The memory related VRMs are all outputting nothing and even getting a little bit warm because of that. Vcore is only about 900mV and the die is a little bit warm after touching it with the fingers. However due to the absence of heat I was able to use the ciruit analyser on the powered up PCB. Of curse the board does boot up (because the other GPU), but the card does not. So plugged the "card" (more PCB, less card) in as secondary adapter on a z97 board.
The ciruit analyser read zero ohms between all upper pins of both 6 pins and all lower pins. It does not look very good because the soldering rod was very weak but it did work. I made the workaround to make it start with one 6pin by soldering two wires between the two. So i took it to a friend of mine which is more experienced in electronics than me and at first we took a closer look on the PCB. And then it died and I always got PCIe error when trying to reboot.
Sometime after I needed it as display output, so I put it in my PC and ran superposition to check it. After a while (2-3 months) it made some issures, so I unplugged it and put it back in the warddrobe. Then I plugged the 970 into "my server" and let it compute for a blender renderfarm. I had my GTX 970 for a while and upgraded to a 5700XT. My guess? Something with the PCB, see below Overclocked? Yes, when I used it in my PC. And before that for 2-3 months CUDA compute
Last thing the card did before? superposition benchmark. Tried different Motherboards? Yes, a H81, (my) Z97 and another Z97 all return PCIe error. How old? Have it since the 30th of december in 2015 (so 4 years) What problem? Motherboard does not boot with card plugged in, see below for more So 2x 6pin, clocks like FE, SKhynix GDDR5 What card? GTX 970 Palit (not Jetstream), PCB as picture below In whatever motherboard I put the GPU the MoBo does not post and returns a PCIe error.