7.12 - speculation time...

[Draku]

I think they will release the driver today.On the AMD Game Forums is posted the Catalyst 7.12 Windows Release Notes : http://forums.amd.com/game/messagevi...&enterthread=y

[yoshi314]

looks like windows users aren't happy with their ati drivers as well.

now that's pretty surprising for a change. here i was thinking that ati simply doesn't care about linux. but it turns out that windows team has its own share of problems.

edit. just looking at the amount of fixes in windows release made me feel kinda jealous. i wonder when will fglrx have that much attention [never?]

[ferreira]

Haha, they got a bunch of crappy fixes, just like us xD
And keep in mind, yoshi314, that most of their fixes are for particular games, while our problems are more general and probebly can't be fixed by some ugly hack...

[yoshi314]

well, taking games aside there's still about 20-something fixes, i guess. i don't know if they're very complex or simple one-liners, for obvious reasons :]

[ferreira]

Yeah, that is the most evil thing about fglrx, when radeon driver refused to build with pageflipping support, I forced it to do that and it's working...
I just want less b0rken AIGLX support... but as long as compositing in KDE4 sucks lemons, I don't care that much.
My Compiz is also badly screwed up

[Svartalf]

looks like windows users aren't happy with their ati drivers as well.

Heh... And here we thought we were the only ones getting the red-headed stepchild treatment.

[Svartalf]

Haha, they got a bunch of crappy fixes, just like us xD
And keep in mind, yoshi314, that most of their fixes are for particular games, while our problems are more general and probebly can't be fixed by some ugly hack...

Keep in mind something else: They're using the same basic code-core with edge hooks to accomodate X11/GLX interfacing as opposed to GDI/Direct3D interfacing. The bulk of the nasty problems we're seeing are either there as lurkers or are are readily present in any OpenGL games or apps under XP or Vista. Even if the leak issue is in their edge code, it's still a design problem with the code core and a regression could happen at any time within the Windows side when they apply some unrelated fix for something else. The same applies to anything else we've seen in the 8.11 driver set.

Combine this with any issues alluded to from the thread over there, it's a mess for anyone using AMD parts right now, I suspect.

[Svartalf]

well, taking games aside there's still about 20-something fixes, i guess. i don't know if they're very complex or simple one-liners, for obvious reasons :]

Some of them are shader hacks. Some of them are memory handling workarounds- either by lowering the overhead involved with the specified mode or by forcibly degrading the AA settings (Often done without telling the user...) to a lower one when there's insufficient host memory to support the multisamples. Some of them are fixes for Crossfire screwups.

Flickering or corrupted rendering of textures is typically due to an "incorrectly" specified shader or similar problem. NVidia's shader compiler on GLSL is a little more tolerant of some things being wrong in your shader specification and will correct your shader code on the GPU machine code side of things. AMD's will do whatever you ask of it in many cases- and if you fail to supply some critical values for some of your calculations, you can end up with their GPU rendering things goofy. There's instances of NVidia's shader compiler doing the same thing and AMD's doing the right things, but it's my understanding that they're fewer and farther between. The end result is that you end up with games going out with busted shaders (they ARE busted...) and the studios not knowing that they have a bug in their game code because they developed on NVidia and didn't ever really test against AMD, or vice versa.

Corruption in a Crossfire mode is usually due to a mis-management of resources (memory contention, etc. within effectively an SMP GPU configuration...).

AA setting problems are usually due to a card not having the resources to attempt the setting, the driver advertising that it can do it (when it obviously can't), and then doing all the wrong things including locking up the title when they ask for the mode it can't do. Sometimes, it's due to the game just bulling ahead with a setting even though it KNOWS that the card can't do it.

Heh... The shader type issues I can forgive- the studios need to a better job of checking their stuff against multiple hardware brands capable of handling the task of playing their game. The Crossfire and AA issues are less forgiveable, with the AA issues being moreso than the Crossfire ones.

[ferreira]

That's cool we can have some really cool and understandable answers on our questions from you, Svartalf. I deeply appreciate that

[yoshi314]

NVidia's shader compiler on GLSL is a little more tolerant of some things being wrong in your shader specification and will correct your shader code on the GPU machine code side of things.

i just had a reminiscence of my own tinkering with custom shaders on playstation emulator.

ati's shader compiler strictness is inhuman :]