GPU User Manual

FutureGrid's supercomputer, Delta (, is a 16-node GPU cluster running Red Hat Linux, with TORQUE (also called PBS) and Moab for job management, and Module to simplify application and environment configuration. Delta consists of 16 nodes with two 6-core Intel X5560 processors at 2.8GHz, 192 GB of DDR3 memory, and 15TB of RAID5 disk storage. Each node supports 2 nVIDIA Tesla C2070 GPUs with 448 processing cores. For details on Delta's hardware configuration, see the Delta page.

The FutureGrid delta cluster is accessible via a batch queue that is managed from india ( To use delta interactively, first log into india:
Then, on india, the following command lets you use one of the delta compute nodes:
 qsub -I -q delta myprgram
If you want to use delta with your job script, please use
 #PBS -q delta

	to indicate that you'd like to use this queue.

For more details about how to manage queues with qsub, see the Delta manual page.

Utilization of GPU resources on Delta:
1) Utilize GPU node
     Running Program on single GPU node
2) Utilize GPU cluster
     Running MPI/CUDA program on the Delta cluster
3) Utilize GPU cloud
4) Mixing utilization of GPU and CPU
5) Non-trivial service or applications using GPU