AMD Linux Catalyst: Hardware Owners Screwed?

[allquixotic]

Here we go again.
--snip--
I'd say windows users are getting even more shafted. Difference is, on Linux AMD gives us the tools to fix drivers ourselves.

HD4830 owner here, by the way.

That was an unnecessarily inflammatory post, but I totally agree that Windows users are getting shafted worse. People with an r700 chip aren't going to continue to receive bugfixes and performance improvements that might be attainable for newer games, even if the new games are DX9 or DX10 and just happen to expose a bug in the existing code. AMD kinda hinted that they might occasionally provide driver updates for these situations, but since the support is "dropped", it'd have to be a pretty big-name publisher to get noticed on AMD's radar. In other words, if the Windows version of a Humble Indie Bundle game is crashing to desktop on HD4xxx and they release a fix in a Catalyst update for HD5xxx and later, you ain't getting the fix for HD4xxx. Sorry.

On the other hand, the open drivers can only continue to improve. We still have people putting devel effort into R100, so I'm sure there are many years left in the community developer interest for r600 and r700. Fortunately, r600g already has pretty good feature parity with the other mesa drivers, and it will in all likelihood be receiving OpenCL and Hierarchical Z support sometime in 2012-2013 timeframe. Performance will probably continue to inch up, and you may be able to play new games (or old ones) that would otherwise not be playable with Catalyst. It's nice to have options.

I wonder, if Gallium3d were closer to being able to provide a graphics driver solution on Windows, would the Windows old-Radeon crowd contribute to Mesa enough to enable a WDDM, D3D9/D3D10 open source driver on Windows? Kind of a crazy idea, I know, but if the support issue is such a huge deal, you'd think people would want to leverage the existing code that works well on Linux.

Or, you know, actually run Linux instead of Windows...

[linux5850]

From the questions and answers from bridgman on this forum in the past, I recall that he said AMD released all the info to write open source linux drivers that would perform at 100% of the speed and capabilities of the Radeon cards. I hope that's still true?

[RussianNeuroMancer]

Intel: You can still file bug reports, can't you? I don't know the specific issues of Intel HD GPUs.
You're implying something won't get supported/fixed while there are lots of bug reports, right?

No, I implying there is no big difference between Intel and AMD FOSS drivers. Yes, you still can fill bugreport to R600g developers too.

NVidia: My last NVidia card has been a 6600GT I bought in ~2004. I just checked the driver support.
It's the oldest chipset supported by the non-legacy drivers. That's quite ok!

nVidia doesn't help nouveau developers, so they have no choice but support hardware as long as possible.

[bourbon36]

Well amd just made up my mind for me! was going with a new amd cpu an ati card in a few months , i will have the money soon i'll just go with intel & nvidia this time no more stuffing about. plus intel has better cpu's an nvida has better gpu's. an less probs on linux.

ps: i'm comming home intel & nvida

[diriel]

Problem is....

...My system has an old 3870, and it is honestly beginning to feel a little bit long in the tooth. By the time OSS drivers come up to relative parity for pretty much any 3D type gaming this system will be flat OLD!! I used to keep my "old" systems for guests etc. I am considering taking this one completely down and either donating it to a needy child or putting it in the closet. No need to pull all this power, when 90% of what will ever occur on it once I De-commission it will be web and maybe the rare Writer instance. Frankly, although I did not expect instant gratification, I was hoping for better than we have at this point for OSS drivers.

[Soul_keeper]

bring on the updates !

I just tested the beta and I can confirm that it fixes a texture render problem with VDRIFT
the transparency for the textures of the trees and spectators does not work with 12.4, the beta renders them perfect.

Too bad the beta has the annoying watermark in the bottom corner.

One issue that is not fixed: my 7850 micro-stutters in the Unigine tropics benchmark.
Also i'd really like to see AMD add a function to query free/used memory to their ADL_SDK similar to cuda's cuMemGetInfo() or OpenGL's glGetIntegerv(GL_VBO_FREE_MEMORY_ATI)
It's a pain to implement the OpenGL query in a non-OpenGL progam, and impossible to implement a cuda call.
No function for this functionality is in OpenCL either last I checked.

[RussianNeuroMancer]

Too bad the beta has the annoying watermark in the bottom corner.

Fix here.

One issue that is not fixed: my 7850 micro-stutters in the Unigine tropics benchmark.

AMD Issue Reporting Form.

[Kano]

AMD should really stop showing watermarks and creating hotfixes only validated for a small card range. The reason is that when you want to package these drivers you have to add a signature file from an old driver to remove the testing watermark. This would not be that hard, but as soon as you get a driver only "tested" (whatever that means in amd terms), you need to replace the control file as well. Ok, an old control file could be found as well, but as it contains binary code it looks even worse when you want to add it to a package. Next thing is the aticonfig tool, should somebody replace that as well or binary patch it to ignore the whitelist? Please remove that full watermark/whitelist code for validated hardware, its a piece of shit.

[Gusar]

No, I implying there is no big difference between Intel and AMD FOSS drivers.

Yeah, hardware video decoding (using not shaders but the dedicated ASIC) and proper power management (now that they finally worked out the kinks in rc6) aren't big differences, no siree. For AMD one *can't* even write proper power management, they haven't released the necessary docs.

You suck at this. You really, really do.

[mirv]

Yeah, hardware video decoding (using not shaders but the dedicated ASIC) and proper power management (now that they finally worked out the kinks in rc6) aren't big differences, no siree. For AMD one *can't* even write proper power management, they haven't released the necessary docs.

You suck at this. You really, really do.

It was my understanding not that proper power management couldn't be written, but that it was low on the priority list of the developers. Unless you've got a link to backup your claims?