ATI reps - Let's have an update on the HD5850 open sources status

[Kjella]

and yes AMD get good PR by this OS activitys. and this is cheap tons of tech websites talk about amd and the opensource driver and amd don't pay for this PR,,,

I visit quite a few tech websites and I can't say I agree, I don't see much press about it at all and when it does then not so enthusiastic. Next month it's a full year since the Evergreen series launched and we might see the first beta (alpha?) quality public hardware acceleration.

I know AMD never promised to write the open source driver for the community, but the question most people ask is when it is usable for them in mainstream distros, until then it's completely irrelevant to most people.

The whole stack sucked when AMD started working on it and it's not their fault. And they had many years of catch-up to do, like any sane company each generation tends to build on the last. But most people aren't interested in listening to the whys, only the results.

Stores sell mostly the last generation of cards, but you won't buy one to use in 1-2 years when the open source support is finally there. And I'm not talking about full high-performance OpenGL 4.0 support, I'm talking about basic 2D/3D acceleration.

I really hope the open source drivers catch up at least some so that you can buy an AMD card and properly use it under linux with the open source drivers while it's still the latest generation. By the time Evergreen acceleration is out of staging and in mainstream distros, I suspect R900 (or whatever Evergreen+1 is) is already out...

[bridgman]

Wasn't the whole point of AMD funding open drivers for their platform to enable seamless support for the data-parallel engines in their upcoming APUs? I'd think supporting compute would be as big a priority as getting triangles on the screen.

If that wasn't the strategy, then why was AMD so intent on meeting the new APU platform with the FOSS drivers?

The main reason was that we saw the Fusion parts being used in a variety of embedded and small-form-factor products and felt that open source drivers would be strongly preferred for those markets.

[trapxvi]

The main reason was that we saw the Fusion parts being used in a variety of embedded and small-form-factor products and felt that open source drivers would be strongly preferred for those markets.

That's what I was thinking, but I thought there'd be a lot of demand for the compute capabilities for DSP-style operations.

[bridgman]

I think the expectation was that anyone trying to get the most compute performance would go with the proprietary driver anyways, since it has a heavily optimized shader compiler. I expect it will come down to a question of whether or not the rest of the infrastructure (particularly an open source OpenCL implementation) is available.