Bleep Qualcomm right in their Qualcomm-hole

[Timex]

Ya, I’m totally aware of all the awesome image improvement stuff that software lets you do, and it’s obvious that better processors let you do it faster.

What I’m talking about when I diferentiate “real time” is a situation where you actually have a dynamic light sensor, and thus the software is able to look at the image, adjust the camera to gather more data, and then snap more pictures. This is a thing which exists, but as far as I know doesn’t occur with cameras on phones, since they are totally static.

For the cameras, the reason I said you just snap the picture, and then stuff occurs, is that technically it doesn’t matter if that processing takes 1 milisecond, or 1 hour… the end result will be the same.

but if the photographer can see the effect of this “post-processing” on their view screen and therefore adjusts something like their location or camera settings, than overall, big win in the end result. And that win comes from desktop class processors in phone sized devices.

This makes sense, in that it improves usability of the device by allowing the user to see the final result.

@Ephraim

And in the case of the portrait lightning, it needs to do 3D image mapping at the time of the picture taken, so that’s a “real time” decision to engage those sensors on the camera, or not. I don’t think of it so much as the sensors doing something different, more a question of whether or not they are used at all, and whether or not the data they capture, if used, is attached to the photo for later “post processing”.

I guess that’s a possible thing, in that the dynamic hardware loop ends up being a modification of which sensors to use… although, in such a case, I’d suspect that you could trivially just activate all of them without much negative impact, all the time, and then just use the data if determined to be necessary later.

[Ephraim]

@Timex

without much negative impact, all the time, and then just use the data if determined to be necessary later.

True, I do wonder about this. It must have some negative impact, or why wouldn’t Apple do it all the time? I suspect it must bloat file size, drain the battery more, or slow the capture in some way.

[Timex]

Well, in terms of permanent storage, I’m sure it’d bloat the size… although the flip side is that I’d have to imagine that storage used by photos on your phone is a pretty tiny slice of the overall storage.

When it’s taking place, it would increase the memory requirements, but again, I wouldn’t think that it’d increase them to the degree that it’d actually have a measurable impact on anything.

Honestly, are you sure they aren’t just doing it all the time?

[Ephraim]

@Timex

Honestly, are you sure they aren’t just doing it all the time?

Sure. Some research (I don’t have a phone that can do this) shows you actually have to enter into “Portrait Lighting Mode”, just like you need to switch to “Slo Mo” if you want to capture at 240fps.

[Timex]

The slow motion capture would result in significant increases of storage used
… not sure about the portrait lighting thing.

[wumpus]

@Thraeg

But I find the resolution to be unacceptably low for a device I’m spending close to a grand on in 2017, regardless of great lab measurements in other categories.

I’m super nearsighted and I often use my devices with my glasses off, with my face basically inches from the screen. So you are arguing there is literally no point of diminishing returns on PPI, we need 800+ PPI displays because that’s necessary. The plus devices I use have 401 PPI and even with my face super wedged up in there because I am blind as a fucking bat without my glasses, I really can’t see pixels.

I 100% agree that

• for VR you need 800+ PPI because you’re magnifying the screen from directly in front of your eye, literally strapped to your face.

• a brand new OLED screen is going to outclass the best LCDs, but it took a few years for OLEDs to get to where they needed to be.

Also hey look Google Pixel, those prices sure do look similar to the iPhone, except for 1/2 to 1/3 the internal SoC performance.

@stusser

It doesn’t actually matter that the A11 is much faster than the snapdragon 835, because it doesn’t noticeably impact the user experience.

In fact, it does, on the web. Let’s look at some webpage load times for CNN:

The bar is length is 148px for the A11 and 234px for the Snapdragon 835. That is 1.6 times slower. On the CNN website. God help you if you have something older than the 835, by the way, it’s going to be substantially worse. And the more JavaScript is executing, the more that ratio tilts as well; Discourse is an extreme example, but it’s also the future.

@mono

All that said, I’m all for going after Qualcomm. The point that you could do more with these phones if they weren’t programmed to accommodate the lagging Qualcomm chips is well received.

Yes, this. The whole thing is an affront to the history of computing, where we expect devices to get faster and cheaper every year – and smartphones are the last exciting frontier where we actually get huge gains every year. Qualcomm is taking a giant stinky monopoly style dump on what was our last fun hurrah party in computing:

Mainframes → Minicomputers → Microcomputers → Laptops → Smartphones

(I guess maybe watches could be next? Hard to get excited about watches for me personally, though.)

Yes, there are clearly points of diminishing return, as we’ve seen on desktop, and it is likely Apple is starting to hit them. But Qualcomm is so far behind, and Android is in trillions of devices. We need more competition here, something that can actually innovate and keep up on the CPU side.

[wumpus]

Great, so Qualcomm are also patent trolls. As if I needed another reason to hate their guts.

Qualcomm prefers to be viewed as an innovative leader in wireless technology, but in actuality, royalties from legacy patents dating from the 1990s are what drive a large share of its profits. While its manufacturing business contributes most of its revenues, patent licensing accounted for $6.5 billion of its pre-tax profit in 2016, compared to $1.8 billion derived from its chip business.

…

For example, past investigations have found Qualcomm has withheld licenses from phone manufacturers that use chip sets from manufacturers other than Qualcomm. U.S. chip makers such as Broadcom, Marvell, Freescale and Nvidea have all exited the wireless market since 2014, decisions the FTC suit claims stem from Qualcomm illegally exploiting its patent monopoly.

[Chappers]

@wumpus

bar is length is 148px for the A11 and 234px for the Snapdragon 835. That is 1.6 times slower. On the CNN website.

But to address the previous point, that is a difference of 90ms which I doubt is noticeable in a real world scenario by the average user. I’m all for better and quicker alternatives to Qualcomm but it is certainly not my priority when choosing my next phone.

[wumpus]

It’s far more than 90ms. Particularly if you don’t adblock, you are getting a ton of JS advertising code on every webpage. The long form version of this is that plain old boring static HTML is no longer representative of the modern web – every time you load a “web page” you are in fact executing a lot of JavaScript.

With Discourse, admittedly an extreme and future oriented example just chock full of nuts JavaScript, the difference between a Nexus 6p and a Nexus 7 to “load a web page” and render a topic is 248ms vs 636ms. And a fast desktop can do it in ~40ms.

Guess which number is closer to iPhone 7 / 8 results?

[Chappers]

Ok wumpus, let’s say the difference is 500ms on a JavaScript heavy page, again I don’t think the average user really can tell a difference that is meaningful enough to them. I think that’s what many of us getting at in response to all your charts and bar graphs. As I said, I agree with your general thoughts in improving this performance, I just don’t feel it is as important as you imply. For me personally, decent performance (which a 835 provides), good screen and most importantly battery life is what floats my boat.

[wumpus]

That’s just one place; what about photo processing? Launching apps? Opening attachments? Literally everything you do on the device is faster when the CPU is faster.

For example, per Tom’s Hardware:

Also

The iPhone 8 also opened more demanding apps faster than its predecessor and the top Android phone right now, although those differences were less dramatic. It took the iPhone 8 11 seconds to fully load the Injustice 2 game, compared to 14.53 seconds for the iPhone 7 Plus and 19 seconds for the Note 8.

So you’re saying users wouldn’t notice any of this because fuck performance who cares?

[Chappers]

No performance is definitely an issue and it would be great if Android phones were as quick as iPhones but in real world scenarios it is fast enough to do what most users need it to do. Admittedly, opening a PDF and being 6x slower is not good but video editing is not a concern to me and probably not to 98% of users. Items such as expandable storage, better battery life, a headphone jack are all things that I would prefer on Android over iPhone despite the performance differential.

[CharlesC]

Man in the world of web development 500ms is an effing eternity. No way a user doesn’t notice that.

[arrendek]

Response Time Limits: Article by Jakob Nielsen

Excerpt from Chapter 5 in my book Usability Engineering, from 1993:
The basic advice regarding response times has been about the same for thirty years [Miller 1968; Card et al. 1991]:

• 0.1 second is about the limit for having the user feel that the system is reacting instantaneously, meaning that no special feedback is necessary except to display the result.
• 1.0 second is about the limit for the user’s flow of thought to stay uninterrupted, even though the user will notice the delay. Normally, no special feedback is necessary during delays of more than 0.1 but less than 1.0 second, but the user does lose the feeling of operating directly on the data.
• 10 seconds is about the limit for keeping the user’s attention focused on the dialogue. For longer delays, users will want to perform other tasks while waiting for the computer to finish, so they should be given feedback indicating when the computer expects to be done. Feedback during the delay is especially important if the response time is likely to be highly variable, since users will then not know what to expect.

I usually tell my users during requirements discussions/fellow developers half that. 500 ms and the user notices it, 5 seconds and you might as well tell them to go somewhere else.

[Gordon_Bleu]

Amazon made the discovery that with every 100ms increase in web page load time, sales decrease by 1%.

[stusser]

500ms total page load time is fine. 500ms multiplied by a number of loading components is deadly, as is 500ms response time to dynamic page elements.

[wumpus]

It is important to realize that everything will take at least twice as long, often 3x as long or more. Everything. And that’s assuming you have the latest 835.

It is not a subtle or small difference; the performance gap has been widening for a few years now.

[wumpus]

Also apologies to @Thraeg if I came across as personal attacks, that is not my intent. I am pissed at Qualcomm, not any individual person here. This is a topic I care intensely about and sometimes, like the Pointer Sisters, I get a little too excited.

[Thraeg]

@wumpus

So you are arguing there is literally no point of diminishing returns on PPI, we need 800+ PPI displays because that’s necessary. The plus devices I use have 401 PPI and even with my face super wedged up in there because I am blind as a fucking bat without my glasses, I really can’t see pixels.

No, that’s a good point. We were comparing the S8 vs. the iPhone 8 non-plus, at 326 ppi, but the plus at 401 is noticeably sharper to the point that I don’t really notice the individual pixels. I’d be fine with a screen like that (though as mentioned, upgrading to Plus tilts the price comparison further toward Android).

@wumpus

Also apologies to @Thraeg if I came across as personal attacks, that is not my intent. I am pissed at Qualcomm, not any individual person here.

No worries, and sorry if I got overly snarky toward you as well.

I actually agree with your main takeaway (faster phones are nice, and it sure would be great if Qualcomm improved their performance and/or there were more viable chip manufacturers in the Android space). I wish you luck in pushing things in that direction, even if I don’t feel strongly enough about it to accept substantial tradeoffs and drawbacks to get a faster phone today.

[CraigM]

@Chappers

Admittedly, opening a PDF and being 6x slower is not good but video editing is not a concern to me and probably not to 98% of users. It

You might be slightly underestimating those factors. For me PDF performance is… non trivial. I do quite a bit with that due to work. The ability to process photos faster is also big, as a father. Especially with the advent of Live Photos, which are basically audioless short videos. I take significant photo amounts, and love that feature. Being able to speed and improve photo quality, while enabling those gains on, essentially, videos? That is huge for me.

And probably more people than you realize.