Each GPU has different performance properties due to the variation in the silicon wafer fabrication process.
As CitricAcid pointed out your video output GPU will always have less performance, as it is also processing video output, even in a headless configuration. My 1080ti’s GPU’s range from 720Sol/s to 790Sol/sec. I do have a rig I converted to deep learning AI via Tensor Flow and it is running full compute mode on all GPU’s and using the mobo video output. However, due to the the way Nvidia implemented the xorg in its drivers, I am uncertain this can be done with a miner… but I am working on it.
You will also notice that each GPU runs at a different temperature. You can blast cool air at them but they will still settle out at different temperatures. Again, due to the variation in the silicon fabrication. You cant really achieve even performance between GPU’s, but you can optimize each GPU’s performance via over clock, memory over clock, and power throttling. Each GPU has a sweet spot that can get you up to 790Sol/sec. Not all GPU’s will go this high but most 1080ti’s easily make 750-760Sol/s.
Power throttling can get you even temperature on your GPU’s. All my 1080ti rigs auto power cycle and seek a set point temp to ±1C.
This is very handy as when the power setting varies outside of daily and seasonal variation, its time to replace the fan or blow the dust off the heat exchange fins.
You can also auto seek the sweet spot for a GPU but you will need to write the code to do this in Windows.