Important: R9500 Defective pipes = NO!

[fury2nd]

I just wanted to let you all know that the theory about broken pipe lines just became a lot more doubtful as far as i am concerned, because I get the checkers IN WINDOWS TOO! Exact same crruption as in 3dmark and everywhere else!

Any new ideas?

[Hellbinder]

Let me guess...

Is this the result of a modding problem????

[HeXiDeciMaL]

Re: Important: R9500 Defective pipes = NO!

Originally posted by fury2nd
I just wanted to let you all know that the theory about broken pipe lines just became a lot more doubtful as far as i am concerned, because I get the checkers IN WINDOWS TOO! Exact same crruption as in 3dmark and everywhere else!

Any new ideas?

Plzzzzzzzz post some screens

[fury2nd]

OK.. i did not HW mod my card.. i tried wizzard's drivers and rivatuner.. both are the same... i'm gonna reinstall them (i'm running on 9500 now) and post some screens... Hexa can i mail 'em to you cause i don't have any usefull web space?

[{Sniping}Waste]

I don't thinkits a pipe line problem but a problem somewhere else. I moded mine by hard wear and I get no artifacts and yes the mod works, my fill rate is now 2X before mod.

look here for more info
http://www.rage3d.com/board/showthre...eadid=33659687

the pipe lines should not effect the 2d but only 3d.

[fury2nd]

Here are the Screenshots (rarred).

and as {Sniping}Waste.. pipes would not affect 2d!

If you look at the single-texture fiillrate you will see that the artifacts are only on one half of the screen - the lower right, but when i tried another driver set they were on the upper right... wtf?

and also.. i think this checkboards could be a distorted mirror image, check out the nature shots -> especially nature2... and look at the top right part of the shot... the non distorted portion is quite dark, but the checkers are bright whereas the more upper middle part that is not distorted is bright, but here the distorted one is dark...

[prometheusdt]

checkerboard windows xp sp1 desktop


http://pics.bbzzdd.com/users/prometheusdt/checker.JPG

copy paste if link doesn't work

[vtsantzalos]

I ve said it before in other posts, but someone whould try forcing 9500 pro drivers onto the 9500 np instead of forcing the 9700 drivers onto it. Maybe its what other people said differences in registrers, memory access etc. Although it is strange some 9500 work without the artifacts.

[mudwhistle]

Fury2nd: Just wanted to comment, i have tested 6 radeon 9500's so far, and yes, I even notice some of the cards exibit a cheesecloth effect, even in windows, I have not seen the checker effect, just the cheescloth heh.

[fury2nd]

mudwhistle.. ermm.. let's not get technical ok?

And you know what i find interesting.. for example.. a window title bar is corrupter, but if I move the window, the corruptions follows the movement as if it was rendered on the window's title bar!

So I guess this rules out any rendering problem or random artifacts due to too low voltages or high frequencies.. i tried underclocking and it doesn't help!

I'm thinking that the data gets corrupted while being copied to the memory, because if i pull the window off the screen and back on it is forced to redraw itself and the artifact pattern changes!

I believe this is connected to the framebuffer/memory of the card and the way it's accessed (somehow connected to the driver..).. timings maybe or memory addressing issues?

[SHrKY]

I've got a couple observations to add too. The first is that after watching the 3dmark tests for a while the checkerboards seem to appear prominently when the pixel shader is used on a texture.

The other thing that seemed kind of relevent was during the ati three headed dog sceen saver, the checkerboard pattern remains the same size on the same pixels dispite zooming the camera. Could this mean the problem might be occuring post-rendering, somewhere in the video output/ display process?

[fury2nd]

shrky that's exactly what i'm saying... i'm willing to bet my ass that it's something with display memory/framebuffer addressing and/or writing/reading...

[fury2nd]

ok, this **** is really strange!

The SW mod (both of them, all dirver versions) give me the checkers and some 3dmarks, single-texture fillrate raises, BUT THE MULTI-TEXTURE FILLRATE IS LOWER THEN THE SINGLE NOW?!?!

What the hell is this?

[T.T.]

It would be very interesting if people with artifacts (after modding their 9500/128 to a 9700) tried the following:

http://www.xbitlabs.com/news/story.html?id=1042578447
Explains how to modify
* 9500/128MB -> FireGL Z1 (256bit memory+ 4 pipelines)
* RADEON 9700 -> FireGL X1 (256bit memory + 8 pipelines)

I got the information on the FireGL's from Ati: http://www.ati.com/products/builtworkstation.html

So still a chance for the owners of a 9500/128mb ?

[Dexx]

Maybe the 9500 non-pro with 128mb mem is a marketing
ploy - maybe it actually uses 64mb of its 128 when stock,
and activating the other 64mb with a mod enables
faulty chips?

[SHrKY]

Originally posted by Dexx
Maybe the 9500 non-pro with 128mb mem is a marketing
ploy - maybe it actually uses 64mb of its 128 when stock,
and activating the other 64mb with a mod enables
faulty chips?

If that were the case flashing a 9500 Pro 64MB BIOS should solve the problem. But I haven't seen that BIOS around so there's no way to test that theory.

[horseflesh]

that windows desktop pic seemed to have the checkering only in areas that had transparency effects, like the shadows under icon names. Is transparency on the desktop is rendered in hardware?

[Clockwork]

DAMN ATI FOR NOT MAKING THE 9500 WORK LIKE A 9700!!!!!!!!!!!!!!!!

[Tyrant]

I'm not sure if this would shed any light on the matter for people having checkerboard issues, but I'll share my experience with them on a Radeon 8500.

I too have had "checkerboard" effects laid across the entire screen in WindowsXP and 3dMark/actual games. However, they only occurred when I oc'ed the core/mem too high. Sometimes it would start to occur during the 2nd test in 3dMark, which suggests that the card was overheating.

Perhaps someone with a software modded 9500 and artifacts could try installing some ramsinks, and maybe even a better hsf for the core?

[Alpha]

It's probably some latency settings in the bios for memory that can be changed with a tweaked bios to fix the checkerboard affect..

But that's just an uneducated guess


But you have to figure when it's only 4 pipes it's using the memory a little different (faster?) and when you enable 8 pipes it trips once in awhile and causes checkerboard.

[mudwhistle]

Well i'm glad I found 1 card that has zero problems, I just wiped all driver files, registry entries, all files from hdd then reinstalled and am now using riva tuner instead of wizards softmod cause i wanted to test how high i could push this card. My 3dmark is about 400 marks lower with rivatuner but it lets me run @ 4X agp with sideband addressing enabled, zero problems so far. I get around 13600 3dmarks with the card a 304 mem / 337 core. At about 310 mem I get errors and at around 350 core I see errors, not too bad, I'm still pushing the speed of a default radeon 9700 pro. i'm happy, it was worth the effort to keep testing different card.

[kbleft]

Originally posted by {Sniping}Waste
I don't thinkits a pipe line problem but a problem somewhere else. I moded mine by hard wear and I get no artifacts and yes the mod works, my fill rate is now 2X before mod.

look here for more info
http://www.rage3d.com/board/showthre...eadid=33659687

the pipe lines should not effect the 2d but only 3d.

Doesn't winXP use directX to accelerate the gui? If these checkerboards appeared in win2k, the evidence would be much more probative.

[fury2nd]

kbleft.. don't think so..

[fury2nd]

*bump*

seems like the problems only appear when transparency is involved... for example the window shadows and fading menus...

also in 3dmark hi-poly demo (carrusel) i don't see any artifacts on the carrusel itselft, but the reflection (which is done with transparency) is totaly messed up!

[ner02]

Try opening up 2 windows media player 6.4 windows. (mplayer2.exe in Start->Run)

The first one you open grabs the hardware overlay...works fine.

The second one uses some hardware accelerated features of the graphics card...checkerboarding.

[prometheusdt]

check Counter-Strike with direct3d

counter-strike works with direct3d. but when you run it with openGL, it gets checkerboard.

[steve055]

I got my 9500 today. Ran it for a while and got no errors. So i applyed the soft mod. I see no artifacts in 3dmark or UT2003. I do how ever, see them in most of the ati demo's. Of the ones tested the bubble one flickers w/ checkers for only 1 sec 2 sec after it starts. The bear and ocean ones however are more interesting. The bear one shows artifacts untill it goes to the light testing screens. The ambiant lighting test however show none. also the ocean demo only shows artifacts on half the screen and the top half ones have a moving area that remains unaffected