Report: Google Wants To Design Its Own Smartphone Chips (arstechnica.com) 90
An anonymous reader writes: Google has been stepping up its efforts to build higher quality Android phones, and one thing holding it back is Qualcomm's SoC technology. According to two reports in The Information (paywalled: [1], [2]) Google is now looking for other partners, and may even jump into chip design itself. The company has already done some design work, hoping to co-develop it with a manufacturer. "The new chips are reportedly needed for future Android features that Google hopes to release 'in the next few years.' By designing its own chips, Google can make sure the right amount of horsepower gets assigned to all the right places and remove bottlenecks that would slow down these new features. The report specifically calls out 'virtual and augmented reality' as use cases for the new chips."
Another big area Google wants better hardware for is video processing tech. The article notes, "Qualcomm has a near monopoly on Android SoCs, but it is more marketing driven than performance driven and has been doing a disservice to the mobile space lately. It rushed to get 64-bit support out the door when it was beaten to the punch by Apple, which resulted in the very hot Snapdragon 810 SoC."
Another big area Google wants better hardware for is video processing tech. The article notes, "Qualcomm has a near monopoly on Android SoCs, but it is more marketing driven than performance driven and has been doing a disservice to the mobile space lately. It rushed to get 64-bit support out the door when it was beaten to the punch by Apple, which resulted in the very hot Snapdragon 810 SoC."
Re: (Score:2)
You don't want your CPU directly serving you ads?
Re: (Score:3)
Do you have an Android phone now? If not, no one cares about your opinion on this. If so, why on earth would you be more worried about a Google chip than a Google ENTIRE OPERATING SYSTEM?
Re: (Score:2)
Right, because it's totally as easy to change the CPU as it is to install cyanogenmod.
Re:If you think I'll allow a Google chip in my pho (Score:4, Interesting)
And it's totally as easy to implement an entire TCP/IP stack, file system, memory, and app code scanner in hardware that would be able to report information to Google without ANY OS/driver support.
Re: (Score:3)
And if you are going to post nonsense as AC, consider the option of not posting.
Re: (Score:3)
why on earth would you be more worried about a Google chip than a Google ENTIRE OPERATING SYSTEM?
Because Google publishes the code to their operating system, but they're not going to publish the netlists for their processors. The latest intel processors have TPM built into the CPU, and they even have voice recognition software built into the CPU that is still powered even when the machine is ostensibly asleep, allegedly for the purpose of waking the machine up. Is that what you want in your phone?
Re: (Score:2)
Because Google publishes the code to their operating system, but they're not going to publish the netlists for their processors
They publish the code to the basic operating system - not the drivers. You know, the drivers that control all of the hardware features of the SoCs? It's pretty irrelevant in that case whether it's hardware/firmware/driver software (a lot of the "hardware" - especially things like TPMs - these days is really microcode loaded into separate controllers on the SoC anyway).
software built into the CPU that is still powered even when the machine is ostensibly asleep, allegedly for the purpose of waking the machine up. Is that what you want in your phone?
1) "software built into the CPU?" That doesn't even make sense.
2) More efficient standby and better battery life? That sounds good to me.
Re: (Score:2)
1) "software built into the CPU?" That doesn't even make sense.
It's hardware and software. Every modern CPU has microcode onboard, but these chips go beyond that. They have a whole little computer in the processor.
2) More efficient standby and better battery life? That sounds good to me...
No, it's just so that it can wake it up with a voice command. It's not more efficient or better battery life.
Re: (Score:2)
1) "software built into the CPU?" That doesn't even make sense.
It's hardware and software. Every modern CPU has microcode onboard, but these chips go beyond that. They have a whole little computer in the processor.
And that whole little computer's software is NOT built into the CPU. It's SOFTWARE, pure and simple. The code it runs is loaded in via the firmware/drivers.
2) More efficient standby and better battery life? That sounds good to me...
No, it's just so that it can wake it up with a voice command. It's not more efficient or better battery life.
Yes, it is. Using a separate hardware component to do this means you don't need to run the power-hungry main CPU all of the time.
Google could buy Qualcomm... (Score:5, Insightful)
Re: (Score:2)
They could add hardware scrolling to their chip and not be dependent on mutli-core processors! /s
Re: (Score:3)
Re: (Score:2)
Why do they need Qualcomm? Just use MediaTek. The story itself is pretty misleading, implying that there's only one solution, and that's QCOM:
If youâ(TM)re using a relatively recent, non-Apple, sold-in-the-US smartphone, odds are good that it contains some kind of Qualcomm SoC.
No, actually, vast numbers of non-Apple phones (why not just come out and say "Android" there) use MTK chipsets (and others, AllWinner, etc).
Another thing they conveniently forget to mention is just how hard it is to design a full mobile phone SoC. It's not like you can get a bunch of enthusiastic coders, lock them in a room for a few months, and a full-featured mo
Re: (Score:2)
Are you being intentionally dense the quote you pulled explicitly says "sold-in-the-US smartphone". Please point to these "vast numbers" of US sold smartphones with MediaTek and Allwinner SoCs.
Re: (Score:2)
My sold-in-the-US smart phone has a MTK chipset. My previous sold-in-the-US smart phone had a MTK chipset. Several friends of mine with sold-in-the-US smart phones have MTK chipsets. Other sold-in-the-US brands using MTK chipsets are Acer, Alcatel, Cubot, Gigabyte, Huawei, Lenovo, LG, Philips, Oppo, and ZTE (among others). You may have heard of some of those multibillion-dollar companies, .
Thank you for your well-thought-out, insightful comments though.
Re: (Score:2)
Way to dodge the question. You didn't actual list a single smartphone model or provide sales figures to prove your "vast numbers" of sales claim. Yes, you can buy some niche junker phone with such a SoC but the only Android phones topping sales charts in the US pretty much exclusively use Qualcomm.
Re: (Score:2)
Except Google is perfectly capable of hiring away mobile SoC designers from anyone they want. Or, of course, just going the Apple route and buying an entire mobile semiconductor company. Remember, they already bought Motorola and kept most of the IP.
Re: (Score:2)
Re: (Score:2)
I'm surprised no one hast mentioned nVidia. $15B market cap (totally in Google's price range), and they already have a tablet and Android TV STB that kicks everyone else's ass. Guess it's currently too expensive, but there's no question it blows away the Android high end...
Re: (Score:2)
Intel have the AVX2 instruction set, which is equivalent to the vector instructions of a GPU shader core. They also have Knights Corner/ Knights Landing / Xeon Phi multi-core CPU designs which are getting close to low-end GPU with a few hundred shader cores. Biggest difference between CPU's and GPU's would be the support for the hundreds of different texture formats that are available, everything from single channel color, to floating point RGBA and all those different compressed formats; RGB10_A2) and the
Re:Google could buy Qualcomm... (Score:5, Insightful)
Also - did you read the article and understand the point?
Qualcomm is the only SoC manufacturer for Android pushing the high end - MediaTek, etc are focusing on the mid-range. Google wants to push the high end to compete with the latest Apple chips, and MTK and others aren't even close. Right now it's either push Qualcomm or do it themselves.
Re: (Score:2)
NVIDIA might like a word with you. They may not have much traction, but it's not for a lack of effort in pushing the high-end.
Re: (Score:2)
I have had "words" with Nvidia - and in fact am working with them on a project for their Android TV-based Shield. They have great hardware for STBs and high end tablets.
But this article was about smartphones, and they gave up on the Android phone market months ago...
Re: (Score:2)
Another possibility: they owned Motorola previously, now they can buy Freescale, or at least the mobile part of its CPU business and run w/ it. Might gain access to Power (as in IBM's CPU) IP.
Re: Google could buy Qualcomm... (Score:2)
Isn't MediaTek horrible in terms of support (i.e. driver updates for newer Android versions etc.), like even worse than Qualcomm?
I'd assume that's one of the things Google is trying to avoid, so as to open up the possibility of longer term handset support a la Apple iOS.
Re: (Score:2)
... and fill the "Q" letter in their "Alphabet".
Insightful indeed. If Google can buy Qualcomm, it would give them all patents related to not just CDMA, but also recent cellular technology. Which would be a huge thing going for it.
Other thing I'd like Google to do is to build chips outside the ARM ecosystem - bring back some variety in the microprocessor market. We already have x64 and ARM, so I'd like to see them make something based on another chip - such as MIPS. The latter has the most experience in its development, and Google could leverag
Alternate headline (Score:2, Troll)
I don't buy it (Score:5, Interesting)
Chip design is very hard and unforgiving. Google knows this, and can't be looking to jump into the business. They might want to help tailor something, but that would be about it...
Re: (Score:3)
Google needs to have a hypervisor, VMs, and have all this work seamlessly for Joe Sixpack. Virtualization takes the right hardware to pull off.
That's the first time I've ever heard anyone ask for VMs running on a phone.
Re: (Score:1)
That's the first time I've ever heard anyone ask for VMs running on a phone.
Yet somehow, everyone who bought an Android phone is using google's JVM. Guess what the VM in JVM stands for?
Re: (Score:3)
Yet somehow, everyone who bought an Android phone is using google's JVM. Guess what the VM in JVM stands for?
That's great, so if you throw that in another VM, it would be a VM running in a VM.
Re: (Score:3)
Yo, dawg...
Re: (Score:1)
These days you have zero security unless you can shitcan your current environment and refresh from memory that cannot be accessed by the vm
Re: (Score:2)
The idea has been around for a while. You have a phone for work and a personal phone. Want to combine the two into one piece of hardware without giving control of your BYOD device to your employer? A virtual phone running in a VM on your own hardware is the answer.
Re: (Score:2)
I'm wondering what Blackberry's plans for QNX are...
If Priv signals the beginning of the end of BBX then does the company merely become an Android handset vendor or do they leverage any technical advantages of QNX and run Android/Linux in virtualization?
Re: (Score:3)
ARM supports virtualization. It's the mode above kernel mode, and ARM even has sample code to setup a VM instance using the ARM hypervisor.
Re: (Score:2)
Who would have thought that mobile devices like smartphones and tablets would be going 64-bit? That's only needed for high-end workstations and servers. Who would have thought mobile devices would have four CPU cores and tens of GPU cores?
Re: (Score:2)
Using a VM will slow things down, and is essentially a failure on the part of the OS designers, added in a way to provide backwards compatibility (like in OS/360) or as a way to provide security that should be in the OS but is not. On a server, VMs are often used as a way to prevent different services from conflicting with each other, or as a way to bundle them up so they can be easily transferred
Well played (Score:5, Insightful)
Re: (Score:2)
Tracking you will be much more efficient. Built in, unblockable analytics.
Google wrecks everything fun.
And that's why we can't have anything nice.
Re: (Score:2)
To be fair, most filtering apps seem to run as fake VPN configurations, routing connections through the App's VPN "server" running on localhost to do the blocking. Putting Apps in a VM with the virtual network hardware hooked directly up to the hypervisor's networking stack, and the VPN config by extension, could make it even harder for apps to bypass that sort of filtering.
Re: (Score:2)
Tracking you will be much more efficient. Built in, unblockable analytics.
No.
I work on the Android security team, and we actively block any attempt to build tracking into the core platform. It's not hard to do, either. Mostly what happens is that people design new features which they don't realize could be used for tracking, we point them out, and they say "Oh, right, guess I need to find another way."
Google doesn't really want to track you, except as voluntary quid-pro-quo for using some service you value.
Re: (Score:2)
I really really want to believe this.
So install the Android sources, and start snooping.
Re: (Score:2)
Tracking you will be much more efficient. Built in, unblockable analytics.
No.
I work on the Android security team, and we actively block any attempt to build tracking into the core platform.
There's only one problem. No way in hell I believe you. I wouldn't be surprised if you aren't even lying, you could be plausible deniability bait. But in a world where us civilians are the product to be monetized, or where the government wants to know all about us, it's almost impossible to believe that either group isn't trying to make their job as easy and convenient as possible.
Re: (Score:2)
I wouldn't be surprised if you aren't even lying, you could be plausible deniability bait.
That would imply there's someone else behind the scenes, hidden from me and subverting my work. But there's simply no room for that to be true.
Re: (Score:2)
I wouldn't be surprised if you aren't even lying, you could be plausible deniability bait.
That would imply there's someone else behind the scenes, hidden from me and subverting my work. But there's simply no room for that to be true.
Which still wouldn't imply that I or many others believe you.
I mean, why wouldn't Google do this? Seems this design is coming out about the same time as more people are becoming savvy to the invasion of trackers, and maladware and are installing software to get around the practice.
Any group I was involved in, especially the one who is the biggest player in monetizing users internet habits, would want to preserve their cash cow, and work at ways to defeat the tracking - indeed - just imagine "total user
Re: (Score:2)
I wouldn't be surprised if you aren't even lying, you could be plausible deniability bait.
That would imply there's someone else behind the scenes, hidden from me and subverting my work. But there's simply no room for that to be true.
Which still wouldn't imply that I or many others believe you.
Sure, you can just believe that I'm lying. But that's only possible because you don't know me.
I mean, why wouldn't Google do this? Seems this design is coming out about the same time as more people are becoming savvy to the invasion of trackers, and maladware and are installing software to get around the practice.
There are many, many reasons why Google wouldn't want to. Included among them is the fact that both the founders and a large percentage of the employees would rebel for both moral and business reasons (though, honestly, the moral reasons are more important). Google has never been particularly comfortable with advertising. Larry and Sergey refused to do it for the first portion of Google's history, until they hit on
That's easy! (Score:1)
All they have to do is contact Apple, I've heard they've been designing their own chips for a while now.
Fight for your bitcoins! [coinbrawl.com]
Re: (Score:2)
I thought that was what their little Motorola adventure was about
Re: (Score:2)
I thought that was what their little Motorola adventure was about
It was more about buying the patents that Motorola owned.
Re: (Score:2)
Having that patents does protect them from litigation as they have plenty of stuff for a counter-suit, but they ended up paying a lot. They probably would have been a lot better off buyi
got $3billion cash, $2.5 deferred tax losses (Score:2)
There are two items missing from that analysis, worth $5.5 billion.
First, Motorola had $3 billion in the bank. When Google acquired Motorola, they aquired those bank accounts. Second, Motorola had paid $2.5 billion too much in taxes, deferring their losses until later. The last time I checked it wasn't entirely clear, but Google should have been able to reap those losses.
Before somebody gets their panties in a wad, that's money Motorola had paid in taxes, but wasn't actually owed. Basically Motorola's tax
Ha (Score:2)
>Google has been stepping up its efforts to build higher quality Android phones
Then they shouldn't have ruined the Nexus 5X by giving it 4-year-old storage options and 4-year-old memory options and removing the wireless charging.
Just because some of us don't want a huge phone, doesn't mean we want weak specs. I was very disappointed because I have loved the Nexus 5 for two years and wanted to upgrade. Now, what is the point?
Re: (Score:2)
Well I don't own a tablet but from my experience of tethering a laptop, buying a tablet without 4G would be a major annoyance if its primary purpose was to pass the time on public transport or watch a video on a summer's
Near monopoly?! (Score:5, Informative)
"Qualcomm has a near monopoly on Android SoCs"
What about Mediatek, Samsung (Exynos SoCs in many of their top-selling phones and tablets), the Chinese fabless semicons like Rockchip, Allwinner, etc, even Intel (Asus Zen phones/tablets)? Statistics please without qualifiers like, a near monopoly on tablets sold by LG, Moto, and so-and-so company.
Re: Near monopoly?! (Score:2)
Correction: Qualcomm has a near-monopoly on SoCs capable of doing LTE on American mobile networks in all licensed bands without additional chips. If you want a single-chip 4+ core Android device that works with 100% support for all relevant bands & data modes on Verizon, Sprint, or even AT&T, you basically have one viable choice: Qualcomm.
Renesas had a competitive alternative chipset ~2 years ago, but the new owners seem to have no real interest in trying to compete with Qualcomm in the US anymore f
Re: (Score:2)
I thought Exynos solved that problem with the S6?
Re: (Score:2)
Re: (Score:2)
Even when OMAP was still a thing it was only ever used by a dozen or so phones. And those were phones from 2010-2012.
Note the source is "The Information"... (Score:3)
Nothing to see here, move along.....
Re: Way behind (Score:2)
As others have noted, Android's biggest performance problems come down to architectural compromises made *years* ago so Android could run (walk?) on 200Mhz devices with 480x320 displays and almost no RAM circa 2009. Just about everything Android does is PIO-based... and the closed nature of Qualcomm's chipsets means end users are still running on an ever-accelerating treadmill with every new device & version of Android just to have a working camera & GPS under the new (and 100% incompatible with bin
Not a manufacturer (Score:2)
Google is not a monolith (Score:2)
Intel Atom chips are the best for Android (Score:1)
Don't get it. (Score:2)
We're in a constant rush to get a better processor in to a smartphone, and yet the current gen chips are still heavily underutilized. I have an old Galaxy S2 i9100 in my desk that still performs well enough to do all modern tasks. My aging HTC M8 which is coming up on its second birthday is still as zippy as ever.
What exactly are we racing towards more powerful phone hardware to do?