|
|
Log in / Subscribe / Register

Wistron Shows Google Android Phone (PC Magazine)

[Posted January 7, 2008 by jake]

PC Magazine plays with a phone that may become the first Android phone. The GW4 from Wistron will be running the Android software by March – which could make it the first – though the version described runs MontaVista Linux. "The GW4 we saw had surprisingly low specs, but that's a testament to the efficiency of Linux, Wistron execs said. The GW4 is based on a TI OMAP 1710 chipset with a 216-MHz processor and only 64 MB of program memory, yet the model we saw ran the Opera Web browser, played video and flipped between a range of Web widget applications like weather and stocks. The user interface was very responsive."
to post comments

GOOG -4.13%

Posted Jan 7, 2008 21:20 UTC (Mon) by proski (guest, #104) [Link]

Amusingly, one of the screenshots shows Google stock losing 4.13% of its value. Coincidence? A hidden message?

By the way, why would not Wistron put its own name of the end product? And what is the brand name they are going to use then? Google?

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 6:31 UTC (Tue) by mtaht (guest, #11087) [Link] (11 responses) 

Just to be grouchy, Montavista was among the first to realize, code for, and ultimately
implement an embedded Linux designed expressly for low power situations, especially cell
phones.

While Android may be getting a lot of press, it's still barely past the vaporware stage.
Montavista has been shipping tools and an OS tuned to cell phones for multiple years now, and
is running on 30+ million shipped phones - including the one demonstrated in this article! 

It bums me out sometimes that mv doesn't get more respect for what they've done, what they
understand, and where they've been. I am far from convinced that google's groupthink can take
over the cell phone market, although I am enjoying their try, and look forward to working with
Android sometime in the future. 

Handheld embedded development, due to Amdahl's law, is far more bound by power consumption
issues than anything else, and not by Moore's law. 10 years ago you could get a power-sipping
208Mhz strongarm processor in a pda (or cell phone) - now you can get a 400Mhz arm variant for
about the same cost - a mere doubling in *10* years, not even remotely tracking the
predictions of a purely Moore's law approach - and the processor mentioned in this article is
actually slower in several respects (and faster, due to the DSP, in others) to the ancient
strongarm.

You'll note the slow processor and lack of memory on these kind of handhelds are incentives
for programmers to still create fast, tight code, and companies to employ them.

Secondly - due to the need to comply with the reliability and legal needs in this highly
regulated market, progress in cell phones is likely to remain far slower than users might
want. 

(confession: I am a former montavistan that helped ignite our cellphone effort back in
2000-03, long before it was a blip on google's radar - back when the company employed some of
Linux's embedded greats)

I personally am an avid follower of Bell's law (http://en.wikipedia.org/wiki/Bell%27s_Law)
about computer classes - for a huge percentage of people - more to come - their most important
computer is the one in their hand, (soon, on their body), not in their lap or desktop. 

Android's approach is going to help create demand for ever more powerful devices in your hand,
and it - and the wonderful iphone - are breaking apart a very closed market - and that's a
good thing. Still, a little respect for those that are no longer trendy seems in order...
especially when they've been shipping working product far longer than google has.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 7:03 UTC (Tue) by dlang (guest, #313) [Link] (4 responses) 

hardware development is faster then you imply

while the speed may only have doubled in 10 years, the power requirements have dropped and the
price has dropped even more drasticly. 10 years ago the 'high end' pda at 266MHz would set you
back close to $1000 and you would be doing good to get one day's worth of use out of the
batteries. today the 400MHz machine is ~300-400 (with quite a few more features added) and
will last several days on a smaller (and therefor lighter) battery pack

for me the big news of google getting involved in the mobile phone market isn't that it's new
for linux to be involved, it's that they may have the money and name recognition to finally
break the stranglehold that the powers-that-be have over the mobile phone market.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 12:12 UTC (Tue) by jonth (guest, #4008) [Link] (2 responses) 

Also, embedded device speed increase is more like 10x faster than that old 266Mhz PDA. I work
on a (commercially available) 3.5G cellular device which contains 2 cores running at 1GHz. As
dlang implies, the embedded space tends to work to different requirements, so we are much more
likely to trade speed for better power consumption. This stands to reason: the processes we
use are basically the same as the desktop chips (we may run one process generation behind),
but we just run the chips at a different operating point (lower voltage, so lower frequency).

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 9, 2008 0:39 UTC (Wed) by man_ls (guest, #15091) [Link] (1 responses)

And, Moore's law speaks about the number of transistors: at a given price, it will double every 18 months. It says nothing about processor speeds.

I think it was a historical accident that processor speeds grew that fast in 1970-2000. In fact, the current generation of processors runs slower than the last (in Intel's i386 offerings), or roughly at the same speed (everyone else). It does not seem to be practical to go beyond 3 GHz, and yet transistor densities are still increasing at the predicted rate. Also in the embedded space, which means that the new processors can do much more in the same space (and with the same power). Speed is not the only factor.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 11, 2008 9:14 UTC (Fri) by Cato (guest, #7643) [Link] 

Intel's Core 2 Duo's run considerably faster than Pentium 4's, even with just one core - they
have a similar clock speed but can get more work done per clock. 

Intel would not still be in business if its new processors were slower than the old ones, and
in fact it is beating AMD at present.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 18:57 UTC (Tue) by mtaht (guest, #11087) [Link] 

Costs for embedded chips has come down - still not anywhere near what Moore's law predicts.
Speed has not gone up, anywhere near what Moore's law predicts. Power consumption is a major
factor. Moore's law does apply to things like flash storage, I note - I have 8GB on my nokia
n800....

I confess to being somewhat disappointed that we aren't seeing more embedded cpus using < 90
nanometer processes at this point.

re:
" for me the big news of google getting involved in the mobile phone market isn't that it's
new for linux to be involved, it's that they may have the money and name recognition to
finally break the stranglehold that the powers-that-be have over the mobile phone market."

I certainly hope so! I remember the major disconnects we had in meeting after meeting with the
major operators in this market... 

...trying to get sensical things like treating voicemail as data, using rss and dropping the
concept of WAP, supporting multiple toolkits and X, etc. They had NIH and revenue_per_call
embedded completely in their every model, not what their customers wanted.

I also particularly bitterly remember battling with TI to free up their omap compiler so
normal people could actually use the DSP. Finally (as part of the openneuros project) - a free
compiler is available for OMAP. It may be too little, too late, as vfp is quite nice and quite
a bit easier to use.

And my whole point to the posting was simply that the article was showing off (and praising)
the Montavista stack, not android. The title was misleading....

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 9:46 UTC (Tue) by davidw (guest, #947) [Link] 

Android has its defects, but it runs on hardware - I'd call it a bit beyond the vaporware
stage.

I write code for J2ME and Android (Hecl - www.hecl.org) because they're "complete" platforms
that are widely available.  You can't just make some embedded tools and an OS and call it
good, you have to have everything from the drivers to the GUI, and have the muscle to make it
widely deployed, something that Sun has done, and Google is credibly able to do.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 11:14 UTC (Tue) by epa (subscriber, #39769) [Link] (4 responses) 

Amdahl's law?  That says (roughly) that if you can only speed up half of the instructions run
by a system, the biggest speed improvement you can get is 2x, since there will still be the
other half of instructions going slowly.  It's usually applied to parallel computing (if you
can parallelize 90% of your program, then you can make it go ten times faster, but no more
than that).  Are you sure this is the law you meant?

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 18:41 UTC (Tue) by mtaht (guest, #11087) [Link] (3 responses) 

Re: Amdahl's law - Amdahl's law applies to many things besides strictly computation and I
should have made that more clear, above. It also applies to all the processes/tasks and
requirements that go into making a product. If your goal is to make a product that does x,y,
and z, size s, weight w, and has a day of battery life, the cost of the battery life and the
weight dominates over how good x,y,z can be - and you end up (currently) with 64MB of ram
(128MB on a good day) and a 200Mhz processor in a typical cell phone.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 8, 2008 19:40 UTC (Tue) by mtaht (guest, #11087) [Link] (2 responses)

I wish I could edit my own comments... the bold clarifies:

Re: Amdahl's law - Amdahl's law applies to many things besides strictly computation and I should have made that more clear, above. It also applies to all the processes/tasks and requirements that go into making a product by a certain date. If your goal is to make a product that does x,y, and z, size s, weight w, and has a day of battery life, the cost of the battery life and the weight dominates over how good x,y,z can be - and you end up (currently) with 64MB of ram (128MB on a good day) and a 200Mhz processor in a typical cell phone.

The equation is roughly the same as amdahls, with some more complicated terms thrown in for things like battery life, I wish there was formally published something that threw in terms for the rate of battery life increase vs weight, etc.

The only "out" of this cruel equation has been "better software" - things like the tickless kernel, for example, have come out of the quest for better power consumption. Also, libraries like what nokia has (maemo's OSSO hooks) for signalling applications when the screen is off, keyboard present, or the device is changing sleep states have been proven useful (and I hope that more Linux applications start paying attention to screen states in particular)

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 9, 2008 14:11 UTC (Wed) by epa (subscriber, #39769) [Link] (1 responses) 

It seems like you've generalized Amdahl's Law so much that it says 'you can't please all of
the people all of the time', or something like that.  To be a 'law' it needs a quantitative
definition - do you have one?

From a quick search I can't see anyone using Amdahl's Law in any sense other than the limits
of speeding up a computation.

Wistron Shows Montavista Phone (PC Magazine)

Posted Jan 9, 2008 18:13 UTC (Wed) by mtaht (guest, #11087) [Link] 

Yea, I'll work on it.


Copyright © 2008, Eklektix, Inc.
Comments and public postings are copyrighted by their creators.
Linux is a registered trademark of Linus Torvalds