Will there be Benchmarks from NVIDIA to check if I have completly correct implemented the cards and if my tuning is going the coorect way?
NVIDIA have done some of this on a few platforms I beleive but really since the server specs are so different and the amount of RAM/CPU clock-speed etc changes things so much that the OEM reference guides might be a better reference. These RA’s are jointly developed with NVIDIA and server vendors like Dell:
http://blogs.citrix.com/2014/03/26/dell-reference-architecture-for-citrix-xendesktop-and-nvidia-vgpu-plus-real-customer-case-studies-of-vgpu-in-deployment-for-graphical-workloads-on-the-dell-servers-at-nvidia-gtc/
NVIDIA have a lot of monitoring tools and you can also use third party tools and benchmarks: http://www.xenserver.org/partners/developing-products-for-xenserver/18-sdk-development/136-xs-dev-gpu-tools.html