(Very) Slow rendering of displaylists

I’m trying to render a displaylist with render-mode GL_SELECT that containing nothing then a huge number of glBegin\Normal\Vertex\End (the whole backend unfortunately need to be restricted to OpenGL 1.2). Running the application in a release build making it close to be not responsible anymore. I’ve tested all also in Debug and with a Mesa driver and in both cases there is a smooth rendering. The application wasting all its time in the nvoglv64.dll as I can easily see by attaching to the process or using ProcessExplorer. Because even the software-renderer is ways faster with the release build (and also spending hours of testing…) I tend to not blame my application anymore. Do you have any idea?
My GPU is a Quadro M1200 and my driver version 431.02 (I’ve tested also 398.01, but got the same result).