Geekbench 6 scores are calibrated against a baseline score of 2500 (which is the score of an Intel Core i7-12700). Higher scores are better, with double the score indicating double the performance.

If you leave worker_node_max_count and _worker_node_min_count parameters with default values, you currently would have no worker node in your cluster.

Benchmark results for the GeForce GTX 1080 can be found below. The data on this chart is gathered from user-submitted Geekbench 6 results from the Geekbench Browser.

In the example below, icgen2host-10-242-64-37 is the only node in your LSF cluser and is served as the LSF management node which is configured not to run users’ workloads.

In addition, since the worker_node_max_count parameter is set to 10, you can only submit jobs that use no more than 10 nodes. Here, we submit a simple sleep job by requesting 2 worker nodes and observe how auto-scaling works behind the scene to dynamically provision two nodes and add to the existing cluster.

Now let’s run a simple MPI job. Go to /home/lsfadmin/shared and this is the directory which is visible to all nodes in your cluster. Threfore, you should place your MPI workloads under this directory. OpenMPI 4.1 is already installed under /usr/local/openmpi-4.1.0 and you can include its bin/ directory to your PATH environment variable.

After the new nodes have been idling for 10 minutes, auto-scaling will destroy the resources and have those removed from your cluster.