The wiki is deprecated and due to be decommissioned by the end of September 2022.
The content is being migrated to other supports, new updates will be ignored and lost.
If needed you can get in touch with EGI SDIS team using operations @ egi.eu.
GPGPU-WG KnowledgeBase Batch Schedulers Torque
| Main | EGI.eu operations services | Support | Documentation | Tools | Activities | Performance | Technology | Catch-all Services | Resource Allocation | Security |
<< GPGPU Working Group main page The latest production version of [| Torque (version 4.2.6.1) ] is not widely used in the production EGI Grid. Moreover, it is normally used in conjuction with the MAUI scheduler.
Support for GPGPUs was introducted in torque 2.5.6. The number of GPGPUs made available on the client node is
controlled through the torque nodes file on the torque server (normally /var/spool/torque/server_priv/nodes).
For example, to indicate that 2 GPGPUs are available on wn001.example.com, we would set the following in the nodes file:
wn001.example.com np=8 gpus=2
Further information on configuring support for GPGPUs can be found under [Torque GPGPU scheduling]
NVidia Support
If you are using high-end Nvidia GPGPUs (Kepler/Fermi), torque can extract and publish information about the status of these cards. However, you must re-compile torque to support these features. See: [NVidia GPGPU support]. This status information could be quite useful for GIP plugins to determine the availabilty and usage of GPGPUs resources.
Caveat This feature does not work with mid-to-low range GTX cards.
MAUI Support
See: [Torque with MAUI]