Imagination Releases New PowerVR GPUs

[pattakosn]

I really do not understand why you are so furious about Imagination.

While I am all for open source drivers, Imagination does not write a linux driver for its chips neither releases any drivers to the public. Instead, it is their customers who release drivers for their SoCs.

Of course I would like it if they released an open source driver, or any kind of driver.

[smitty3268]

I really do not understand why you are so furious about Imagination.

While I am all for open source drivers, Imagination does not write a linux driver for its chips neither releases any drivers to the public. Instead, it is their customers who release drivers for their SoCs.

Of course I would like it if they released an open source driver, or any kind of driver.

Imagination is the company blocking release of any drivers. I assume they license their own internal drivers to their customers, and then allow those customers to make minor changes on top. If it's somehow different then that, they are still the company releasing all the docs under NDA and agreements to never open that information up.

[pattakosn]

Imagination sell intellectual property (ie their hardware designs) to customers like intel, apple, TI, etc. It is their customers who release drivers for their products. How do you know that it is not their customers who block drivers or spec releases ? Even if Intel and Imagination are willing to be more open, there may be agreements with other imagination's customers blocking it.

Given the fact that Imagination 's android driver is pretty good on a lot of devices (smartphones and tablets from so many vendors) and it is only the netbooks from Intel that the driver is crap, maybe this outcry against imagination is a little too harsh.

On the other hand the linux driver apparently is so bad that your comments (ie all of them) are fair!

[Adarion]

pattakosn, ask the developers that tried to get a driver written.
Word is that imgtec changes the chips even within a series from customer to customer. So e.g. you reverse engineered chip type, say, A-1 that is uses on a Vendor-X board. Now, you see that this chip is used on a Vendor-Y board. Wohoo? No. It likely won't work with chip A-1 on Vendor-Y board because something is different. Great, isn't it?
Furthermore, yes, there are surely NDAs all over the place. But since none of any of the firms hold a 100% of Imgtec they can't do much about it.

[jonnyh]

pattakosn, ask the developers that tried to get a driver written.
Word is that imgtec changes the chips even within a series from customer to customer. So e.g. you reverse engineered chip type, say, A-1 that is uses on a Vendor-X board. Now, you see that this chip is used on a Vendor-Y board. Wohoo? No. It likely won't work with chip A-1 on Vendor-Y board because something is different. Great, isn't it?

I would expect this to be similar for all soc-based graphics - Imagination aren't selling a whole chip design, but the graphics is integrated into the whole system so have different display controllers, memory management, cache, CPUs etc.
I wouldn't then be surprised if, from the driver standpoint, there is a large difference between the 'same' graphics core on a TI OMAP compared to samsung exynos, although both claim to be an sgx540.

This may be different to ARM and their mali, as they effectively are selling the chip design as a whole.

I also wouldn't compare android drivers to their windows, there is a big difference between the cut-down Opengl-es (on android) compared to full-blown directx (Though it could be argued that, at the performance/power point imagination are aiming at, many of the features of directx10+ are a bit of a waste, and extra complexity for the driver)

[liam]

I would expect this to be similar for all soc-based graphics - Imagination aren't selling a whole chip design, but the graphics is integrated into the whole system so have different display controllers, memory management, cache, CPUs etc.
I wouldn't then be surprised if, from the driver standpoint, there is a large difference between the 'same' graphics core on a TI OMAP compared to samsung exynos, although both claim to be an sgx540.

This may be different to ARM and their mali, as they effectively are selling the chip design as a whole.

I also wouldn't compare android drivers to their windows, there is a big difference between the cut-down Opengl-es (on android) compared to full-blown directx (Though it could be argued that, at the performance/power point imagination are aiming at, many of the features of directx10+ are a bit of a waste, and extra complexity for the driver)

Exynos, at least the Exynos of the last two generations, used Mali not pvr.
Prior to last year, in the galaxy s, they used hummingbird which was arm 8 + sgx400.