Allwinner Shakes Up Mobile Chip Industry With $5 64-Bit Chip

Status
Not open for further replies.

morerice

Distinguished
Sep 29, 2010
90
0
18,660
Can someone more familiar with the mobile chip architecture explain why Intel hasn't been able to manufacture a competitive alternative? Intel really hasn't been able to keep up with ARM architecture in all the specs that matter for mobile platforms.
 

yhikum

Honorable
Apr 1, 2013
96
0
10,660
Can someone more familiar with the mobile chip architecture explain why Intel hasn't been able to manufacture a competitive alternative? Intel really hasn't been able to keep up with ARM architecture in all the specs that matter for mobile platforms.

The explanation might be as simple as there is much less demand to run x86 based OS and their applications on small screen devices. Look at success of Microsoft Surface. Even it is popular, the concept will never be promoted down to smaller screen devices such as phones. And when you consider hardware requirements for phone you will see that SoC is just a part of whole solution, as you'll be needing battery, screen, cell connection module and make it into a case. Intel is trying to bring x86 legacy to mobile market, which already moved beyong windows. Running a form of Linux for x86 might be a case, but you can achieve much more with cheaper and less power hungry ARM implementations already.

Hence, we see articles where Intel is somewhat an oddball when it comes to mobile devices.
 
Can someone more familiar with the mobile chip architecture explain why Intel hasn't been able to manufacture a competitive alternative? Intel really hasn't been able to keep up with ARM architecture in all the specs that matter for mobile platforms.

ARM has been designing their architecture for low power embedded use for over two decades now. Intel has only just recently begun trying to shoehorn X86 architecture into mobile devices.

On top of that, ARM strictly develops the technology; they license it to companies to use. Intel not only is trying to design their mobile X86 technology, but also fabricate the actual chips themselves.
 

kenjitamura

Distinguished
Jan 3, 2012
195
3
18,695
I think Allwinner is like the only Chinese SoC that makes their source available to the open source community, right? If so awesome to hear they're putting this out on the market. We might get some decent non-chinese tablet manufacturers to put these chips into their products and then we'll have some very good and cheap devices that Cyanogenmod can be loaded on.

I say non-chinese tablet manufacturers because even if the Allwinner SoC has source code floating around the tablet also has WLAN, touch screens, etc that need device drivers and we all know a Chinese tablet manufacturer will not release source code for those.
 

bit_user

Polypheme
Ambassador
Can someone more familiar with the mobile chip architecture explain why Intel hasn't been able to manufacture a competitive alternative? Intel really hasn't been able to keep up with ARM architecture in all the specs that matter for mobile platforms.
The simplicity of the ARM instruction set translates directly into a smaller, simpler chip, which burns less power. This also means layout is easier and cheaper. And manufacturing costs are lower, because more chips can fit on a wafer and yield is higher.

ARM also has a fundamentally different business model. It licenses its designs to customers, some of whom do various levels of customization. This also benefits from ARM's relative simplicity.

In the past, Intel has gotten burned every time they strayed from x86. I think they might have learned their lesson too well. When it comes to mobile, nobody cares about x86.
 

bit_user

Polypheme
Ambassador
I find it interesting that both MS and Intel are threatened by essentially the same problem. They built huge businesses based on complex, high-margin desktop and server products.

These are difficult to scale down to mobile. And when they do, they find that actual competition and cheaper end products prevents them from achieving big enough margins to replace the cashflow they're losing from the shrinking PC market.
 

jasonelmore

Distinguished
Aug 10, 2008
626
7
18,995
Reaching $5 price point:

Yeah right, Easier said than done. these kind of promises always end up being broken by the time products hit OEM's. and i'm pretty sure the ARM licensing a lone is $1 per core
 

MyDocuments

Reputable
Jun 21, 2014
39
0
4,530
This is good news on the competition front, poor Intel, they always wanted to make big margins, oh well, perhaps those days are over(?).
However, aside from the larger tablet form-factors, I see nothing of modems (2G, 3G, 4G, etc) and connectivity being mentioned (BT, WiFi, NFC, etc.) and integration is the big prize in mobility and the race to the bottom; everything built in, and of course a good support network for helping customers get these devices into their products.

Overall a nice effort on this product, let's see where the early adopters are, oh, and they've previously positioned their devices (A10, A20) in the Cubieboards and Banana-Pie so well done!
 

fagiano

Reputable
Jan 17, 2015
1
0
4,510
@jasonelmore : AllWinner has been selling dirt cheap SOCs since a while. Few years ago you could already buy an A10 for 7$ in small volumes. 5$ is proabably for larger volumes. I think they are serius about it. Also, I'm not sure what's exactly ARMs license cost but is definively much less than 1$ per unit(ARM would have 100 times the revenue if that was the case).
 

ET3D

Distinguished
Mar 12, 2012
99
33
18,660
"Allwinner's $5 quad-core 64-bit chip should make possible the creation of mobile devices that cost as low as $50, come with Lollipop out of the box, and deliver surprisingly good performance for the price"

According to ARM's site, a Cortex A53 is marginally faster than a Cortex A9 at the same clock speed. Put it in a $50 device and it will have 1GB of RAM at most, making the 64-bitness worth very little. I don't find that particularly exciting. Yes, it will be a decent improvement over the A7 that's prevalent at the low end, but I'd rather buy a Cortex A17 tablet.


Far as I checked, a Cortex A53 is slower than a Cortex A17. Its main benefit is
 

bit_user

Polypheme
Ambassador
Well, they both have a decode width of 3, but the A53 has an 8-stage pipeline, while the A17 has an 11+ -stage pipeline. This is possibly explained by the A53 being in-order, while the A17 is out-of-order.

Interestingly, the A17 is typically used as the big core, in a big.little configuration, whereas the A53 is typically the little one. I guess that sheds some light on its pipeline architecture.

You make a good point in questioning their decision to go with 64-bit, for such cost-sensitive applications. I don't know if the ARMv8 architecture provides any other advantages, for an in-order chip, but the A53 has a claimed DMIPS/MHz of 2.3, which is higher than any of the in-order ARMv7 cores.

I'd guess that the A53 is the most power-efficient ARMv8 core. So, we can say this will probably be power-efficient, but not particularly fast. But the price suggests it might not be on the latest process node, so probably not the highest DMIPS/W out there.
 

austenwhd

Distinguished
Dec 21, 2011
267
0
18,810
Govt. of India's shuold use same in Akash Tab.

@Tumesh Kakkar, they just might do that, considering the Make in India and Digital India plan is in motion, we might see newer technology more frequently available than before. Also, the tablet is still working on dual core and 512 MB RAM which is seriously too low, quad core and 1 GB at least should be the target, and with the help of cheaper chip, we might as well get these tabs, with all these upgrades, but still in same price range.
 

Vorador2

Distinguished
Jun 26, 2007
472
12
18,785
Well, the perfomance of Allwinner SOCs is not very good, but the price is very attractive. I bet most OEMs will use this chipset for their lower cost models.

On unrelated news, Intel is losing money in mobile because it's paying manufacturers to include Atom in their devices. Which i believe should be forbidden since it's the same as dumping.
 

bit_user

Polypheme
Ambassador
Exactly what I was wondering: how is this legal?

But I'm sure Intel lawyers have found ways.

On the plus side, one consequence is that you can buy a J1900 mobo + CPU on a popular online store for $63 shipped, right now, yet the tray price of the CPU, alone, is $82 (http://ark.intel.com/products/78867/Intel-Celeron-Processor-J1900-2M-Cache-up-to-2_42-GHz?q=j1900).

This is good news for anyone wanting to build a NAS or a low power server of some sort. I almost went this route, but sorta upsold myself to an i3 for its ECC support.
 

epobirs

Distinguished
Jul 18, 2011
197
13
18,695
"Exactly what I was wondering: how is this legal?"

It's legal because it is not remotely the same thing as dumping. Dumping is putting product in a market at below your cost to disadvantage the competition until they leave the market and you can raise the product's price to sell it profitably and unopposed. That is not what is happening with x86 based tablets. You aren't seeing Surface Pro 3 tablet going for $400 to take out Apple.

What Intel is doing is most often manifested in co-op advertising, which is extremely common across a wide range of businesses. When an endcap at Target is used to prominently display a particular brand of product, you can bet money changed hands for that valuable display location. When an ad for a new EA game plugs GameStop as the place to buy it, that is another example.

It is entirely legal to offer a retailer or a company downstream in your manufacturing chain incentives to carry product in their stores or use your product in their designs. The cost of such incentives is a big chunk of the tally for Intel's losses in mobile. Because the product is not being given away or absurdly discounted and nobody is required to buy no matter how heavily advertised and prominently displayed in the stores.
 

bit_user

Polypheme
Ambassador
What would Intel gain by buying them? It already has far more design & manufacturing capability than they do, and their current revenues would be but a blip on its financials.

I think the issue is that Intel is stubbornly sticking to x86 and trying to use it as a differentiator. In some sense, it's hard to blame them for this, because suppose they did build yet another ARM SoC... It might be the best out there, but how much more could they charge for it? It'd be a race to the bottom, for them. I think they rightly see that they need a differentiator to set them apart from the competition and help them justify higher prices (and thus margins). It's just that x86 seems like more of a disadvantage than a value-add.
 

Giroro

Splendid
Is $5 actually a good price for a tablet processor? What does the competition cost? I've always assumed that the tablet SoC isn't actually that big of a part of the final price compared to the screen, batteries, shipping, and assembly.

Also, to people asking why Intel is pushing x86 over ARM. Intel owns x86 - if that architecture wins, they win. Licensing out a competitor and further supporting it's dominance in the market does not help them when their goal is to push out ARM altogether. Once Windows (Real Windows, not Windows RT) runs on ARM, Intel will be at real risk of losing their near-100% monopoly on the PC market if they can't kill ARM before that happens (AMD has to pay Intel for x86 processor they make, remember). Then again, it is probably close to technically impossible for an OS and it's applications to be cross-compatible between different hardware architectures, so that may not be a real risk to intel.
 
Status
Not open for further replies.