add cpus-per-gpu docs for sofia

source/brussel/tier1_sofia.rst

@@ -208,9 +208,12 @@ More general information about Globus is available at :ref:`globus platform`.
 Job submission
 --------------
 
-sofia uses the :ref:`Slurm job scheduler <running jobs>` . Only Slurm-native
+**sofia** uses the :ref:`Slurm job scheduler<running jobs>`. Only Slurm-native
 commands are supported for managing your jobs.
 
+Users must specify a Slurm partition when submitting jobs. Loading a
+``cluster`` module is not required.
+
 .. _sofia_job_environment:
 
 Job environment
@@ -232,6 +235,8 @@ If your workflow requires your full shell environment to be propagated, please
 refer to the VUB-HPC documentation on `how to copy your full shell environment into your job
 <https://hpc.vub.be/docs/faq/advanced/#how-can-i-copy-the-login-shell-environment-to-my-jobs>`_.
 
+.. _sofia_job_memory:
+
 Job memory
 **********
 
@@ -248,6 +253,20 @@ allocated because insufficient memory remains available on the node. It also
 helps keep benchmark results representative of production runs by ensuring that
 all jobs follow the same linear memory-per-core allocation.
 
+.. _sofia_job_gpu:
+
+GPU jobs
+********
+
+In the ``zen4_h200`` partition, Slurm jobs are allocated a fixed ratio of
+``24`` CPU cores per GPU (1/8 of the CPU cores on a node).  Job requests
+that do not follow this ratio will be rejected.
+
+This policy avoids situations where GPU credits are available but cannot be used
+because there are insufficient CPU credits. It also helps keep benchmark results
+representative of production runs by ensuring that all jobs use the same
+core-to-GPU ratio.
+
 .. _sofia_help:
 
 Getting help