
As they say....
"Imitation is the finest form of flattery"
The smartphone chip market it dominates is nevertheless an increasingly challenging one for Qualcomm. Slowing growth, litigation and rising Chinese competition are among its problems, but also the trend for the largest, richest handset makers to design their own chips. Apple has its own processors, and given its current legal …
Actually, it's the design of Android that gets in the way of timely updates. True, Qualcomm don't always release blobs to ODMs for older chipsets, but then they have no incentive to - and such costs will always have to be paid by the downstream customer anyway.
Google are beginning to put the pieces in place so that Android handsets can be effectively updated without the code first going through the chipset and Original Device manufacturers; if it were that easy, Google would have done so years ago because *they* do have a strong incentive to do so.
@Dave 126,
"Google are beginning to put the pieces in place so that Android handsets can be effectively updated without the code first going through the chipset and Original Device manufacturers; if it were that easy, Google would have done so years ago because *they* do have a strong incentive to do so."
Yep, Project Treble, a well overdue and welcome development. They're turning Linux into a microkernel. I can't see Linus liking that!
In a way it's an unspoken acknowledgement on Google's part that they seriously cocked up the entire roadmap for Android from the very beginning. It's like they chose Linux "coz it's free and cool", without even beginning to think about the possible consequences. And here we are, years later, with appalling and, frankly, embarassing update problems. No one else in the history of anything even remotely related to computers has ever to build such a large market with such a crap software upgrade path.
Project Treble might start solving some of these issues. What'll be interesting is to see how this plays with the wider Linux development roadmap. It's another split away from the kernel mainstream (bad), but they'll be ending up with a kernel with fewer hardware dependencies (good).
In a way the Linux world needs Linux to go the microkernel route too. This will allow things like WiFi and graphics drivers can be done by hardware manufacturers jealously guarding their IP or not wanting to have an enourmous team dedicated to driver updates every time Linus and chums change anything. However I can't see the Linux kernel development community, or Linux himself, being particularly keen on the idea.
That's not correct - Android in its entirety is a copy of every single aspect of iPhone down to its recent "flat design" change that suddenly occurred overnight. Android started out life as a blackberry clone until the initial iphone keynote - so anything beyond a keyboard and a small 240x360 resolution screen is ALL BASED on an iphone. But android is a solid $29 worth of phone as long as you're not paying more than that.
Another BB10 diehard here...
If Google are so keen on the microkernel idea (Project Treble), why don't they just buy BlackBerry with a bit of loose change and actually get hold of a proper, solid and highly respected QNX microkernel OS?
If literally every manufacturer is going to have to re-write their drivers for Treble, why not re-write them for an actual microkernel OS instead, make a complete break with Linux? The rest of the task is, in comparison, fairly straightforward; it's another POSIX compliant unix-like OS environment, recompliation of Android's userland doesn't seem too great a task, especially as Google are also in control of their own application development environment.
This suggests that Google will pursue a similar strategy to the one it has adopted for its cloud servers, commissioning custom designs from its chip suppliers
If the article would perhaps focus on how differently Google operates in the server and consumer space then it might be quite interesting. As it is, to suggest that Google is suddenly going to flip its business model, in which the own-branded devices serve largely as PR for the latest Android release, flies in the face of all the previous suggestions that this was about to happen.
A handful of SoC designers would let them design a custom SoC using standard ARM cores, like Apple did with the A4 and A5. They'd need a whole CPU design team if they wanted to do custom cores.
I don't believe this is for mobile. Their volumes are too low, even if Pixel becomes more successful they can't justify the cost. Maybe if they wanted to add some custom hardware to an otherwise stock SoC to give future Pixels an advantage not just in software but also hardware over the high end Android competition (i.e. Samsung) but that's a risky strategy. They can only upset Android OEMs so much before they start considering options that hurt Google's bottom line a lot more than Pixel helps it. Samsung has already tried to go its own way with S Voice, Samsung Pay and so forth. They could make a deal with Microsoft for the missing pieces like search and maps, and drop Google entirely like Chinese Android OEMs already have.
There are two categories where they might want a custom SoC. One, the watch/glasses wearables market. No one is making SoCs custom designed and targeted at that market, so if they think there's something there for them, they'll need their own.
Two, for their servers. They make those more and more custom, so maybe this is an SoC with some special support for search optimization, or neural networks to work alongside x86 CPUs and perhaps eventually replace them. They won't care if ARM cores are slower than x86 cores if most of their code eventually runs on DSPs or specialty ASIC blocks that are far faster than either for their particular needs.
Why would Google want to hundreds of millions developing their own SoC, just to hand it to others? Why an OEM want to become even MORE dependent on Google, sacrificing even MORE ability to differentiate from the competition, by choosing this over similar (or better, if Qualcomm, Samsung etc. can continue designing better cores than ARM standard ones) offerings? What's the value add of Google's SoC over Snapdragon, especially since Google's would require they buy their cellular baseband etc. separately?
I still think Google's SoC efforts will not be targeted at smartphones.
You;re absolutely right but Google shareholders don;t mind paying BILLIONS for blimps and solar panels when Google has NEVER HAD ONE SUCCESSFUL consumer product - they are an ad sales company - a GREAT AD SALES company with delusions of being called a genius like Steve Jobs, E.on musk or even Jeff Bezos. Yes, Google is plowing billions into grocery deliver.
I sincerely hope they get their act together before they attempt to launch any more Pixel phones. The supply chain for the first Pixel has been horrifically bad. It wasn't available in most markets, and the countries that it was available in seemed to suffer from endless 'out of stock' messages on the store.
Google shareholders don't ind Google wasting their money. If t took Apple years and spending at least $2-5 B to get it up and running - of course, Apple needs about 300 MILLION chips a year ... now that R&D is amortized to a few dollars per chip. Google will spend nearly as much as Apple for their devices that sell about ONE MILLION a year (that's 1/300th the scale of business - yes, just pointing out Google doesn't do that math). So, like Google wasted $12 BILLION dollars in 2 years buying Moto, they will plow BILLIONS into this and after 3 years, realize no plant will charge them less than $150 PER UNIT phone sold to manufacture one million chips - they will literally waste more chips than have completed ones that pass inspection. Another BLIMP boondoggle. It took Google 8 months to sell 1-million phones - about what Apple sells EVERY 2 days.
You tend to think quite narrowly, don't you? There's no reason that a Google designed bit of silicon couldn't be used by other Android vendors, for free or for a modest fee - that would easily give them the volume.
Apple have done very well tweaking and designing silicon to work with their own OS, and their split with Imagination Technologies could well be that they want a GPU that can be efficiently used for AR and DSP duties - indeed, a recent Reg article suggested that using Apple's AR toolkit on existing iPhones depleted the battery too quickly for consumer use. This sits well with what an ARM rep (in this article) said of the power savings of AR-optimised silicon over existing mobile GPUs.
What? Google has been creating processors for a couple of years now. At Google I/O they dropped their second generation TPU where a pod is processing over 11.5 quadrillion floating point operations a second. Doing it on less than 1/30 the wattage of the next best.
Nobody else is anywhere close to having a processor like this. Nobody was close to the first generation and Google is that much more ahead.
Google hired the lead architect of the MIPS chip a few years ago and has multiple teams. Google is also going to demonstrated a 47 qubit quantum processor later this year. They have created their own networking silicon for many years.
This article does NOT accurate and does not make any sense.