I solved shaking with bigger passive cooler on Silicon image chip, fortunately card has some screw holes around.. so i found some bigger chip passive from some vrm module or so.. paint new holes.. found someone willing to drill new holes and its running for hours.
Now im not only sure about memory chips, because they are without any passive, so far no atefacts etc.. but once i got just black screen instead of picture in Unigine Sactuary.. but it was otherwise still rendering regarding of FPS counter.. but because screensaver and monitor shutdown was infolved, memory chips could be fine, this not typical problem for them. In any case, i learned that problem was really Silicon image chip and its cooling - good to know, i theory i can fix other cards with same issue.

Thank you . Good to know . Do all your cards with the affected Silicon Image chips have the same Silicon Image chip model number ?

I am asking, because I have used several cards with Silicon Image TMDS chips on 1600x1200 and 1920x1200 monitors and never seen issues . I am even running one at 1600x1200 70Hz with reduced blanking (to fit under the 162MHz limit. I do have excellent airflow, however.

I would like to compare with the chip models with those on my cards .

Open benches are worse for thermal performance than a closed air case. The pressure created by properly configured intake and exhaust fans in a sealed case will always perform better from a thermal perspective than an open air test bench. At least this was the consensus last time I saw any real testing performed on the theory.

Do all your cards with the affected Silicon Image chips have the same Silicon Image chip model number ?

There is picture of at the start of thread. Chips are fine, they only need proper cooling, i dunno if my card cooling was good enough, or there was some bad contact between cooler and chip. As i said cooling in this era was bad one.

Open benches are worse for thermal performance than a closed air case. The pressure created by properly configured intake and exhaust fans in a sealed case will always perform better from a thermal perspective than an open air test bench. At least this was the consensus last time I saw any real testing performed on the theory.

Well, this is theory, i tneeds to 1 not totally silent fans to have decent air flow (slow ones make not enough flow), big case, cable and card management what is not even possible, you also need some big holes to intake and outtake of air (what means more noise out of case) and more costly on components and energy. These additonal fans cables has "nice " feature that there are easily accidentally disconnected.
For me its not worthy, i just cool only main components a bit more (better) to not need additional airflow. In theory show that flow keep main components a bit cooler, to not need to run their fans faster, but old cards no usually such clever, or you need some additional thermal probes and fan cables and controllers.. and for new components, if noisy water cooling all in one block is solving it.
I kept airflow fans in the case for long time, had some profiles to start them, only with some higher temperature but it was still annoying, last time i did because too high VRM temperature on MB and it has very small effect.