The new rCUDA version is able to ...
| 
| Hits: 2003
/ 2
The new version of rCUDA is easier to be used. Now, the server side of rCUDA can automatically adapt to the network fabric used by the client side. That is, once the rCUDA server is up and running, it can concurrently accept clients that use Ethernet and clients that use InfiniBand. In previous versions of rCUDA, the rCUDA server could only accept either Ethernet clients or InfiniBand clients, depending on the configuration specified when launching the rCUDA server. We will provide more details about the new rCUDA version in next posts.
The new version of rCUDA is born
| | Hits: 1770
/ 0
The rCUDA Team is glad to announce that the new version of rCUDA was born a few days ago. In addition to many bug fixes and minor improvements, the new version of rCUDA integrates three big developments: (1) a new internal architecture intended to provide better support to CUDA applications as well as close-to-native performance; (2) a new communications layer able to get all the bandwidth from the underlying network fabric; (3) support for Slurm. As every baby, the new rCUDA version is small right now. However it is quickly growing. We have been already able to execute a dozen NVIDIA samples. This number grows and grows every day.
rCUDA listed in the TOP 100 Spain Influencers, Blogs, Podcasts & Youtubers in 2020
| | Hits: 1724
/ 0
According to Feedspot, rCUDA is one of the TOP 100 Influencers, Blogs, Podcasts & Youtubers in Spain. We thank Feedspot for these great news!
Three talks about rCUDA worldwide during 2019
| | Hits: 2987
rCUDA live demo in SC19 at Denver
| | Hits: 3062
New rCUDA release is closer
| | Hits: 3296
/ 1
The rCUDA Team is working hard for months in order to get the new rCUDA version ready. We expect to release it within a few months. Nevertheless, we are getting some initial results. The figures next to this text show the performance of the new rCUDA version when moving data, located in pageable memory, to/from a V100 GPU. The figure labeled "V100 H2D pageable" depicts the bandwidth attained when moving data to the GPU. The traditional case (shown in green for CUDA, used locally in the same node where the GPU is) is compared with the performance of the previous rCUDA version and the performance of the new rCUDA version when using a GPU located in a different node. It can be seen that the new rCUDA version provides 2x performance over both the previous version and the original CUDA case. Similar results are achieved when moving data from the GPU back to the host (figure labeled "V100 D2H pageable").
rCUDA now supports Slurm
| | Hits: 1932
/ 0
The rCUDA Team is glad to announce that the support for the Slurm job scheduler has been completed. This support will be included in the next release of rCUDA. With this support, jobs submitted to the Slurm queues can use the remote GPUs provided by rCUDA in a transparent way, without having to modify the applications and without having to modify Slurm.
