Green visual artifacts?

[Melan]

I have been getting this bug on this PC for a while. What happens is that while I am playing, odd, jagged green artifacts flicker on some brush faces (apparently including visportals - that's what I think, anyway). Unfortunately, I haven't been able to take a screenshot; the effect either lasts a very short time, or it isn't captured on shots.

 

What's interesting is that I didn't have it before - might it be caused by turning Catalyst A.I. off to fix the rotating sky bug? I will have to investigate this, but if you have an idea, please post it.

 

I've got an AMD Athlon 3200+ 2.01 GHz, 2 GB RAM and a recently purchased ATI Radeon HD 4800; TDM runs very well outside this problem.

[Myagi]

Sounds like "classic" ati POS. My old computer with a radeon9800 had sparkles over the entire screen when playing demanding games, like doom3. The solution was to clock the core down, using ATI Tool. Inexcusable IMO that a non-overclocked out-of-the-box card needs to be clocked down to work as advertised.

 

For my new box I got a 4850, 2 or so months ago, and I was extremly PO'd when I saw artifacts like you describe, even in Quake bloody 1 (ezquake) on occasional surfaces. Never going to buy ATI again (luckily I only bought it as a temporary solution). I also got the same artifacts when trying TDM. It's not the same as on my 9800, those were single pixel sparkles, the new ones are larger blocks usually green but I think some red sometimes too, confined to surfaces or portals.

 

Long story short, while I haven't gotten around to testing it yet, if possible you could try clocking it down a tad and see if it makes any difference. It's just pure speculation though based on my experience with my 9800.

Edited January 15, 2010 by Myagi

[7upMan]

Myagi, it's interesting that you mention this. I had a ATI 9800SE@Pro back in the days. The 9800 SE was a 9800 chip with some shader units disabled. This could be due to production errors as well as to sell the card at a lower price. My 9800SE was a All in Wonder card with a TV decoder and had the shader units re-enabled by the shop where I bought it (which made her the ATI 9800SE@Pro). However, the shader units must really have had some defects, because not only I got the white noise in Doom 3 you mentioned, but also serious graphics glitches in 'Riddick: Escape from Butcher Bay'.

 

On none other my of ATI cards (X1600, X1900, HD3870, HD4870) I had similar errors, not even the green/red glitches both of you have. I'm using Catalyst 9.10 and stock frequencies for my gc. What are the temperatures of you GPUs? I have AtiTrayTools set up so that my CPU never goes over 80°C, most of the time it stays at 76°C.

[Melan]

Myagi, I have looked into this - there is an "ATI Overdrive" panel in the control center; would I need to downgrade the GPU/memory clocking to a lower MHz frequency, or am I off base with this?

[Baddcog]

Most likely the gpu would need bumped down. Maybe the mem clock too. Keep track of stock (there might be a reset button).

 

Probably try only 10 or 20 at a time.

[Myagi]

My 9800SE was a All in Wonder card with a TV decoder and had the shader units re-enabled by the shop where I bought it (which made her the ATI 9800SE@Pro). However, the shader units must really have had some defects, because not only I got the white noise in Doom 3 you mentioned, but also serious graphics glitches in 'Riddick: Escape from Butcher Bay'.

 

On none other my of ATI cards (X1600, X1900, HD3870, HD4870) I had similar errors, not even the green/red glitches both of you have. I'm using Catalyst 9.10 and stock frequencies for my gc. What are the temperatures of you GPUs? I have AtiTrayTools set up so that my CPU never goes over 80°C, most of the time it stays at 76°C.

 

Interesting, mine was a 9800pro, however original unmodified. If they re-enabled shader untis, that weren't supposed to, it was from factory. In which case that's still unacceptable because I'd personally call that fraud. (IIRC more people from the same brand had those issues.)

 

As for temperature with the new one, that wasn't the issue. Those glitches happened right away, while the temperature was still very low (well below 50°C even I think). I don't remember the exact number but it was among the first thinigs I checked. It even happend in quake1, no pixel shaders or anything that would push the card to its limits or make it break a sweat.

 

In any case I realize that many ATI cards work, which is why I put "classic" in quotes However the statistical abnormally that my last card had issues out of the box, and the very next one I buy 5 years later with new computer / OS / drivers again screws up, was enough to make my head explode. That coupled with the lousy linux support made me have enough of ATI.

 

On a side note, the glitch is no longer there. I don't know if that is because it's colder here for the past month. Just tried with TDM again and saw nothing, GPU temp is at 37°C after running around training mission for ~5 minutes. Either the card just needed to get broken in or the glitch maybe happens when the card has slightly higher temps (in the 40 range). Will have to wait for warmer weather to find out.

 

 

Myagi, I have looked into this - there is an "ATI Overdrive" panel in the control center; would I need to downgrade the GPU/memory clocking to a lower MHz frequency, or am I off base with this?

 

It's just a guess on my part, I don't know if it helps anything at all and can't try because it works atm (see above). But yes that would be were you'd do it. I'd try GPU clock first, then if that dioesn't help, memory. Just a notch, if the default GPU is at 625, try 600 and see if it makes any difference and work from there. I'd also write down the defaults, even if there is a Defaults buttons.

 

Another alternative would be too search the net. Now that we're two with the issues, it's not unlikely there are more and maybe someone already has a verfied solution or knows the cause.

Edited January 16, 2010 by Myagi

[Melan]

I am a bit late, but thanks for your suggestion, Myagi! Turning the GPU clock down didn't have effect, but adjusting memory from 993 to 910 MHz got rid of the glitches. Or so I think, I haven't seen them since!

[7upMan]

Melan, it might be a good idea to check if your graphics card blows the hot air out of the backside of the computer or if the air stays within the case, only to be exausted through the PSU fan. I could imagine that if the temperatures inside the computer case are too high, this could cause the problems you mentioned.

 

On the other hand, Catalyst Overdrive gives you the option to increase the cooler fan speed manually. You could try that and see if the artifacts still remain when you crank the MEM frequencies up to the original value.

[Baddcog]

Some cases allow an extra fan to be put in too, a pretty easy job. 4 screws and a 4 pin molex connector (should be a loose one hanging around).

 

Also, have you cleaned out the dustbunnies lately?

[7upMan]

"Dustbunnies"? LOL they are "wool mice" in German 

[Melan]

And "dust kittens" in Hungarian. WRT my fan, there aren't any in there. The ventilator is set up correctly, air goes out at the back.

[7upMan]

On http://www.cpuid.com/hwmonitor.php you'll find a software called HWmonitor. It's great in that you don't need to install it, just launch the executable. Here you can quickly check your system temps, among lots of other info. GPU core temps and GPU fan speed is measured too. If there's anything wrong with your system that doesn't obviously show, HWmonitor is the tool that most of the time will help you find the issue.

[Myagi]

I did a net search the other day, and I saw some other reports of the same, some RMA'd their cards. My computer was brand spanking new, no dust there, quality PSU and a monster case with huge exhaust fans (aswell as gpu temp being near idle). It happend from the first day for me.

 

I've run the training mission a couple of more times now, also after the room temperature was 4 - 5°C higher than last time, still no artifacts.

 

If it is related to RAM temperature, maybe some of that compund for heat connectivity between chip and heat sink wasn't top notch and melted into place in my case *crosses fingers*

 

 

Good to hear that downclocking appears to work (even if it's annoying having to do it). I would be interested to hear if it was a lasting remedy or just coincidental, after you've gotten the chance to run for some extended amount of time.