I hired one of A100 equipped VM from vultr. It appears to be partitioned temporally, but it failed to run UnifiedMemoryPerf from official CUDA demo suite.
Exception threw at unified memory allocation.
Please clarify the availability of unified memory with vGPU.