Slashdot Mirror
Review of the First Medfield Phone
Google85 writes "Beginning April 23rd, Intel, through Lava International, began selling the Xolo X900 smartphone in India for $420, Anandtech has just published a review of the smartphone which runs Android on x86 and uses binary translation as the mitigation for both libraries and NDK applications that haven't yet been ported to x86."
When push comes to shove, they make more money on PC CPUs. When they have a choice of making a wafer of high-margin vs low-margin CPUs, who do you think will win? Don't be surprised if there are major supply problems every time the PC market takes an uptick.
"Though it may take a thousand years, we shall be FREE."
Am I the only one wondering why they're trying to make money off a $420 phone in India of all places? Surely that's a better market for cheaper, sturdier feature phones, like Nokias for example...
One thing I know, and that is that I am ignorant...
Aside from abstract contemplation of "progress" vs "backwards" it probably burns power and generates crazy heat.
I have a X86 netbook on my desk running Android ICS. An old Asus EEE model 900. (my wife has like a 700 also running ICS). It works great, really. The keyboard, although icky netbook size, is better than any smartphone I've ever seen, and the speakers, although icky netbook speakers, are better than any phone or tablet speakers I've ever experienced. The problem is that even running non-emulated (limited selection) apps without an emu layer, it pumps out so much heat even just idling that the cooling fan never turns off. Whirr 24x7. I would imagine an emulation layer would consume even more power.
I would theorize that much as laptops which burn laps are VERY old news, in the future, phones that burn ears and hands are going to be news.
Who will make the first smartphone with a cooling fan? Or a monster solid aluminum heatsink case like a handheld land mobile or ham radio HT?
My experiences show Android X86 would make a pretty good desktop OS for the average user. I'm looking into adding a desktop running android and putting it on the KVM with the other 4 machines on my desk at home (I guess making it my 5th machine)
"Science flies us to the moon. Religion flies us into buildings." - Victor Stenger
If all the existing applications for a given device class are designed for a different instruction set, then of course you're going to have to emulate if you want users of said applications to buy your device. Power Macintosh required a 68K emulation layer, Macintel required a PowerPC emulation layer, and Wii requires 6502 family, Z80, 68000, and MIPS emulation layers to make Virtual Console work. As for this device, it appears only NDK applications not yet distributed in an x86 edition will run in emulation. The rest (over 75% according to the article) will run in an x86 port of the Dalvik VM, and just as with the Macintosh ISA switches, library calls will of course run natively.
I would imagine an emulation layer would consume even more power.
Not always. A game running in PocketNES on a Game Boy Advance SP, especially later versions of PocketNES that can recognize common idle loops in NES games, doesn't use substantially more power than a native GBA game. And with a lot of Android applications spending most of their time either in libraries or blocking on I/O, the emulator isn't going to be running all the time.
They are doing a static binary translation before you download the app. Sure, that won't be as efficient as something compiled natively, but according to their power stats it seems to make little difference as it is just as efficient as the OMAP4 SoC.
The population of India is huge. About 25% of them live below the poverty line. And another 25% have barely enough income to survive. 500 million such people drown out the other 500 million people with some disposable income. The top 25% of India are solidly middle class by American standards. They have steady income, are willing to spend humongous portions of their pay on their children's education. The predatory education industry makes more money than you can imagine. Anyway the richest of the rich live in a kind of opulence that defies comprehension. One guy named Mukesh Ambani built a private residence in downtown Mumbai for the cost of some 1 billion US dollars. It is a 25 story high rise as a private residence! Then some astrologer dude told him such wanton flaunting of wealth would attract the evil eye, and the owner decided not to live there!!!
sed -e 's/Chuck Norris/Rajnikant/g' joke > fact
Is this a first? Premiering a phone in India?
Going forward, as Chindia rises in income, is this going to become the new norm? With the huge markets in China and India (even as a fraction of their total populations), will they become the global arbiters of taste?
Consider the case of 1920x1200 monitors. They're harder to find than a girl on Slashdot. They've all been replaced by 1920x1080 monitors because of economy of scale issues. So ... 15 years down the line, will 350 million people each in China and India dwarf the gadget markets in the US? (Yeah, I know, the mini-countries of Europe. But, again, two countries, two markets, hundreds of millions of people vs 35 markets.)
I'm not a lawyer, but I play one on the Internet. Blog
Performance is roughly on par with the Cortex-A9 SOC's released a year ago, accounting for the clockspeed advantage. Compared to Krait, it's behind in performance, and likely battery efficiency based on the One S. Outside on Win8 tablets, I don't see any compelling reason for using this...unless it's cheaper than the ARM equivalent, which I don't think is the case.
Instead of theorizing why not look at their measured stats? The phone was pretty much as power efficient as current OMAP4 phones. Secondly, Intel is using static binary translation for native ARM NDK apps so there is no emulation layer running on the phone itself for that. In those cases Intel pushes an x86 binary out to the user that has been translated and validated on their side. For most things, though, they are straight Dalvik apps and so there is no more translation overhead than Dalvik on ARM.
nes is what like 30 years old now? emulating it could probably be done by a dumb phones possessor with out to much strain. A modern arm system will take a lot more power.
---Saying gnome 3 is better than windows 8 not so much a compliment as it is damning with light praise.
It is not amazing, but it is competitive. Battery life is average, performance is average. GPU performance is currently a bit below average, but this is a solid first attempt. It clearly proves that x86 CPUs can compete on battery life.
The important question is this: why would you pick Intel over the established Android ARM cortex architecture? It is possible that price and Intel's famous production and supply can win over some manufacturers, but you'd expect something a bit more amazing was required to gain a considerable market share.
Why didn't they just put Windows XP on it? Then it would be *really* useful and have one clear advantage over every other phone. A simple dialler application wouldn't be hard to write to make phone calls. (Linux is better technically, but lots of people are tied to Windows for particular applications and would love to have something more portable than a netbook to run them. In my case, it's a VPN client used to connect to work.) I know Intel wants to push x86 as an embedded platform, and Android is kinda the standard for phones these days, but I'm surprised they give up on the old Wintel model so easily. Heck, I would suggest that replacing Windows Phone with something based on Windows XP and x86 processors is an easy way for Microsoft to grab marketshare in the corporate market.
-- Ed Avis ed@membled.com
Only those methods in the application that use JNI have to be emulated. The main body of the application runs in the Dalvik VM, and the Android system libraries are still native. Besides, the article mentions Intel's tech evangelism to the developers of top applications to get them recompiled.
So it looks the performance and battery life are on par with the current crop of Android phones, but no one has mentioned the main advantage of the Intel device: binary compatibility with x86 architecture, i.e. tons of software that already exists. So there is huge potential of this being the first phone able to run windows and linux binaries by side-loading Debian/Ubuntu with Wine/Virtualbox.
;)
A 1.6GHz Atom should be enough to run Windows XP sufficiently fast, imagine using all your favorite desktop apps on your phone, the screen's not too shabby either with 1024x600. Sure, most won't be optimized for touch input, but that trade-off is worth it for this kind of flexibility. Apps with source code can have their touch-friendliness added, for those that really require a mouse and/or keyboard, those could be added via Bluetooth (or USB?). I see no reason why it wouldn't be able to run apps like desktop Firefox/Chrome (with touch-input extensions), Gimp/Photoshop, MS Office, VLC, maybe even XBMC, or games like Warcraft/Starcraft titles, Counter strike or Quake (I was really missing the Quake3 benchmarks in TFA
Connect it to a monitor and use it as a *real* PC that fits in your pocket and you can bring anywhere - how cool is that?
And when you gaze long enough into the code, the code will also gaze into you.
[...] it pumps out so much heat even just idling that the cooling fan never turns off. Whirr 24x7.
That sounds more like a serious bug ICS on x86 than an indication of platform efficiency.
I would imagine an emulation layer would consume even more power.
That's a reasonable assumption, but the Anand review didn't see any indication that emulation was a big power draw. If anything, the battery life and efficiency (battery life normalized to battery size) were middle of the pack. Disappointingly unexciting for people rooting for either a big win or a big loss.
A native game is likely to be doing a lot more than a NES game, a NES game ported to run natively would use considerably less power than the emulation or a more modern game.
http://spamdecoy.net - free throwaway anonymous email - avoid spam!
I'm sorry but all of them even the one you linked to is the same design, they are just polished different. same as how jet liners are the same design they just are done differently.. it's called a dominate design
http://en.wikipedia.org/wiki/Dominant_design
'...if only "Jumping to a Conclusion" was an event in the Olympics.'
Whoever thought that designing a future product to use an ISA from the 1970s which emulates an ISA from the 1990s would be a good idea?
Well, yeah, except that's not all- AFAIK all Intel x86 CPUs since the Pentium Pro and Pentium II have been designed around a non-x86 RISC-like core, using an internal translation layer to convert x86 instructions on the fly (and hence are still "x86" compatible chips to an external observer).
Actually, I've heard some say the core isn't really that RISC-like, but the point is that it's still *not* x86 and relies on translation.
And yes, I did note in the other comments that Intel's solution uses static translation, but the whole thing is still quite silly when you think about it- a chip using a sort-of-RISC core pretending to be an x86 via a dynamic translation layer is being used to run ARM instructions converted to x86 format.
Hmm
"Slashdot - News and Chat Sites Deviant". (Click "homepage" link above for details).
Presumably those are by far the most computationally intensive routines, otherwise the developers wouldn't have bothered with JNI.
That or they're just ports of an application for another platform whose model layer is written in C++ or another language that doesn't compile to Dalvik bytecode.
You'd be surprised how bad some of the code was in GBA games, even first-party ones. WarioWare Mega Microgames, for example, didn't even have the CPU sleep between frames.