Poulsbo mess casts a shadow on Intel's Moblin project (ars technica)

They needed a low power, low cost, high performance video chipset for the
embedded market. The laptop-style chipsets where you got GMA900+ GPUs use
up much more energy then the Atom processor does.

So licensing the Powervr core was probably much cheaper then designing a
new one. Intel had licensed other GPU designs from those folks in the
past. I think for their old 8xx series chipsets had a older PowerVR design
derivative. So this is nothing new from Intel's perspective.

-------------

Keep in mind that this is a common problem with All embedded Linux stuff.
The Beagle Board and N900 use similarly licensed PowerVR. I don't know of
one ARM platform that has open source 3D graphics.

So if there was a actual open source driver then it would be fantastic as
it could provide a common code base for many devices on multiple
architectures. But, alas, it does not seem like the PowerVR folks are
interested in it.

They won't. I hear the some of the first ARM netbooks will be Qualcomm Snapdragons, and Qualcomm are one of the most Free-unfriendly ARM implementors.

"It would make sense if they really want to use upstream kernels and be able to tweak the OS at every level and update it at will."

They haven't seemed terribly fussed about upstream with android.

"because ARM is PowerVR invested territory too"

Not so. ARM have their own Mali GPU (just as bad Free driver territory). Nvidia have their Tegra (which is even worse I think).

These drivers can make or break a hardware platform. ATI's hardware from
all reports is currently trouncing NVIDIA's, but NVIDIA has an absolutely
freaking fantastic set of driver developers. Even on Windows, I constantly
run into bugs and performance issues on ATI hardware while NVIDIA's
hardware has worked amazingly well on Windows (and Linux) for years now.

The problem, from a corporate perspective, is that if you have a great
driver, the last thing you want to do is give that driver source away. The
competitors having access to a great driver framework for video
acceleration would essentially level the playing field. If ATI had
NVIDIA's driver quality, NVIDIA might not even be around in a few years.
If all those little oddball graphics chips had the drivers that some of the
top-tier embedded graphics companies had, they'd pose a much larger threat.

The success of Open Source and Free Software in the corporate world has
almost always been around commodities, not Freedom itself. Companies
invest in Linux because it's made embedded OS development cheap and easy,
not because they actually give a damn about Freedom. Until video drivers
are as much of a commodity as disk drivers or network drivers or OS kernels
or web servers or interpreted languages or whatever, there is far more
incentive for a company to keep them closed -- and the keep the upper hand
-- than there is to give it away to a bunch of a people who aren't even
going to repay them with dollars, but are instead just going to port it to
some other vendor's hardware.

Gallium3D, being a cross-platform generic modern-day graphics driver
toolkit, may very well change things around. When there is a top-notch
GLSL front and middle end for Gallium3D, a top-tier memory manager, and an
API that makes hooking up a high-quality front and middle end API with a
tiny bit of per-hardware glue code, graphics drivers stand to become a
commodity. Once that happens, companies will no longer benefit from hiring
huge in-house driver teams and keeping their work locked up, because all
that will do is cost them more than just leveraging all the work already
done in Gallium3D.

There will always be holdouts. Broadcom, for example, still is totally
Linux-unfriendly with their wireless chips even though there are a
bazillion Free drivers for wireless already in Linux. That is no longer
normal, though. Just a few years ago, wireless in a Free/Open OS was a
chore. These days, so long as you avoid a handful of hold-out vendors,
you're pretty much golden with wireless. (Broadcom is, unfortunately, one
of the more popular vendors in common hardware, surpassed only by Intel, I
think.) NVIDIA will probably stay away from Open drivers for many years
even if Gallium3D becomes as much of a success as I hope, only because
NVIDIA is so far ahead of the curve that Gallium3D being a great success
won't actually close the gap; Gallium3D will have to pull off a small
miracle to really meet or surpass NVIDIA's team. Or the majority of users
will have to suddenly start caring more about Freedom than Functionality
and drop NVIDIA on principle (likely in the Linux world, very iffy in the
embedded world, highly unlikely for the majority of people who don't really
care about computers at all other than using them as tools to get things
done, especially where "things" equates to professional video/modelling or
gaming).

So far as people hating on Intel for this driver... get over it. Intel is
a huge company. No company that size is going to always make the best
decisions. Look at Sony. The PS3 division was working in complete
opposition to their Music and Movie divisions for many years (and still are
in ways, although at least the Movie division is pretty happy with the PS3
just for helping them crush HD-DVD in the market). Microsoft is in the
same boat; you have people and even whole divisions at Microsoft that are
all about real technology and are _actually_ all about developers and they
actually love the idea of Linux or Open Source, while other divisions wish
Open Source would disappear off the face of the earth. When you have
thousands and thousands of employees and more VPs than some mid-size
companies have total employees, it's only natural that your company's left
hand will be working against the right on a constant basis. The people at
the top are far more concerned with very broad, very general strategy than
they are with piddly specifics like licenses or development models.

Even if the CEO has an opinion on the matter, he really doesn't make those
decisions. Ballmer has next to no say in most of Microsoft's technology,
despite what a lot of the business-clueless Linux users think. Ballmer
actually has less say than he _should_ because he's not even that popular
with a lot of Microsoft's technical folks. Microsoft has been going
through a lot of internal upheaval since Gates left, because he was the
only exec that actually had any clout at all with the developers at the
company. I say this with a fair bit of insider knowledge, given that I
live across the street from Microsoft's main campus, my apartment complex
and every nearby complex or neighborhood is literally filled with people
who work at Microsoft, many of my professors are ex-Microsoft developers,
and quite a few of the people I know are currently employed by or contract
to Microsoft... and most of those people I've talked to are huge Linux fans
for the same reason many Linux users got into the OS in the first place
(development, development, development), although they (like me) have a
far, far more realistic view of Linux's shortcomings and Windows' strengths
than the average Linux fanatic fanboy, especially in the areas of Linux
being an actual contender on the desktop (hint: it's nowhere close and
isn't even moving in the right direction to get there). :)