ATI's Drivers suck

squall_leonhart · April 18th, 2009, 03:15 AM

Quote:

(i know i shouldn't respond...)

and yet somehow their cards still perfom quite well in ogl. how was dx10 supposed to fix an ogl bug?
oh what will nv do when ati releases that lost 50% of performance you mentioned?

routinely, you spew this counter-productive vitriol, with every new ati driver.
why?, no one is sure but you.

in fact, more folks here will begin to ignore you as you repeat this same story monthly.
is that what you want to accomplish?

do you play ogl games on ati cards?
are you losing business because of this alleged terrible bug?

you act as though you have some vested interest and you talk about ati like they were your ex-girlfriend.

why do you care enough to complain and even rant about it?
do you have a goal?

(the above are rhetorical questions on my part, perhaps you should ask yourself)

and yet somehow their cards still perfom quite well in ogl. how was dx10 supposed to fix an ogl bug? -
OpenGL's functions are at this time tied to the development of DirectX, However, the point made at the time was indicating that DirectX applications require texture hacks on ATI cards to work properly, as much as OpenGL titles.

and yet somehow their cards still perfom quite well in ogl. how was dx10 supposed to fix an ogl bug? - You're reading ability sucks. Its already been told that the developers apply hacks on their end to fix issues within the driver.

oh what will nv do when ati releases that lost 50% of performance you mentioned? - Nvidia will be happy to finally have a decent competitor that doesn't require hacks to work. Btw, the 50% loss is incurred at development, and is worked around. ATI are only making it harder for softfirms to make games that work on both cards, having to provide a seperate code path for a number of effects, simply to not suffer the slowdown on ATI cards.

routinely, you spew this counter-productive vitriol, with every new ati driver.
why?, no one is sure but you. Google is your friend, as are various opengl and directX developers. They are as sure as i am.

in fact, more folks here will begin to ignore you as you repeat this same story monthly.
is that what you want to accomplish? Let the trolls ignore, they are the ones deluding themselves in the end.

do you play ogl games on ati cards?
are you losing business because of this alleged terrible bug? No, but i do engage in opengl and directx development for these cards which is where the problems exist the worst.

why do you care enough to complain and even rant about it?
do you have a goal?
The gaming industry is fucked enough as it is, without ATI making it harder on softfirms to produce decent games. You wonder why ATI's Best played program has hardly any softfirms engaged..

mudlord · April 18th, 2009, 03:28 AM

Quote:

why do you care enough to complain and even rant about it?
do you have a goal?

Yes. Because I develop on the OpenGL API. So this is a big issue: driver quality.

Note: I can provide proof of concepts to back squall's claims (eg the texture format bug).

squall_leonhart · April 18th, 2009, 03:35 AM

should we invite Shash to join the fray?

mudlord · April 18th, 2009, 03:37 AM

There is enough proof of concepts found in demoscene productions to show issues which clearly occur on ATI cards and not Nvidia cards.

So yes, a demoscene coder's views would be very interesting.

**Regeneration** · April 18th, 2009, 03:47 AM

I don’t think it’s fair to blast away the entire Catalyst driver for its poor OpenGL support. Besides, we don’t know the story behind ATI’s poor OpenGL support, but there are two possibilities:

1. The Khronos group is biased towards Nvidia.
2. AMD is poor with communications with initiators such as the Khronos group.

When Khronos released OpenGL 3, Nvidia came up with a supported driver in three days. AMD came up with OpenGL 3 driver after six months. This fact should lead us to believe that there is some kind of issue between Khronos and AMD, but we don’t know which side responsible for that.

squall_leonhart · April 18th, 2009, 03:52 AM

And its not because the demo's were made for nvidia, rather they were made to fit a the standard calls in both OGL and D3D

squall_leonhart · April 18th, 2009, 03:55 AM

Quote:

Originally Posted by Regeneration

I don’t think it’s fair to blast away the entire Catalyst driver for its poor OpenGL support. Besides, we don’t know the story behind ATI’s poor OpenGL support, but there are two possibilities:

1. The Khronos group is biased towards Nvidia.
2. AMD is poor with communications with initiators such as the Khronos group.

When Khronos released OpenGL 3, Nvidia came up with a supported driver in three days. AMD came up with OpenGL 3 driver after six months. This fact should lead us to believe that there is some kind of issue between Khronos and AMD, but we don’t know which side responsible for that.

Nvidia is active in opengl development, while ATI is not (usually coming up with similar but incompatible extensions in their drivers)

However, The issue is not limited to OpenGL, DirectX Texturing and Shading is also a problem, and has been since the Radeon 8500 days

I might ask Pete Bernert to make an appearance here since he has POC to demonstrate the issues between the cards.

mudlord · April 18th, 2009, 04:10 AM

Figured I might post a POC for own of the projects I am collaborating with:
http://code.google.com/p/glidehqplus...trunk/Glitch64

If you read through main.cpp and textures.cpp, it gives away a texture format bug. You may notice the "ATI_SUCKS" in the code. That is to work around a texture format slowdown issue. The exact texture format can be noticed in textures.cpp

In fact.....

Code:

 case GR_TEXFMT_ARGB_1555:
                if (ati_sucks > 0) return -1; // ATI sucks as usual (fixes slowdown on ATI)
                factor = 2;
                *gltexfmt = GL_RGB5_A1;
                *glpixfmt = GL_BGRA;
                *glpackfmt = GL_UNSIGNED_SHORT_1_5_5_5_REV;
                break;

I dont think we can make it more obvious.....

squall_leonhart · April 18th, 2009, 04:40 AM

even intel's 950 supports ARGB 1555

mudlord · April 20th, 2009, 01:53 AM

Yes, which I find quite interesting.....

So this is a pure driver bug. We tested on other cards, and the code in question works fine. For ATIs, we have a slowdown of 50%, hence the texture format bypass hack.

I find it very interesting why ATI fanpeople haven't responded to make a rebuttal.