MicrosoftNice to see decent gains like this on production infrastructure without forklift upgrades!
The memory capacity fix may also eliminate an occasional bug for CycleCloud Slurm HBv2 users where HBv2 VMs in a NodeArray might fail to come online for Slurm due to the reported memory not matching the expected memory in the Slurm cluster configuration file.
Lastly, customers targeting thread/core distribution to specific L3 domains should use one of the constrained cores VMs to have full visibility into the topology (which cores share an L3 cache).
