Obligatory iPhone 4 Thread...

[wco81]

I'm assuming dual-core because I don't think that Apple can afford to ship a single core CPU in its new devices in 2011. IMHO Apple doesn't have to have SoCs that are a generation ahead, but they need to be at the top of the current generation because their devices have to be at least good enough for a year while the competition will crank out new and improved devices every month. And in H1/2011 we will see dual-core CPUs in tablets and high-end smartphones from other manufacturers for sure.

I don't know, in their history, they've shipped inferior hardware before, though probably not since they started using Intel.

Issue is whether they can get good enough performance, with same or better battery life than the competition, if the competition is at dual-core and Apple devices remain single-core next year.

Some of the early impressions of Tegra2 tablets with Android suggest that the software isn't taking advantage of the faster CPU. Maybe that will change completely with 2.2.

But Apple devices don't have to be the fastest or have the most megapixels in the camera, etc. They just have to be perceived as smoothly-performing and can do enough "multitasking" to make people think about something other than performance.


That said, since they intro'd a new form factor this year, they probably will shoot for improving the internals next year.

 

[Mike11]

I don't know, in their history, they've shipped inferior hardware before, though probably not since they started using Intel.

AFAIK every iPhone model was at least on par with the competition in terms of internals at it's release date. In recent memory the 3GS was the fastest with it's 600MHz Cortex-A8 (I think Palm Pre was the runner up at the time with a 550MHz Cortex-A8 and a SGX530) and it looks like the iPhone 4 just gets beaten in terms of raw specs by the Samsung Galaxy S which comes out just a week earlier (assuming the A4 isn't running too much below 1GHz).

 

[wco81]

I think the pace of releases, especially by Android vendors, is going to be hard for Apple to keep up with, especially if they stick to an annual release cycle. OTOH, look at the iPhone4 preorders. Looks like the annual cycle has created an event and a lot of pent-up demand gets unleashed all at once.

HTC and others can grab the latest SOCs as they come off the line. Plus no single Android model is shipping anywhere near iPhone volumes so they don't have to wait for production to ramp up on an SOC. They can release a model and get some publicity about the incremental features, including the internals, then release another incremental upgrade 3 months later. That seems to be HTC's pace, at least for this year.

But I don't think the average consumer is going to be demanding phones with such and such clock speed or dual-cores or specific SGX chips or even a specific level of DX or Open GL support.

You have to show them new applications or capabilities. Maybe Face Time helps. But for a lot of people, it seems the shiny, sleek industrial design is a big draw. After all, that's what gets noticed by other people when you pull out your latest and greatest phone.

Beyond that, responsiveness and subjective sense of speed probably helps too. With every new release of iOS, legacy hardware performance degrades. So iPhone doesn't have to be the fastest smartphone out there. It just has to be faster than the previous iPhone (with new OS features).

 

[Lazy8s]

With the MBX+VGP Lite/ARM11 and SGX535/Cortex-A8 sets of cores evenly spanning the four years of releases of Apple mobile devices, they get two prime-time years out of each set using clock speed upgrades.

They're definitely due for new cores next year. While a dual-processor Cortex-A9 will be ready, SGX543MP (possibly in a 2, 3, or 4 core configuration?) will be cutting it close for availability.

 

[Mike11]

With the MBX+VGP Lite/ARM11 and SGX535/Cortex-A8 sets of cores evenly spanning the four years of releases of Apple mobile devices, they get two prime-time years out of each set using clock speed upgrades.

They're definitely due for new cores next year. While a dual-processor Cortex-A9 will be ready, SGX543MP (possibly in a 2, 3, or 4 core configuration?) will be cutting it close for availability.

But, assuming Samsung's still gonna be the manufacturer, will their 32nm process be ready in time for mass production (for the first next gen Apple device, the 2011 iPad in maybe March)? If think that's the bigger question here. AFAIK some customers' 32LP tape out is supposed to start in Q3, that seems quite late (for the iPad).

 

[ltcommander.data]

http://www.appleinsider.com/article...one_4_has_512mb_ram_doubling_ipad_report.html

Leaks from the WWDC presentations report that the iPhone 4 actually has 512MB of RAM. Double the iPhone 3G S and notably double the iPad. It's also double what the leaked iPhone 4 prototype had. I guess we'll never know whether it was doubled as a response to the leak or was 512MB always the intention for production models. I didn't think that 256MB of RAM was really limiting apps, but it'll no doubt be valuable for multitasking/background tasks.

 

[wco81]

Word is that they doubled the RAM so it could run iMovie.app, which only runs on iPhone 4.

 

[Mike11]

http://www.appleinsider.com/article...one_4_has_512mb_ram_doubling_ipad_report.html

Leaks from the WWDC presentations report that the iPhone 4 actually has 512MB of RAM. Double the iPhone 3G S and notably double the iPad. It's also double what the leaked iPhone 4 prototype had. I guess we'll never know whether it was doubled as a response to the leak or was 512MB always the intention for production models. I didn't think that 256MB of RAM was really limiting apps, but it'll no doubt be valuable for multitasking/background tasks.

The more the better, especially since fast app switching and state-saving will be such an inconsistent and random experience for the user once state-saved apps get kicked out of RAM.

 

[Entropy]

The more the better, especially since fast app switching and state-saving will be such an inconsistent and random experience for the user once state-saved apps get kicked out of RAM.

For clarity, even in this case iOS4 is still better or equivalent to the typical multitasking behavior, where competing apps are free to hammer away at the subsystems even as you are trying to load the application you actually want.

Apple didn't go with the multitasking behavior they already had in place (but didn't make accessible) because they felt it wasn't good enough for this platform. What they have introduced is what they feel is a better compromise. You may not agree. But responsiveness is what they have optimized for, so if you want to piss on them, it would be wise to choose another aspect of their multitasking implementation.

 

[Mike11]

For clarity, even in this case iOS4 is still better or equivalent to the typical multitasking behavior, where competing apps are free to hammer away at the subsystems even as you are trying to load the application you actually want.

Apple didn't go with the multitasking behavior they already had in place (but didn't make accessible) because they felt it wasn't good enough for this platform. What they have introduced is what they feel is a better compromise. You may not agree. But responsiveness is what they have optimized for, so if you want to piss on them, it would be wise to choose another aspect of their multitasking implementation.

I totally agree with Apple, at least in principle. It's the execution and inconsistent user experience in 4.0 that I have problems with. And every bit of RAM helps minimize that, so I welcome the bump to 512MB. But that doesn't mean that it isn't a great and appropriate system once all the kinks are gone and all developers have properly updated their apps (and implemented full state serialization by themselves). I expect this to happen around iOS 5.

 

[wco81]

Speculation is that they added the RAM partly to run the iMovie app, which only runs on iPhone4.

 

[BRiT]

And who in the US would ever run the iMovie app on the iPad considering AT&T's 5Gb a month limit?

 

[wco81]

Actually the limit is less now.

You could do the editing but wait until Wifi is available to do the upload.

I don't know that there's this great need to have to edit and upload on the go though.

 

[ltcommander.data]

Speculation is that they added the RAM partly to run the iMovie app, which only runs on iPhone4.

iMovie HD 6 only required 256MB of RAM, so it probably would be possible for Apple to get iMovie to work on the iPad or third gen devices. Not that I'm complaining about them putting in more RAM of course.

 

[Mike11]

As usual, iFixit took it apart:
http://www.ifixit.com/Teardown/iPhone-4-Teardown/3130/1

And Chipworks' in-depth analysis of the iPhone 4's components:
http://www.chipworks.com/iPhone-4-teardown.aspx

 

[BRiT]

Seems like iPhone 4 has some reception issues depending on how you hold the phone [ http://gizmodo.com/5571171/iphone-4-loses-reception-when-you-hold-it-by-the-antenna-band ]

Speedtests while not touching iPhone, while touching with hands, and while touching inside leather case with hands:

as soon as you connect the left side with the bottom, that's when reception starts to drop.

 

[mboeller]

this could have something to do with the metallic frame of the iPhone4:

According to the c't article about the iPhone4, the 2 small "plastic" openings in the metallic frame are the areas for the antennas. Therefore if you cover this area with your hand reception could start to drop.

 

[RudeCurve]

The small openings are there to physically separate the two antennas. When you bridge this "gap" with your hand you are "shorting" them together. The two-piece steel frame are the antennas.

At 2:56 SJ describes the "gap".

 

[mboeller]

The small openings are there to physically separate the two antennas. When you bridge this "gap" with your hand you are "shorting" them together. The two-piece steel frame are the antennas.

At 2:56 SJ describes the "gap".

Thanks

 

[Entropy]

I totally agree with Apple, at least in principle. It's the execution and inconsistent user experience in 4.0 that I have problems with. And every bit of RAM helps minimize that, so I welcome the bump to 512MB. But that doesn't mean that it isn't a great and appropriate system once all the kinks are gone and all developers have properly updated their apps (and implemented full state serialization by themselves). I expect this to happen around iOS 5.

Apologies for my tone.
Even before smartphones, I always found the "preemptive multitasking is the pinnacle of OS technology" attitude of those who had read a bit on the net or passed Computer Science 101 annoying. The fact that it originated for multiuser mainframes catering to users attached at 300 - 9600 baud running text interfaces and may not have been a very good model for supercomputing, real-time computing, embedded systems, and eventually even single user systems never seems to have registered.

You'd think single user computers should be a piece of cake, but since the MC68000 Apollo workstations of the early 80s computing resources has risen by easily a factor of 1000 or more, we've had two or three decades to improve scheduling and I/O and interrupt handling, and I still get stalled occasionally sitting at my 2,8GHz four core/eight thread system. The method of personal computing to improve responsiveness has largely been to throw hardware at the problem, assuming that with fast enough hardware the time that the system gets stuck in la-la land should get short enough to be unnoticeable. While the idea obviously isn't without some merit it is a horrible, horrible approach for handheld devices where long battery life and low weight are of critical importance.

Being aware of the issues involved and trying to improve the situation by being more intelligent on the software side, rather than just filling the check box (and silencing semi-educated tech-nerds) by using what they already had working, does Apple a lot of credit. Seeing them criticized on the net for not taking the easy way out is frustrating.