.. _Hawk supercomputer:

Hawk supercomputer
==================

.. _Hawk overview:

Overview
--------

`Hawk <https://www.hlrs.de/solutions/systems/hpe-apollo-hawk>`__
is a Tier-0, 26 PFLOPS peak performance,
Hewlett Packard Enterprise Apollo supercomputer at the HLRS,
available to the University of Stuttgart for research.

See the PDF slides linked on the `HPE Hawk Hardware and Architecture
<https://kb.hlrs.de/platforms/index.php/HPE_Hawk_Hardware_and_Architecture>`__
page for more information. See the `HLRS status page
<https://websrv.hlrs.de/cgi-bin/hwwweather?task=viewmachine&machine=hawk>`__
for maintenance information.

+-----------+--------+-------+-----------------+----------+-----------+
| Type      | Nodes  | Cores | RAM             | GPU      | Partition |
+===========+========+=======+=================+==========+===========+
| CPU       | 5632   | 2x64  | :qty:`256 <GB>` |          | ``cpu``   |
+-----------+--------+-------+-----------------+----------+-----------+
| GPU       | 24     | ?     | ?               | 8x A100  | ``gpu``   |
+-----------+--------+-------+-----------------+----------+-----------+

.. _Hawk partitions:

Partitions and nodes
--------------------

Hawk uses the PBS scheduling system!

CPU nodes
^^^^^^^^^
There are 5632 nodes equipped with 2 :cpu-amd-sheet1:`AMD EPYC 7742 <8761>`
(64 cores, 128 threads, :qty:`2.25 <GHz>`, :qty:`200 <W>`), :qty:`256 <GB>` of RAM (:qty:`380 <Gbps>`),
Dual Rail InfiniBand HDR interconnect (:qty:`200 <Gbps>`).
There are no GPUs.

GPU nodes
^^^^^^^^^
There are 24 nodes equipped with 2 CPUs,
Dual Rail InfiniBand HDR interconnect (:qty:`200 <Gbps>`).
Each node has 8 :gpu-nvidia-sheet1:`NVIDIA Tesla A100
<Data-Center/a100/pdf/nvidia-a100-datasheet-nvidia-us-2188504-web.pdf>`.

.. _Hawk access:

Access
------

Please refer to the :ref:`HLRS access policy <HLRS Applying for compute time>`.

Access can also be gained through calls from the Gauss Centre for Supercomputing
(`instructions <https://www.gauss-centre.eu/for-users/hpc-access/instructions-for-hawk>`__).

Once access is granted, refer to the
`Hawk user documentation <https://kb.hlrs.de/platforms/index.php/HPE_Hawk>`__.

.. _Hawk obligations:

Obligations
-----------

Usage of Hawk comes with reporting obligations and acknowledgment requirements.
The specific details vary depending on whether Hawk resources were acquired
through EuroHPC (:ref:`instructions <EuroHPC obligations>`), the :doc:`HLRS
</_pages/_hpc_facilities/hlrs>` (:ref:`instructions <HLRS obligations>`), or
the Gauss Centre for Supercomputing (`instructions
<https://www.gauss-centre.eu/for-users/hpc-access/instructions-for-hawk>`__).

.. _Hawk publications:

Publications
------------

* :cite:`zhong23a`: benchmarking of Tensorflow :abbr:`DNN (Deep Neural Network)` on CS-Storm
  coupled to :abbr:`FEM (Finite Elements Method)` on Hawk
* :cite:`hpcg20b`: Hawk ranked in 18\ :sup:`th` place of the conjugate gradient benchmark