There's more ways to Rome than the virtual one!
Let's for the moment asume you want two WUs of Einstein running on your GPU. This can be arranged in two ways:
1. Choose to run GPU-only WUs for Einstein and in the Einstein project settings choose 0.5 as setting for GPU utilization factor of apps in http://einstein.phys.uwm.edu/prefs.php?subset=project.
This will run only two Einstein GPU WUs, but has a severe performance hit as it will not use the CPU most efficiently (it will try to run too many other -CPU- tasks along the GPU ones).
2. Again choose to run GPU-only WUs for Einstein, then via an app_config.xml:
<app_config>
<app>
<name>einsteinbinary_BRP4</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
<app>
<name>einsteinbinary_BRP5</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
<app>
<name>einsteinbinary_FGRP</name>
<gpu_versions>
<gpu_usage>0.5</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>
You can then choose to run the six free cores for another project by suspending all others (except for the nci ones).
The advantage of this approach is that you do not need to 'pay' for the performance hit of running one or more VMs.