Bleep Qualcomm right in their Qualcomm-hole


#321

Isn’t this the very same video @chappers already linked and we discussed?


#322

Was it? Sorry, I was reading the write up instead.


#323

hey Wumpy @wumpus , are these Kirin chipsets any good? Could they dethrone Qualcomm as the Android Soc of choice?


#324

No, because much like Qualcomm they’re using ARM Cortex A73s. The neural network stuff smells like bullshit.

All mobile phones, both iOS and Android, work on a platform called ARMv7/v78, which is analagous to x86/amd64 on the PC side of things. (Yes I’m aware that intel tried to break into this space but they weren’t successful so don’t bring it up!)

ARM has two product offerings. First, they design actual CPU cores, like the A53, A73, etc. Every Android SoC manufacturer licenses their CPU designs and embeds them inside their CPUs, including Qualcomm (snapdragon) and Samsung (exynos). Any time you see an article talking about “cortex A73 cores” or the like, that means they’re using ARM’s CPU designs.

ARM also licenses the ARM instruction set. If you go this route, you need to design your own CPU that happens to speak that same language, much like AMD has done with intel computer CPUs. Only Apple has done this, and that’s why only Apple SoCs perform differently. There’s a rumor that Google did it too, and that would result in greatly performant Pixel 2s, but that did not happen. Maybe next year.


#325

Oh sweet merciful Jesus

Let the shittiness take on a whole new dimension of suck. On the other hand probably has to be faster than the utter garbage that was Intel Atom.


#326

Thought you might be posting that. It’s nuts they’re talking about using an 835, which isn’t even all that impressive as a phone CPU. Their emulation layer had better be really good.


#327

A credible tablet part like the a10x would be more sensible. If it existed on the Qualcomm side of the house?


#328

Will these be using Windows 10 Mobile, or some other type of Windows 10?


#329

Broadcom is supposedly about to unveil a $100bn bid for Qualcomm.


#330

Yep.


#331

Wow, the bid went up $30bn in just an hour and a half.


#332

Time to embrace the Qualcomm hole @wumpus

Once again real world tests including web browsing show no considerable difference between my LG V30 and your iPhone X.


#333

I did recently check on my one plus 5 va iPhone 8 and factoring for Discourse doing half the work (if we detect Android we only send 10 posts down per chunk instead of 20, for example) it was only noticeably slower on the categories page, topics were fairly close. It is half the work, though.

Android Chrome has gotten quite a bit faster after we prodded them. So on the software side it is probably 50 percent faster than it was a year ago on heavy JavaScript, if you use Android Chrome like you should be, it is the default and all.


#334

More data for you, @chappers.

Being that this is a Google employee, the lede is heavily buried here. Saying the “average” phone is a Moto G4, when Ye Olde iPhone 6 from 2014 clobbers it in perf, is far from the complete story. Take a peek:

Let’s tighten up a bit there and see what we can find in the good area…

The good news is that Pixel 2 is getting reasonable (now that Google Chrome / Android has fixed perf in latest Chrome releases), so that’s very encouraging, that implies the 835 is in the ballpark of an iPhone 6s which is an OK, if not great, place to be. The rest of the stuff in the pink box is slower than an iPhone 6, more around iPhone 5s level. Which is … not OK.

What about a real-world site, like CNN.com?

On the high-end iPhone 8 it takes just ~4s to parse/compile CNN’s JS compared to ~13s for an average phone (Moto G4). This can significantly impact how quickly a user can fully interact with this site.

Average phone, my ass! Unless the argument is that the “average” person carries around a four year old iPhone 5s. And you can probably get a used iPhone 6, which would be dramatically faster, literally 2x faster, for less than the Moto G4.

The lede is that Qualcomm fuckin’ sucks balls and mobile performance should currently be 3x what it is on their crap-tastic chips. 🖕


#335

I wouldn’t find that a surprising result actually.


#336

Especially if you are including international markets


#337

Well yeah if you are building for the third world, sure. But for traditional US/europe websites…


#338

I did until last month! On the other hand, I am a decidedly above average person.


#339

Even in the US I’d suspect the average phone is way worse than what you would consider acceptable. Lots of folks have garbage phones.


#340

{{citation needed}} on this one. Give me actual data.