CDSNA

GPU & DPU Solutions

Use some help reaching your ML, Analytic & HPC goals…

Through our On-demand GPU Reservation models, you can now reserve GPU supercomputer and workstations to meet your workload requirement. Cloud reservation is an option as well as appliance wheeled to your datacenter. With DPUs in the stack, performance will not be a constraint. Learn more on how to combines these possibilities  with no hard requirement for hardware ownership.

 

DPU & Data Acceleration

Through On-demand GPU Reservation models, you can reserve GPU supercomputers and workstations to meet your workload requirements.

Being able to further accelerate the data throughput to enhances your workloads & trainings are now possible via DPUs and SuperNICs.

With BlueField in the stack, performance will no longer be constrained, even when the last “bit” must be delivered for a Synchronous handshake. Learn more on how to combines these possibilities with no hard requirement for hardware ownership and pay-as-needed.

 

DPU BlueField

The HGX Systems will allow most of your workload requirement customizations. But if you are rethinking or reviewing the justification and business cases for running incorporating DPUs in your stack, contact us for a comprehensive discussion.

Contact us today to try our single and multi-instance “Cloud Based DPU” accelerated platform on on-demand. They are available to try as pre-configured instances for high performance workloads, AI training, AI, ML frameworks, all from your personal web browser

  • Experience the advanced capabilities of the NVIDIA DGX/HGX.
BlueField-3 DPU
BlueField-2 DPU
BlueField-3 SuperNIC

SuperNIC

With tens of DGXs and linear scalability, a successfully testing and integrating with the world’s flagship clustered-parallel filesystem, Weka FS, storage stack, the NVIDIA DGX SuperPOD provides a robust solution for scalable AI development and deep learning tasks. Key features include:

  • Streamlined model training directly from Spectrum Scale.
  • Automatic use of local resources as cache to reduce data re-reads over the network.
  • Dedicated workspace for long-term storage (LTS) of datasets.
  • A centralized hub for acquiring, manipulating, and sharing results via standard protocols like NFS, SMB, and S3.

Send a Message

Follow Us

Youtube Linkedin