AMD Radeon R9 290 Released At $399 Price Point

[phoronix]

Phoronix: AMD Radeon R9 290 Released At $399 Price Point

Just a few weeks after releasing the AMD Radeon R9 290X ultra high-end Hawaii graphics card, AMD today unveiled the R9 290. The R9 290 is one step below the 290X but will only set you back $399 USD...

http://www.phoronix.com/vr.php?view=MTUwNDQ

[zanny]

I wish one of the GPU companies wouldn't drop the ball on Linux entirely. At least, besides Intel. Though I'd consider the fact Mesa still isn't GL 4+ compatible yet a pretty big ball drop on their part.

Nvidia just flat out is hostile to the ecosystem and requires you use a binary blob. AMD doesn't try hard enough. I wish there was a way to broadcast to these companies that I, and hopefully enough people to matter, want open spec hardware and some effort to contribute to the ecosystem to support our hundreds of dollars purchases.

[bridgman]

I wish there was a way to broadcast to these companies that I, and hopefully enough people to matter, want open spec hardware and some effort to contribute to the ecosystem to support our hundreds of dollars purchases.

Intel and AMD both have substantial GPU programming documentation and multiple teams contributing to the ecosystem. What else are you looking for from them (us, whatever) ?

[zanny]

Intel and AMD both have substantial GPU programming documentation and multiple teams contributing to the ecosystem. What else are you looking for from them (us, whatever) ?

Michael's wording makes it seem like AMD isn't taking Linux and more in general what Valve is pushing seriously enough. Especially considering how the beta Steamboxes are really pushing Nvidia hardware, it seems like everyone is in some way dropping the ball on this console market Valve is trying to create.

I mean, the fact Mesa is still 5 versions of opengl behind doesn't help. If it isn't up to snuff and both companies cared enough they would put the developer resources into the project to get it up to grade with the binary drivers, or rewrite it, or something.

I'm not trying to critize the developers who are working on it and are paid or doing it for free, you are all great amazing and wonderful and I love you all. It is just obviously not enough, when RadeonSI is 2 years old and still only half as performant as Catalyst, and the Intel parts are hoping for opengl 3.3 by years end while we may see opengl 4.5 next year.

I mean, I know writing opengl code is kind of a PITA after having written shader code before, and implementing it has to be even worse. Maybe Mantle is nicer? I guess well see.

[pingufunkybeat]

I'm not trying to critize the developers who are working on it and are paid or doing it for free, you are all great amazing and wonderful and I love you all. It is just obviously not enough, when RadeonSI is 2 years old and still only half as performant as Catalyst, and the Intel parts are hoping for opengl 3.3 by years end while we may see opengl 4.5 next year.

You are ignoring the progress that's been done. When AMD got into the open driver business, they had zero documentation, partial support for old hardware, zero support for any cards available on the market, and about 3 years of delay between a product's release and reaching OpenGL 2.1 support. No OpenCL, no power management, no hardware decoding.

Now all available chips are documented, there is UVD and dynamic power management virtually across the board, there is OpenGL 3.3 (almost), there is functioning OpenCL support (still needs work), and newly released cards can do OpenGL 3+ out of the box. And everything except the latest generation is almost at binary blob performance (r600g).

Given limited resources, this is incredible. RadeonSI was a driver written from scratch, for a completely new architecture. And here it is, with almost feature parity with the mature r600g driver, beating Catalyst in some workloads. It still needs some polish, but considering the situation 5 years ago, the progress is incredible.

In a year or two, you'll have out-of the box OpenGL 4+ performance out of the box with every AMD card you purchase. It's just that it can't come overnight.

[bridgman]

It's just that it can't come overnight.

There is still the idea of keeping thousands of "already up the learning curve" zombie programmers frozen in the basement, thawing them out when needed for "overnight" projects and tossing them back in the freezer when the project is done, but we aren't allowed to do that under Ontario labour law.

Best option so far seems to be floating a barge offshore in the San Francisco harbor and doing the work there

[pingufunkybeat]

There is still the idea of keeping thousands of "already up the learning curve" zombie programmers frozen in the basement, thawing them out when needed for "overnight" projects and tossing them back in the freezer when the project is done, but we aren't allowed to do that under Ontario labour law.

...or just taking 50 Windows developers working on Catalyst and making them write kernel modules and work on Mesa and LLVM. That would work so brilliantly...

Seriously, though, I give you guys credit for hiring Marek, Tom, Christian and other community developers with a proven record (wasn't Cooper not also originally a community developer?) It shows that you're willing to expand the team with good people. It's just that I wouldn't mind if you hired 5 extra ones and doubled your team Sure, good GPU developers don't grow on trees, but if Intel can find them, they must exist.

I understand that constraints come from higher up and that FLOSS drivers are not a priority for the big heads. I'm really impressed by the progress that's been done so far. But if it came a bit faster, it would be even better

[bridgman]

Seriously, though, I give you guys credit for hiring Marek, Tom, Christian and other community developers with a proven record (wasn't Cooper not also originally a community developer?) It shows that you're willing to expand the team with good people.

Also Alex and Michel, of course.

... but if Intel can find them, they must exist.

There is a teensy-weensy difference in our revenues and R&D budgets, of course... IIRC we look pretty good relative to company size.

[duby229]

Oh shit don't even start this conversation. It'll end with a "down" and a "size".

[chithanh]

Sure, good GPU developers don't grow on trees, but if Intel can find them, they must exist.

Intel does not even attempt to fully support all their graphics hardware with open drivers.