Edit karolina-slurm.md

docs.it4i/general/karolina-slurm.md

@@ -131,6 +131,9 @@ The following code requests 16 GPUs on exactly two nodes.
 #SBATCH --nodes 2
 ```
 
+For large jobs that require more than 16 GPU nodes (equivalent to at least 128 GPUs), the "qgpu_big" queue is designated, with a limit of 64 GPU nodes (corresponding to up to 512 GPUs).
+
+
 ## Using Fat Queue
 
 Access [data analytics aka fat node][6].