Virtual GPU workstations were deployed in blade servers for efficient resource sharing.

NVIDIA has introduced the TESLA P6 GPU, built on the company's Pascal architecture, to enable dense GPU-enhanced compute deployments for demanding workloads including graphics-intensive applications and machine learning. The P6 features 2,048 CUDA cores and 16 GB of memory in an MXM form factor, allowing it to slot into blade servers for high-density configurations. Beyond traditional compute applications, NVIDIA is positioning the P6 alongside its P4, P40, and P100 GPUs to support virtual workstation deployments.

The company has developed new software to facilitate this shift toward virtualized workstation infrastructure. While NVIDIA's existing "Grid" software enabled multiple desktops to share GPU resources across both Pascal and Maxwell architectures, the new "Quadro Virtual Data Center Workstation Software" is purpose-built for sharing a single GPU across multiple workstations. According to NVIDIA, the software ensures "encoding of graphics is done on the GPU, not the CPU, so that the two work in harmony to make virtual workstations speedy," with this capability now available on Linux as well. Additional management software tracks virtual workstation performance, allowing administrators to assign resources to individual VMs based on their actual workloads rather than estimates.

NVIDIA points to early market adoption to validate the virtual workstation opportunity. The company notes that "a Japanese auto-maker is moving 10,000 engineers to virtual workstations and Pascal GPUs," suggesting significant enterprise interest. However, the broader workstation market remains diversified: Microsoft last week announced a new version of Windows 10 designed specifically for physical workstations, while HP recently refreshed its physical workstation lineup and introduced a backpack workstation targeting virtual reality exploration.