Apple to Use Intel Modems in AT&T iPhone 7, Qualcomm Modems for Verizon and China

So, this article is leading me to believe ATT will be an inferior product as opposed to the Verizon version. Is this what we are to take from it?

Yes. And this will mean that people will use anecdotal evidence to "prove" that there's a big difference between the two. In reality, the performance difference will probably be negligible, especially given the fact that speed is mainly limited by the tower capacity than the modem.

Oh man... This will be Samsung or TSMC all over again in the forums

I guess no modem for T-mobile.

You sure on that? I'd agree with you if people were getting moderately high speeds. We're talking about 450 mbps vs 600 mbps LTE modems. I haven't heard of any regions getting anywhere near that.

Yes, if you're nearing the peak (say, AT&T is delivering you 400), you might get better results with the 600, because you rarely get a perfect signal. But currently, the 2016 4G average is 9.9 Mbps.

Even if you're in a region getting 50 mbps peak, I don't think you're going to be affected. By the time we start hitting 200 mbps peaks and this might start to matter, the phone will likely be obsolete.

It's like debating whether 802.11n or 802.11ac will get you better speeds out of your 10 mbps internet connection. (Ignoring the range differences between those standards.)

Yes, I am 100% sure about that. Let me explain.

When it comes to LTE modems, the advertised peak speed is essentially short-hand for a bundle of underlying technologies.

For example, an LTE modem that supports 450 Mbps peak download speed (Cat 9) means that it features:

* Aggregating three LTE connections simultaneously
* Receiving data on two antennas simultaneously
* Enough signal processing horsepower to decode a max of 6 bits out of information out of every LTE transmission received from the tower.

On the other hand, a modem that support 600 Mbps peak download speed (Cat 12) means that in addition to all of the above, it can decode 8 bits out of every LTE transmission from the tower. It also supports features like receiving data on 4 antennas simultaneously instead of only 2.

Now, why do you as the user care?

Because receiving data on 3 connections simultaneously is faster than receiving on only one connection. The typical LTE smartphone has peak 150 Mbps LTE download speed, which is possible with only one LTE connection. Let's say that out of the 150 peak, you're getting 9.9 Mbps in the real world. Well, now let's say you have a modem with 450 Mbps peak. That 9.9 could become ~ 30 Mbps. Triple the actual real-world speed. And if you had a modem that went further by supporting the more sophisticated signal processing, that real-world speed gets an addition 33% boost, going from 30 Mbps to 40 Mbps.

So will you ever achieve those peak theoretical speeds? No. But what you will really get is the *relative gain*. That's the important thing here. These features are all speed multipliers, independent of what the absolute value of the speeds you're getting.

Here's a video that shows the effect of carrier aggregation - going from 110 Mbps to 220 Mbps peak on Sprint's network. Did the phone with carrier aggregation actually get 220 Mbps? No. But it did get twice the speed of the phone that doesn't support carrier aggregation.
[MEDIA=youtube]jJmpdMKpo4o[/MEDIA]

And yes indeed, these feature are actually really launched in other networks around the world. Australia, South Korea, Japan...and they will indeed be launched in the US over the next year. How long does the average user keep their phone? Having these features built into the phone means that a year after purchase the phone gets better as the operators turn on those features in their networks. How many other technologies inside a phone get better with age instead of worse?

So where does that leave unlocked phones......uter confusion?

So, this article is leading me to believe ATT will be an inferior product as opposed to the Verizon version. Is this what we are to take from it?

What's the point of faster LTE when it just means your data cap runs out sooner?

500 web pages is still the same amount of data at faster LTE or at 3G speeds.

If instead of working more efficiently and just being done faster, you choose look at another 500 web pages then yes, you will use more data.

So, this article is leading me to believe ATT will be an inferior product as opposed to the Verizon version. Is this what we are to take from it?

Perhaps, but not necessarily.
There are two aspects to "better modem".

- The less interesting aspect is the high end feature support. If the modem supports a bunch of features that have not yet been rolled out by US (in particular AT&T) carriers, and that are not scheduled to be rolled out for two or more years, then their absence is not really a big deal for most people. Yes, if you plan to use your phone in South Korea (or whatever wonderful place has rolled out these leading edge features) you might care. Yes, if you plan to hold onto your iPhone for five years you might care [but people who care about cellular speeds and specs DON'T hold onto their phones for five years...].

- The more interesting aspect is that there are significantly different algorithms that can be applied to the problem of extracting the desired digital signal from the (very noisy) analog radio signal that is received. These algorithms differ in how much hardware they utilize, how much power they utilize, how long they take to run, how complicated they are to encode in a chip, etc etc. QC has been in the business a long time, and what *I* mean when I say that QC has good modems is that their implementations of these algorithms are remarkably good.

The question I would consider far more important than what specs the Intel modem supports is the quality of their decoding algorithms. The generous possibility would be to say that Intel is well aware of how important this is, and has thrown a vast number of resources at creating algorithms and implementations every bit as good as, or better than, QC. The cynical possibility would be to say that it took Intel years to implement the best known algorithms in various aspects of their CPUs and memory controllers, and that sort of cheap "do the minimal possible job you can get away with" attitude is baked into the company's DNA. Certainly the stream of rumors over the past few years regarding Intel's modem business is substantially more of the "OMG, what a cluster****" variety than of the "Damn, that's impressive" variety.

My suspicion is that the Intel modems ARE subpar (though not enough subpar for it to matter to most users) and that Apple is taking a hit here for a larger goal; that larger goal being a substantially closer relationship with Intel that allows Apple to specify what they want in future modems, and even to license the modems for direct inclusion on future A# SoCs. (So essentially to create as close a relationship with their modem vendor as Apple has with Imagination and the PowerVR GPU. QC seems unlikely to grant that, whereas Intel is desperate enough to be a lot more flexible.)

Meh, just keep the headphone jack this time.

Oh man... This will be Samsung or TSMC all over again in the forums

Yes. And this will mean that people will use anecdotal evidence to "prove" that there's a big difference between the two. In reality, the performance difference will probably be negligible, especially given the fact that speed is mainly limited by the tower capacity than the modem.

Sorry but that is not entirely true. Even under completely identical conditions in a controlled environment, performance varies from modem to modem for various reasons. The difference can be up to 20-30%, again under completely identical network conditions (e.g. both modems assigned the same amount of resources from the network).

You can read about some modem vs. modem tests here:
https://www.qualcomm.com/news/snapdragon/2016/05/11/modem-vs-modem-recap-4-reasons-why-snapdragon-modems-are-superior