3dfx did it, but that's only because they knew they were either going to go out of business or get bought out. Voodoo drivers were improved by community efforts, who's to say the same can't be done here?

And a counter-argument to the "the driver is crap" line. Take a look at the wikipedia article on PowerVR and scroll down to the MBX section. A lot of devices would be crashing left and right if that were true. A few major ones: N95, iPhone, iPhone 3G, Sony Ericcson P1

Except that those devices don't use the same driver.

True. So why is the linux driver so horrible compared to say....the iPhone driver which would be the most comparable? resolution? design of OS? Lack of time and energy put into developing it?

The latter can be your best guess. Probably PowerVR didn't see the need to develop it further at the time.

See my earlier comment, far more devices ship which use Symbian than use Linux. I'm not sure where the iPhone lies in all of this.

To re-iterate my understanding, the driver works but is buggy, Nokia won't release a buggy driver (it will crash, people will complain, they will have to support it and will also get bad publicity for releasing a crap bit of software).

To make the driver less buggy, would require paying ImgTech to re-write it (or possibly paying them quite a bit of money to obtain the source under NDA so Nokia can do it, but I'm not sure if this would be allowed). This is where the money thing comes in, not in the per unit licensing cost, afaiu.

And finally, I feel that Nokia probably see the N8x0 as old technology now, Nokia are presumably looking to the next generation to make changes, and the old stuff is just in a holding pattern.

Still no reason we couldn't get some beta software though, if we could persuade Nokia that is (and there's not some cost involved - we would be a small group, or some licensing issue).

Apple insists on writing all the drivers for the graphics cards they use in their traditional computer lines. I've no idea who wrote the graphics driver in the iPhone, but it wouldn't be out of character for Apple to pay whatever licensing is necessary to do it in house.

Plus again the iPhone, etc. have sold quite significant numbers of units so it would make sense to do.

I wonder if the driver from the iPhone could be stolen and made to work for the N8x00. It's Darwin-based, but I'm sure it could be of some help.

And as for licensing, it seems that PowerVR provides the SDK for free, and TI provides the SDK free as well, even more recently a gaming API for the 24x0 chipset. I've already requested the API (which is for the 2.6 kernel)

From the presentations I've on the MBX chipset, it seems that there is no specific limitation and it can actually drive resolutions as high as 1024x768. The other thing I noticed is that the use of the MBX chipset vastly reduces power consumption for raster and 3d graphics usage versus software-rendered solutions.

Has anyone ever seen this?
http://linux.omap.com/pub/patches/www.muru.com.mirror/

There is a kernel patch that has several references to the DSP on the 24xx chipset as well as the on-die graphics processor.

The funny thing is, look at the credits on the patches. Nokia devs.

The driver for the OMAP2430 can't easily be stolen, why do you think the driver for a completely non-similar piece of hardware like the iPhone could? :\

because it has the same chipset and a similar OS?