NVIDIA Announces Enhancements to Quadro vDWS
This week at the GPU Technology Conference in Munich, Germany, NVIDIA announced new capabilities to its Quadro Virtual Data Center Workstation (Quadro vDWS). According to the blog post, "Quadro vDWS with multi-GPU performance enables professionals to work remotely on any device, while their designs and intellectual property are secured in the data center."
In addition to multi-GPU workloads, which allow up to four NVIDIA Tesla GPUs in a single virtual machine (VM) and aggregation of multiple Tesla GPUs and GPU sharing across multiple VMs, live migration on the Quadro vDWS, GRID vPC and GRID vApps with VMware vMotion and vSphere 6.7 u1 is now supported.
"Through our close collaboration with NVIDIA, Red Hat Virtualization will soon be the first virtualization platform to enable multi-GPU workflows in support of NVIDIA's Quadro Virtual Data Center Workstation software. Customers will soon be able to use NVIDIA's newest GPUs on a fully open, enterprise-ready virtual foundation built on the backbone of the world's leading enterprise Linux platform in Red Hat Enterprise Linux," said Joe Fernandes, vice president, Products, Cloud Platforms Business Unit at Red Hat Inc. in a prepared statement.
The 70-watt Tesla T4 GPU will provide "2x the framebuffer" as its predecessor, the Tesla P4, helping virtual desktop infrastructure workflows that require more advanced "rendering, simulation and design."
Deep Learning containers from the NVIDIA GPU Cloud (NGC) for TensorFlow, PyTorch, MXNet, TensorRT and more will now be tested with the Quadro vDWS. These containers include NVIDIA CUDA Toolkit, NVIDIA deep learning libraries and an OS, according to the announcement.
The company said that support for multi-GPU, VMware vMotion and NGC containers is expected in late fall, while NVIDIA Tesla T4 support with NVIDIA vGPU software is expected by year-end.
Wendy Hernandez is group managing editor for the 1105 Enterprise Computing Group.