doc.zih.tu-dresden.de/docs/jobs_and_resources/slurm_limits.md

@@ -89,3 +89,14 @@ additional threads using the Slurm option `--hint=multithread` or by setting the
 variable `SLURM_HINT=multithread`. Besides the usage of the threads to speed up the computations,
 the memory of the other threads is allocated implicitly, too, and you will always get
 `Memory per Core`*`number of threads` as memory pledge.
+
+## Guidelines for Fair and Proportional Resource Allocation
+
+In shared computational environments, efficient and fair use of resources is essential to
+ensure that all users can benefit from the system. To achieve this, it is critical to allocate
+resources — such as CPUs, GPUs, and memory—proportionally to your actual needs.
+For example, if you request only a single GPU from a node, you should adjust your memory and CPU
+allocations accordingly, rather than consuming disproportionate portions of
+the node’s total resources. This approach prevents resource bottlenecks
+where underused components block other jobs from running,
+resulting in wasted capacity and delayed workflows.