bump

This looks like pixel clock generation issue, likely in the analog domain. You'll probably have to carefully inspect some passives around the main chip, maybe even replace a tantalum.

I woult try a bios re-flash.

nuno14272 wrote on 2024-05-16, 12:49:

I woult try a bios re-flash.

I took the opportunity to upgrade to the latest version, but as expected it didn't make any difference.

Tiido wrote on 2024-05-16, 12:01:

This looks like pixel clock generation issue, likely in the analog domain. You'll probably have to carefully inspect some passives around the main chip, maybe even replace a tantalum.

Bummer, I did desolder and check the 5 yellow tantalum capacitors around the main chip and they all measure fine (10-11.5 μF). Also I soldered them back in different positions, just to be extra safe, but nothing changed. Then I checked for shorts (continuity/near zero resistance between the poles) all the small capacitors and resistors in circuit, but none beeped. My inexperienced eyes see nothing suspicious around the chip, if anyone spots something please say so!

the problem are the eletrolitic capacitors.. Those are the ones to change... That's my second guest..

nuno14272 wrote on 2024-05-20, 14:05:

the problem are the eletrolitic capacitors.. Those are the ones to change... That's my second guest..

All 8 are new,

konc wrote on 2024-05-08, 10:49:

I did solder the 4 missing capacitors and while I was at it I changed the other 4 electrolytic as well.

Check this, very similar issue:
http://3dfx.pl/public_html/yabbfiles/Attachme … terferencja.mp4

Repair & fix (pl lng, use translator):
http://3dfx.pl/cgi-bin/yabb2/YaBB.pl?num=1617221455/111#111

Final result in attachment.

smola wrote on 2024-05-20, 18:06:

Check this, very similar issue:

This does indeed look very similar! The only problem in this case is that after replacing all 8 big electrolytic and measuring the 5 big yellow tantalum capacitors, there are only the tiny ones left. And I wouldn't know where to start replacing since there is no obvious short. Are the small ones really capable of causing this behavior? If so I guess I could start slowly replacing them, provided that I also find a component map to get their values.

You can do the finger test, find a place where the finger starts to make the problem worse (avoid the crystal and its capacitors, that will absolutely affect stuff and give a false lead).

Well, that 8 ele caps could be weak quality/bad type. They look like common ones. I'd go for better quality caps, especially for voodoo cards. Usually I use hq poly/tantalum caps, very rare low esr ele type, never common type. You can measure noise level on power lines, but it requires oscilloscope. "Safe" ripple level is, I'd say, under 30mV (good if less than 3% of nominal voltage, best if under 1%) . Low quality/common caps can increase it, it was confirmed few times in the past during other repairs. Also, resoldering same tantalum caps is not good idea, they really don't like high temperatures, and can lose their orig params. They need to be soldered very quick, approx. in 2 seconds max, heating over 3s can damage them - there is pdf with explanation "how to solder" tantalum caps made by vishay (try to find: "Guidelines for Replacing Tantalum Capacitors Using a Soldering Iron"). Good luck!

I suspect Tiido is right, it looks like an analog domain issue. The analog domain on the Avenger is powered from the core voltage, usually 2.7V. The only electrolytic cap on your card that is on that rail would be C51, and perhaps C50 (can't tell if it's tantalum or electrolytic cap).

There are three rails powering the analog domain:
-first rail powers AGP_PLL_VDD (Avenger ball D18)
-second rail powers PLL_AVVD0 and PLL_AVVD1 (Avenger ball B18 and C19)
-third rail powers the DAC_AVDD (Avenger ball C11 and D11)

Each rail is tapped from the 2.7V rail, has a series ferrite bead (600 ohms impedance at 100MHz), followed by : 1x tantalum capacitor, 1x 100nF ceramic cap, 1x 10nF ceramic cap, 1x 1nF ceramic cap. All these parts are located here:

I doubt the ceramic capacitors are damaged, but it's not impossible (cracked ceramic capacitors can fail open, and behave like they're not there, without shorting the rail).
You could check the voltage on both sides of the ferrite beads, and the output voltage of the regulator. You should only see around 2.7V.
As smola indicated, if you have an oscilloscope, measure the ripple on this rail.

Also check for damage on the Avenger IC, it looks a bit weird right there the PLL/DAC supplies are located:

You could also try to replace the 14.318MHz crystal.