I have a Dell R730 and a GeForce RTX 3080 TI GPU. I deployed these as a server for AI LLM testing and development. Due to the power requirements of the 3080 TI as well as logistics of power connector on the R730 mobo I was not able to power it from the motherboard of the R730. I purchased a Corsair RM850X and ran two 8-pin PCIe cables from inside the R730 out to the Corsair, which is outside the server. Since the RM850X is not connected to any motherboard it would not provide power even after I connected two 8-pin PCIe power connectors to the 3080 TI. I used a paperclip to short two pins on the 24-pin motherboard connector and it powered up the Corsair PSU and, consequently, the 3080 TI powered up and is now working. For the first time the R730 sees the GPU and I was able to install drivers and have seen HUGE performance improvements in the AI interactions.
I have seen posts and been told that this is not a good long-term plan and that there is even a danger in doing this. Given the constraints that I don't have another server or PC I can use for this project nor do I have an option to get another GPU that has less power requirements at this point I need to make this work with the hardware I have. So my question is--what are my options? This is working. I left it running over night and there have not been any signs of overheating or any other problems. That said, I don't want to play the fool and either ruin the GPU, the server, or both. How can I provide power to this GPU for the long-term in a way that is both safe and provides adequate power to it?