Computex ARM Netbook Frenzy

[INKster]

I don't disagree darkblu, but NOKIA never wanted in the past either to depend on just one semiconductor manufacturer. Meaning that it's most likely a mix of different reasons.

Synergy between Moblin OS and Maemo OS software development is another strong focus point of their alliance, not necessarily just the sharing of hardware components and wireless IP.
Linux distros on ARM and low-speed x86 for smartphones, MID's and netbooks still have plenty of maturing to do.

 

[Manabu]

I'll be curious to see how aggressively TI pushes OMAP4 for this market - in theory it's well suited for it, in practice given the large amount of competition at this point it's not clear whether they will or even should bother. We'll see.

Well, it seems it will not push at all:

http://www.bloomberg.com/apps/news?pid=20601087&sid=ausdfSe0vOc0

TI seems afraid of Wintel and will try to hang strongly it's position where they think that Intel has no chance to win: the smartphone market. So, no expenditure to make OMAP4 chip more attractive to netbooks. At least, not for a while...

An interesting quote from the interview: “I’ve watched companies like AMD spend their whole lives in a full front assault [against Intel] and I don’t have any interest in trying to do that,”

At least, ARM is not so afraid of Intel. They really geared up this year. Look at last 30 days news alone. Lots of work with foundries for 32nm and beyond and the 2ghz Cortex A9 hard-macro. This last one is a clear sign that ARM is aiming at the new Intel's territory.

The Cortex A9 wasn't meant for go so high in clock speed, especialy in multicore. It was designed for smartphones and set-top boxes, as you can see in the original press release of A9 in 2007. Then Asus/Intel created the netbook market, and it was an opportunity and a threat to ARM. The netbooks do not canibalize only notebooks, but also smaller internet tablets powered by ARM chips, for example. Intel also announced it's plans for smartphone processors. ARM tried to stay away from intel, but this is clearly not possible anymore, and they are up to the challenge.

The thing I'm hopping for now is an announcement of an Cortex A11 64bit architeture, soon. For servers, and even net/notebooks in 2012~14, being able do address +4GB of memory is an must. MIPS already have an 64 bit architeture out for an while...

 

[rpg.314]

Intel recently allowed 2GB of RAM in netbooks with pinetrail. So 64 bit netbooks will be here before 2012.

I think the real non-x86 netbook revolution will have to wait for Chrome OS to take off. IOW, an OS designed to run off the internet in crucial ways. That will be necessary to get rid of the ISA lock in. Not sure how successful MIPS and PowerPC will be in this regard though.

 

[Manabu]

Yeah, Intel will probably be the first with +4GB netbooks, but I don't think it will be a "must" for some time. Remember that we are entering the "good enough computing era", so computer requirements will probably grow slower than the means to provide it. That is actually the only reason why ARM will be competing with intel, because it is still light years away from nehalem chips.

I have an Phenon X4 with 4GB of ram, but I'm still using XP 32bits out of laziness. It is rare for me to use more than 2GB of ram. I understand that Win Vista and Win 7 for example "help" you to easily fill-up 1GB of your memory at boot-time, but even then, 2~3GB of ram is plentiful for basic netbook tasks today.

Cheap 2~3GB ARM based netbooks will still rock in 2012, but by 2013 even for cheper netbooks it may be too low. I'm actually more worried if the processing power of even 2Ghz A9 quadcores will be good in 2012... the hypothetical Cortex A11 would have also to address this issue.

And actually I'm not much excited by Chrome OS... all apps runing only in browser interpreted
environment on an ARM CPU... It will be slow, and I don't trust in chrome's stability and security compared to any mature OS kernel, like linux's one. And how to call Eigen2 to accelerate things in ChromeOS, for example? I hope that it will be easy to "webbreak" ChromeOS netbooks. I hope for Ubuntu's port, or Maemo, or something like that.

 

[rpg.314]

Yeah, Intel will probably be the first with +4GB netbooks, but I don't think it will be a "must" for some time. Remember that we are entering the "good enough computing era", so computer requirements will probably grow slower than the means to provide it. That is actually the only reason why ARM will be competing with intel, because it is still light years away from nehalem chips.

Bloat always increases to take up all the RAM available.

Cheap 2~3GB ARM based netbooks will still rock in 2012, but by 2013 even for cheper netbooks it may be too low. I'm actually more worried if the processing power of even 2Ghz A9 quadcores will be good in 2012... the hypothetical Cortex A11 would have also to address this issue.

Quad core A9 @2GHz would be pretty good actually, even in 2012, as you said, we are entering the era of godd enough computing.

And actually I'm not much excited by Chrome OS... all apps runing only in browser interpreted
environment on an ARM CPU... It will be slow, and I don't trust in chrome's stability and security compared to any mature OS kernel, like linux's one.

V8 is mighty fast with javascript, so I am not concerned with the performance atm.
And Chrome OS is based around the linux kernel.

And how to call Eigen2 to accelerate things in ChromeOS, for example?

Most of the apps are web apps, that does not mean that there will be no native code apps at all. The browser, for a start is a native app. You will certainly be able to write native apps in chrome OS.

I hope that it will be easy to "webbreak" ChromeOS netbooks. I hope for Ubuntu's port, or Maemo, or something like that.

There will be no need to xyz-break anything in Chrome OS. I will be opensource, Prolly, Apache 2.0 licensed.

 

[Arun]

You will certainly be able to write native apps in chrome OS.

Any link? Because TBH I'm skeptical of that. At the very least I think it's clear there won't be a "Native App Store" for Chrome OS.

 

[rpg.314]

There may not be a native app store for Chrome OS, but you won't need google's permission to install native sw on chrome, will you? And what are the odds, no body will fork it if google gets pesky about things like that?

Infact, Debian ships with Iceweasel instead of firefox just because they don't like Mozilla's trademark crap.

 

[Mike11]

Cheap 2~3GB ARM based netbooks will still rock in 2012, but by 2013 even for cheper netbooks it may be too low. I'm actually more worried if the processing power of even 2Ghz A9 quadcores will be good in 2012... the hypothetical Cortex A11 would have also to address this issue.

Is there even gonna be a Cortex A11? I thought the ARMv8 architecture was already planned for the 22nm process. So is there a need for a Cortex A11 as ARM's High-End CPU after the Quad-core A9 in 32nm/28nm and before ARMv8's introduction in 22nm?

And actually I'm not much excited by Chrome OS... all apps runing only in browser interpreted
environment on an ARM CPU... It will be slow, and I don't trust in chrome's stability and security compared to any mature OS kernel, like linux's one. And how to call Eigen2 to accelerate things in ChromeOS, for example? I hope that it will be easy to "webbreak" ChromeOS netbooks. I hope for Ubuntu's port, or Maemo, or something like that.

I don't think that performance should be a problem for a Chrome OS Netbook with a 2GHz dual-core Cortex-A9 and a SGX54x (plus VXD etc.). At least with custom Chrome OS applications, even with almost everything running in the Chrome Browser, if that's in fact what Google is gonna do. Google's Chrome Browser already has (experimental) support for Google's Native Client, O3D and WebGL etc. HTML5, Gears, V8 and upcoming support for Flash on ARM doesn't hurt either. So IMHO Web-apps and custom packaged offline apps from their app store should be able to run fine (assuming software and drivers are optimized for these extended web standards). Palm's WebOS already does something similar.

 

[Laurent06]

Is there even gonna be a Cortex A11? I thought the ARMv8 architecture was already planned for the 22nm process. So is there a need for a Cortex A11 as ARM's High-End CPU after the Quad-core A9 in 32nm/28nm and before ARMv8's introduction in 22nm?

Why do you think there's a relationship between an ARM architecture (ARMv8) and a process? These are not related (unless of course the spec is so huge, an design using it couldn't fit on a particular process ).

 

[Blazkowicz]

Cheap 2~3GB ARM based netbooks will still rock in 2012, but by 2013 even for cheper netbooks it may be too low. I'm actually more worried if the processing power of even 2Ghz A9 quadcores will be good in 2012... the hypothetical Cortex A11 would have also to address this issue.

These specs are good enough for a multi-user machine . Exciting tech. imagine a computer with that much cpu and ram running debian or ubuntu lts on an SSD (or google OS?), an amount of < 5W thin clients (3 to 8), a low-footprint OLED display plugged on each machine (should be a reality by then) and here you have a computer room for an african village or a community living on an ecological lifestyle, serving computing needs from a single affordable photo-voltaics panel.

12" (or what the trend is) OLED netbooks with such high specs are a good prospect too.

I have a very pessimistic mindset about the development of technology in other domains (giant wind farms, hybrid cars, GMOs : lures!)
But the near-future of computing gives me hope. We're talking about the highest technology ever built (28nm, 22nm semiconductors), which allows both to scale back power by an order of magnitude (from <100W to <10W servers and computers), and bring access to independant media and the whole humanity's knowledge to the masses, all around the planet.

A technology that can empower the people from humanity's 80% bottom brass.

 

[Mike11]

Why do you think there's a relationship between an ARM architecture (ARMv8) and a process? These are not related (unless of course the spec is so huge, an design using it couldn't fit on a particular process ).

Maybe I misunderstood, but that's how ARM talked about it. At least according to the article I read a while ago:

ARMv8 instruction set will be application driven, says exec

LONDON — ARM Holdings plc is working on a next-generation instruction set architecture to be the basis for a future set of processor cores. When introduced its capabilities and its extensions to the current architecture – ARMv7 – will be driven by applications and partners, according to a senior company executive.

However, it is not certain that the instruction set will be introduced in the 32-nm/28-nm process node for which ARM (Cambridge, England) is developing an SoC design platform.

"I don't think it is necessary for us to move to ARMv8," said Simon Segars, general manager of ARM's physical IP division. Segars said that there is plenty of scope for the current generation of ARM processor cores to take advantage of increased performance and reduced power consumption offered by the 28-nm nodes being worked on by both the IBM Common Platform group and foundry Taiwan Semiconductor Manufacturing Co. Ltd. [...]

http://www.eetimes.eu/210700147

 

[Laurent06]

Well, to me it looks like Simon Segars was trying not to afraid adopters of ARMv7 CPU's by putting ARMv8 further in the future.

I still fail to see why from an engineering point of view an ISA would require a specific process (except, again, if it's huge).

Of course as an engineer, I might not correctly understand marketing/manager language

 

[Arun]

Of course as an engineer, I might not correctly understand marketing/manager language

Don't worry, there's nothing to understand in that quote Well, unless what it's actually saying is that just like they made these hard IP versions of the A9 on 40nm running at 2GHz, they'll do the same at 28nm and doing that for ARMv8 will have to wait until later no matter when that comes out. That'd be a genuinely weird interpretation though, even if it's plausible.

 

[Ailuros]

Of course as an engineer, I might not correctly understand marketing/manager language

OT but I'd be extremely worried if you would.

 

[Mike11]

Don't worry, there's nothing to understand in that quote Well, unless what it's actually saying is that just like they made these hard IP versions of the A9 on 40nm running at 2GHz, they'll do the same at 28nm and doing that for ARMv8 will have to wait until later no matter when that comes out. That'd be a genuinely weird interpretation though, even if it's plausible.

That makes sense. My interpretation was: He used ARMv8 instead of the more appropriate next-gen core (a la Cortex-A9) because... it hasn't have/had a name yet but it's clear to everyone that v8 comes after v7 . And regarding the manufacturing process: I think he might have just meant that "ARM" can increase the performance of ARMv7 based CPUs significantly until 28nm (introducing quad-cores in 32nm for extreme high-end and then shrink to 28nm for higher clock speeds or lower power requirements). But after that Cortex-A9MP is maxed out, or at least "ARM" couldn't increase performance at the same rate as before (4 cores is AFAIK the maximum for Cortex-A9MP) --> new architecture needed for 22nm.