Tuesday, January 7, 2020

The Snapdragon 865 will make phones worse in 2020, thanks to mandatory 5G

Qualcomm is so obsessed with 5G, it's hurting 4G performance to spur adoption.

Qualcomm Snapdragon 865 Flagship Processor
Qualcomm recently took the wraps off its flagship SoC for 2020, the Snapdragon 865. As usual, we can expect this chip in all the high-end Android smartphones in 2020, and it's 25 percent faster than last year, with fancy new camera features and AI-accelerating co-processors. What's unusual is the way Qualcomm designed the LTE modem in the Snapdragon 865: it doesn't have one.

This means nearly every flagship Android phone will be a 5G phone in 2020, and putting the 5G and 4G on a giant extra chip means smartphones are going to use way more power, no matter which cell network you're connected to. When 5G networks are only going to be in their infancy in 2020, this sounds like an across-the-board downgrade to me.

In 2019, 4G had a big power and size advantage over 5G thanks to the all-in-one SoC with an integrated modem solution. In 2020, Qualcomm is so desperate to make 5G a thing that it's making 4G worse.

More power-hungry modems for everyone

2019's Snapdragon 855 offers 4G connectivity in a single, simple package. 2020's Snapdragon 865 has no onboard modem, and it needs an extra chip.
We've spent the past year railing against early 5G hardware from Qualcomm because 1) the networks it supports barely exist and 2) the hardware requires significant compromise in your smartphone design, even if you never use 5G. The whole reason a modern smartphone works so well is due to the SoC, the System on a Chip.

This combines every major computing component onto a single chip, which you can see in the diagram above. There's a CPU, GPU, an "ISP" for camera functionality, a Qualcomm "Hexagon" co-processor, Wi-Fi, Bluetooth, and—in every flagship SoC released in the last seven years except for the Snapdragon 865—an onboard cellular modem. Mashing everything into a single chip saves power, and it saves space, which means room for a bigger battery.

We've already seen what happens when Android phones ship with separate modems. The first 4G phones with separate modems, like the HTC Thunderbolt, were legendary disasters. The Thunderbolt was a hot, slow, buggy mess, and it had a battery life of only a few hours. It was so bad that one HTC employee even apologized for releasing it. Qualcomm's 2019 5G package was the Snapdragon 855 with a separate X50 modem, and those were fireballs, too.

PCMag's Sascha Segan has been on the road testing 5G networks, and he wrote that heat was a constant issue:
On a hot Las Vegas morning, my two Galaxy S10 5G phones kept overheating and dropping to 4G. This behavior is happening with all of the millimeter-wave, first-generation Qualcomm X50-based phones when temperatures hit or exceed 85 degrees. We saw it with T-Mobile in New York, with Verizon in Providence, and now with AT&T in Las Vegas. It's happened on Samsung and LG phones, with Samsung, Ericsson, and Nokia network hardware.
While this has so far only been an issue with 5G phones, the Snapdragon's 865 could make it an issue with all flagship phones, even when using 4G LTE.

We can let Qualcomm explain the disadvantages of a separate modem in its own words, if we go back to this 2012 press release. That's when the company talks up the single-chip LTE solution it debuted in the Snapdragon S4. "Typically, the more chips that are involved in building a device, the more challenging it is to conserve battery life while maintaining performance," Qualcomm wrote then. "Consolidation means good things for your battery."

Qualcomm threw all that out the window with the Snapdragon 865. At least with the 2019 5G design—with its onboard 4G modem and the extra 5G chip—you could turn off the extra 5G chip and just use the SoC's onboard modem with the usual power savings. Now, with no onboard modem at all, even in 4G-only mode you'll be using more power by needing to light up that second chip.
Smartphone manufacturers have spent the past few years aggressively protecting the space used by components. How many manufacturers have we seen remove headphone jacks, telling us they need the space to simplify the design and free up more space for batteries? And a 3.5mm headphone jack is useful. When 5G networks remain so limited, who would want to waste so much extra internal space on 5G hardware that might never be used?

So far, the industry's response to 5G's bigger space requirements has been to make smartphones even larger. Consider the Galaxy S10 as a data point: you can get 4G versions like the 6.1-inch Galaxy S10 and the 6.4-inch Galaxy S10+, but if you want 5G, you will have to pick up the ultra-huge 6.7-inch Galaxy S10 5G, which is basically a new S10++ sku. For the Galaxy S11's 2020 release (which, remember, has to have 5G in every model), rumors point to across-the-board size increases for every model.

Financially, it's easy to see why Qualcomm made this decision. The company is probably going to make a ton of money on the Snapdragon 865, since every phone manufacturer using the 865 has to also purchase a separate modem. Qualcomm gets to sell double the amount of chips compared to last year! Since Qualcomm, according to the Federal Trade Commission, has a monopoly on high-end smartphone chips, no one can do much about this. Instead, the public will pony up for this design change in the form of more expensive devices.

We already have some data points for what the increased size, additional chips, and greater complexity mean for the cost of 5G phones. A year ago, OnePlus CEO Pete Lau told The Verge that 5G would raise the price of phones by $200 to $300, and that's about what we've seen in the market. The 5G version of the OnePlus 7T Pro costs $900 in the United States, while a similarly configured 4G OnePlus 7 Pro was $750. A Galaxy S10+ is $1,000, while a Galaxy S10 5G is $1,300. Plus, don't forget some carriers will charge extra every month for 5G—on Verizon, your bill will increase $10 a month.

Pick your poison: Sub-6GHz or mmWave 5G

5G will be a requirement for any phone with a Snapdragon 865, but there are two kinds of 5G that manufacturers can choose to implement.

So first—some definitions.

"5G" with no other qualifiers doesn't mean a lot because, as in the early days of 4G, the cell carriers are shameless about misleading the general public and will label anything with the "5G" stamp if they think it will help sales. The biggest abuser is AT&T, which one day just decided to label its entire 4G LTE network as "5G," with no technical upgrades made.

In the United States, honest uses of "5G" over the last year have mostly meant "mmWave," which uses a big chunk of higher-frequency spectrum (24GHz and above) for big speed increases. This spectrum is available because it's actually not that great at hosting a radio signal, and it comes at a cost of greatly reduced range, reduced signal penetration, and more complicated, power-hungry mmWave hardware. Implementing mmWave is a technical nightmare. When corporations talk about how 5G will revolutionize the way you watch cat videos or whatever, they're talking about mmWave.
mmWave runs from 24GHz up to 100GHz, while 4G and sub-6GHz 5G are below 6GHz.

The other kind of 5G is "sub-6GHz" 5G, which lives in the lower bands and offers only modest speed increases. In the US, carriers have started rolling out sub-6GHz, but the reason they focus so much on mmWave is because that 24GHz-and-above frequency is available in the US, while sub-6GHz frequencies are quite limited. The implementation of sub-6GHz is considerably more doable than mmWave, with less extra handset hardware and a longer range. International mentions of "5G" in 2019 almost always meant "sub-6GHz." Other countries have auctioned off sub-6GHz frequencies for 5G, but the US has not.

Verizon's first public mmWave speed test hit an outrageously fast 760Mbps down. This was done under ideal conditions—right next to the mmWave tower and with no other traffic to contend with, but it's still a staggering speed for a mobile network. Sub-6GHz 5G was recently launched by T-Mobile, and the results are markedly less dramatic—the company says to expect 20 percent speed increases. PCMag's hands-on testing of T-Mobile's sub-6GHz 5G peaked at 297Mbps—but in some locations it dropped to 51Mbps, slower than 4G networks in the area.

Back to Qualcomm's chip packaging decisions. Snapdragon 865 phones will all support some form of 5G, but which one will be used is up to the manufacturer. The mandatory X55 modem supports both sub-6GHz and mmWave, so it sounds like all phones will at least come with sub-6GHz support. If you want mmWave, that takes more work, and it will be something you have to design the entire phone around. MmWave penetration is so poor that your hand can block the signal. Qualcomm's solution to this problem is to buy more chips from Qualcomm, in the form of multiple QTM525 mmWave antenna modules, which can be placed all around the phone. By putting these mmWave antennas into the phone, allowing for reception at the top, bottom, and sides of the device, you can hopefully prevent the "you're holding it wrong" scenario, allowing for 5G reception in any orientation. Again, though, this just contributes to 5G's big problems: these antenna modules are sizable chips, and you'll have an even bigger, more expensive, more complicated phone.

Qualcomm is rushing 5G, just like it rushed 64-bit

Assuming the Snapdragon 865 turns out to be as big of a regression in practice as it looks now on paper, this won't be the first time Qualcomm has totally borked the smartphone upgrade cycle for the entire year. Toward the end of 2013, Apple unveiled the world's first 64-bit smartphone, the iPhone 5S, and it wasn't running a Qualcomm chip. Apple completely shocked Qualcomm by beating it to a 64-bit mobile SoC—one report even quoted a Qualcomm insider as saying it set off a "panic" inside the company. Qualcomm set about rushing a 64-bit flagship SoC out the door, and just over a year later, the Snapdragon 810 was born.

The Snapdragon 810 is infamous for being Qualcomm's worst yearly upgrade ever. The fast turnaround time resulted in a chip that was extremely hot. It throttled constantly, to the point that it was hard to see the SoC ever hit its 2GHz rating. When it was warm, it turned in benchmark scores that were lower than the previous year's Snapdragon 801, and the battery life was terrible. Samsung actually dumped the chip from the Galaxy S6, citing overheating concerns, and shipped it worldwide with its own Exynos SoCs instead. No one else really got out of 2015 unscathed, and it was a bad year for Android phones.

That Qualcomm insider mentioned above even admitted there was no technical reason to be in such a rush with 64-bit. "It's not that big a performance difference right now, since most current software won’t benefit," the insider said. "But in Spinal Tap terms, it's like '32 more,' and now everyone wants it."

This is the way Qualcomm thinks about designing and marketing its SoCs. "Make the best chip you possibly can" doesn't always look like the guiding criterion. "What can we sell as the next big upgrade?" often seems like a higher priority. The company has openly admitted to this line of thinking in the past, saying it started making eight-core chips not because they were technically useful but because customers in China liked having the bigger number on spec sheets.
This is the Qualcomm I think we're getting in 2020—a Qualcomm where marketing concerns force a feature out the door before it's ready, and every flagship smartphone suffers because of it. The move to separate modems will bring bigger, hotter, more expensive hardware, all for the sake of getting 5G in your smartphone, and you'll be lucky to actually connect to a 5G network.

5G networks aren't ready for primetime. MmWave is sounding more and more like it's not a nationwide answer for 5G coverage. The limited range, low penetration, and need for dense tower distribution led T-Mobile's CTO to declare that mmWave would "never materially scale beyond small pockets of 5G hotspots in dense urban environments." For companies like Verizon, which are trying to roll out mmWave, their own coverage maps reveal that mmWave is limited to specific addresses, rather than whole cities or sections of cities. US carriers are starting to turn to sub-6GHz for a more scalable form of 5G, but again, this lacks the huge speed increase of mmWave and (in the US, at least) has limited spectrum.

Android phones are not long-term investments—manufacturers typically only support phones with major updates for one or two years, after which you're expected to buy a new one. With such a short lifespan, investing in a 5G device when 5G networks are not immediately available in your area doesn't make a ton of sense. You'll be upgrading again in a year or two anyway, so just hold off on the expense of 5G for a bit longer. Just as I didn't think anyone should buy a 5G phone in 2019, I doubt 2020 is the year of 5G, either. The carriers still need to get their networks together, and by 2021 we should be able to get a flagship device with an integrated, single-chip 5G SoC. The big problem is that Qualcomm turned "Don't buy a 5G phone in 2020" into "Don't buy any flagship Android phone in 2020."

Forget 2020, wait for 5G in 2021

A bad flagship chip from Qualcomm doesn't leave a lot of other options for high-end smartphones in 2020. The one ray of hope for 5G phones is the Snapdragon 765G, which was also introduced last week. This is Qualcomm's first-ever chip with a 5G modem integrated into the SoC, and it should do plenty for 5G's battery consumption.
The Snapdragon 765G block diagram. Note the juicy "connectivity" block at the top, denoting the integrated 4G/5G modem.
One downside to getting the 765, though, is that it's slower than a Snapdragon 865. Another is the limited phone selection of this chip. For instance, if you wanted a Galaxy S11, you probably won't be able to get one with a 765. We actually have no idea what the distribution of this chip will be like. In the past, the 7-series chips have been hard to find in US devices, and they have been more popular in China and India. The good news is that the 765 shows us that Qualcomm is working on integrated 5G; integrated 5G is just not ready for flagships this year. This first chip is a good signal that 2021 will see a flagship SoC with integrated 5G. Until then, things could be rough.

If you really need a new smartphone, may I suggest one of the fine 2019 devices that are still available for purchase? The best Android phones of 2019 are easily the OnePlus 7 Pro and 7T, and the 7 Pro can be had for $150 off, making it $550. If you need a good camera, the Pixel 3a has one of the best out there, and it costs just $400.

As for 2020, it's not looking like a great year for smartphones. We still have to wait for actual devices to come out to confirm all of this, but I'm not optimistic. We'll have larger, more expensive devices with worse battery life, and none of those things should make you want to "upgrade."

Source: Arstechnica

Whatsapp Button works on Mobile Device only

Start typing and press Enter to search