The rCUDA team is pleased to announce the release of rCUDA 5.0 during the SC'14 conference at New Orleans, LA. The new rCUDA version supports the CUDA Runtime API and CUDA Driver API version 6.5 as well as the routines in cuFFT, cuRAND, cuBLAS, cuFFT, and cuSPARSE.
rCUDA was successfully installed in one of the clusters that University of Almeria owns. The SLURM patch developed by the rCUDA team, which enables the use of remote GPUs, was also installed.
The rCUDA technology has been presented at #ISC14. On the one hand, a talk titled "Increasing the efficiency of your GPU-enabled cluster with rCUDA" was given by Federico Silla at the HPC Advisory Council European Conference 2014, on June 22nd. On the other hand, two different live demos comparing the performance of CUDA and rCUDA on real time were concurrently running at the Mellanox Technologies booth.
This week rCUDA has been present in three different forums: (1) a tutorial entitled "On the use of remote GPU virtualization for managing the GPUs of your cluster in a flexible way" and a poster with title "Boosting the Performance of Remote GPU Virtualization Using InfiniBand Connect-IB and PCIe 3.0" were presented at the IEEE Cluster 2014 conference. (2) A talk entitled "Increasing the throughput of your GPU-enabled cluster with rCUDA" was given at the HPC Advisory Spain Conference 2014, slides and video. (3) Finally, a talk entitled "Extending SLURM with Support for Remote GPU Virtualization" was given at the Slurm User Group Meeting 2014.