

The Tablet Turning Point


#1

Remember how people in the year 2000 used to say how crazy and ridiculous it was, the idea that Anyone Would Ever Run Photoshop in a Web Browser? I mean come on.


This is a companion discussion topic for the original entry at http://blog.codinghorror.com/the-tablet-turning-point/

#2

They’re not “x64.” x64 is a contraction chiefly used by Microsoft as shorthand for x86_64 (which IMO is a terrible shorthand not least because it is numerically lower than its predecessor)

Neither the Tegra K1 nor the iPad Air uses the x86 architecture. They’re both ARMv8 based devices. Specifically, their 64-bit mode is called AArch64.


#3

While I gotta say that iPad Air 2 is impressive (just bought one myself) the Ember benchmark where i7 is only ~20% faster is deceiving.

Let’s take Octane 2.0 for example, you’ve included scores for it in your graph.

My desktop i5 (3570) scores ~33K on Chrome x64 on Windows 8.1.

That’s 3.5 times faster than iPad Air 2.

Which, considering the clock speed - almost exactly 3.5x of the iPad Air 2 (it’s overclocked) - isn’t that impressive, I’ll give you that. But it’s still way faster than 180ms vs 225ms of the Ember test would indicate.

I guess you’re reassuring yourself that going Web was a good decision. Long term, I agree.

Short(er) term, until JavaScript tooling, libraries and intermediate languages like TypeScript mature enough for a reasonably comfortable development experience/productivity, I’ll stick with building native apps, on desktop and mobile alike.


#4

I like to use history as my guide, and I believe it’s going exactly the same place it did on desktops and laptops — that no-installing-anything friend of every lazy user on the planet, the inevitable path of least resistance, the mobile web browser.

This reminds me of an XKCD post:

I’ve always had the same belief - One day we’ll come full-circle and start using the web browser primarily once again. The current method of having separate applications per OS goes against DRY methodology.


#5

I wonder how Windows tablets would perform. They are missing in your statistics.


#6

I’m not certain that mobile will ever be as fast as desktop. However, desktops a few years ago became fast enough that they’re more than fast enough for most common tasks. I think we’ll see the same with mobile.

At home we have a general desktop… that was top of the line back in 2008(quad core 3.0Ghz with 8Gigs RAM) upgraded with an SSD and win8.1. It was slow before putting in the SSD, and now it’s fast enough. I’m not home right now, but doing some searches it looks like that Octane benchmark on that machine would score somewhere in the 9-10k range. (my Dell m4600 i7 laptop does 28k)

Perhaps your point is that the iPad Air 2 has finally hit the point where it’s fast enough.


#7

I maintain that the biggest (but not only) obstacle to web apps over native apps on mobile is connectivity!

I work in London, and live just outside. Most of the time my phone has either no signal or just manages GPRS (which, apparently, these days is not enough to load any webpages without timing out). Sometimes I’ll get 3G, and very very occasionally I’ll see that 4G label pop up!

So anything that requires me to have a constant connection I assume it is essential not available to me. Thanks.

Obviously some things are inherently web based (“real” web pages, blogs, news sites etc) and I’m not arguing they should be native. I’ll conceded that Discourse belongs in this group (as does Stackoverflow).
Also anything you only use very occasionally (e.g. travel sites) make no sense in a native app.
But for anything you use frequently and which can work perfectly well (perhaps with some limitations) offline then they really should be native.


#8

Web apps absolutely can be made to work well offline by the way: http://diveintohtml5.info/offline.html

Obviously that feature is still in its infancy, but it’s coming.


#9

Agreed. I think of x64 is an obsolete marketing term invented by Microsoft for 64-bit operating systems.


#10

Samsung Series 3 ARM Chromebook Octane 2.0 == 6209


#11

OK, I’m late to the x64 party, so I’ll be more constructive. ARM calls its new 64-bit instruction set (which the iPad Air runs and to which all Apple iOS software has been updated) A64 (what AMD called x86-64 in the case of x86 chips) and calls the corresponding execution mode AArch64 (what AMD called “64-bit mode”).

But one is ambiguous and the other is a mouthful, so we in the iOS development community use what Apple suggested, ARM64 (which is, to keep the parallel, the equivalent of what Microsoft’s “x64” was for 64-bit x86)

EDIT: As far as I can tell, Google uses the term ARM64 as well for Android.


#12

For context, here are the default browser (IE11) scores of a Surface Pro 3 with Core i5 that I just ran myself. Next to each number is the iPad Air 2 results AnandTech reported using its default browser (Mobile Safari):

  • Google Octane 2 – 8745 / 9430
  • Mozilla Kraken 1.1 – 3310 ms / 4014 ms
  • Sunspider – 106 ms / 285 ms

Notice the iPad is faster than the Surface Pro 3 in one of those tests! And it is close in Kraken. The two major JS tests agree. Sunspider isn’t a terribly accurate test these days.

Hopefully MS can close the gap here in IE12, but if you use the default browser in Windows, you could get better web performance on an iPad now in some cases. This blows my mind.


#13

Are there corresponding statistics on how desktop/laptop environments have improved over time? The relative stats between the two are more interesting to me. Say if you had a series of stock Macbook Airs and you saw that Chrome performance had hit a wall while tablet performance somehow was doubling.


#14

Jeff, something is wrong with your Surface. My i5 (Surface Pro 3) just scored 19091 in Octane 2 in IE 11. Never mind that it’s a benchmark specifically tuned to Google’s priorities.

In the real world, I have more issues making things run smoothly in Chrome than IE. I was just complaining earlier on Twitter than Chrome can’t even do a simple CSS3 translate animation smoothly on my 4k monitor, where IE is like butter. Chrome also can’t handle something as simple as keeping a position:fixed element in place while scrolling with the mouse wheel or touch. And it doesn’t even support CSS3 animations without a prefix still.

The days of knocking IE for being behind are long over. Even legacy support is as bad or worse for me these days with iOS 6.0 users outnumbering IE <9 users, and it lacking more CSS features I find myself missing.


#15

Interesting. I terminated the browser from task manager and tried again:

  • Kraken – 2160 ms
  • Octane – 11914

I do tend to have a ton of IE11 tabs open at any given time, though I use my Surface Pro 3 mostly as a browser, so that’d be all that was running. Let me install pending Windows updates, reboot, and try yet again.

edit: after reboot and Windows updates applied. All tabs closed except for the ones with the two benchmarks

  • Kraken – 2164 ms
  • Octane – 11666

I am running this from the touch / modern / x64 IE11, on battery, with afaik default settings. I haven’t tweaked anything that I know of.


#16

Hmm, forgot that mine is on the Win10 preview, so it’s a newer build of IE (though it still calls itself 11 it’s probably got a good chunk of IE 12 work). It’s possible they’ve started tuning for some of the Octane tests like Chrome and Safari do. But that’s a pretty big jump from your number.

Are you closing all IE instances and pasting the URL of the page with the Start Octane 2.0 button into a new one, as they suggest?

Being on battery will make a small difference, but for me it only dipped down a few hundred points.


#17

You are both comparing one tablet to another tablet. The article suggests tablets are getting closer to pc’s.

They are not because modern pc’s are still much faster.

Sadly though due to the lack of competition on the high end PC market that may change. Intel is not feeling any pressure to improve there. While they are feeling tons of pressure to improve on processors used in smaller devices, like tablets…


#18

I’ve been hearing that argument for at least the last 7-8 years. Despite a few, mostly aborted, pushes in that direction we still seem as far away as ever. Don’t get me wrong - if offline caching and storage ever really get off the ground that will address at least half my argument (there’s more to it - but the offline part is by far the biggest).
Honestly I think it’s a shame the RIA movement has died out (for now) as that seemed to show the most promise (much as I hate Flash - but Silverlight was actually quite good. Well, potentially). I know that raises issues of its own but I hope to see another go in that direction in the next few years. I agree with the sentiment in that xkcd, just not the current execution.

I wrote about about all this a few years ago (some of what I wrote is already cringe-worthily out of date - especially my faith, at the time, in iCloud - but that’s all in-line with the title).


#19

Towards the end of your post you mention issues with Chrome’s V8 engine on Android. Have you tried running the same benchmarks using a different engine / browser, e.g. Firefox? Of course it wouldn’t be representative of user experience, as few people use it, but it would be a good indication of what that tablet could do.


#20

Most everyone I know already use their phone as their primary devices outside of work. I don’t even see that many tablets now, and I’m seeing newer and cheaper tv’s running Android which replace my need for even a media center machine.

I’d be interested to see how competing Android devices would rate on those charts, like the new Nexus devices or the LG G3. Of course it is progress for Apple to start outperforming Android devices that are a year or two older than them.