Nvidia’s GPU Technology conference that’s underway this week in San Jose California typically attracts a large swath of engineers, data scientists, content creation and high performance computing (HPC) professionals from many industries, professions and walks of life. Workstation professional types like this typically need a lot of horsepower from very specialized machines that require rigorous software certifications for the high performance design, machine learning and data analytics workloads they’re built to address. As such, Nvidia CEO Jensen Huang wasted no time introducing a new data science workstation platform the company is setting forth as a reference design for its high performance computing OEM partners Dell, Lenovo, HP and others.
At the core of these new systems will be a pair of Nvidia’s pro graphics Quadro RTX 8000 or RTX 6000 Turing-based GPUs with up to 48GB of GDDR6 memory each, connected over its high speed NVLink multi-GPU interconnect, for a total of up to 96GB of contiguous memory space and the ability to address massive data sets without the need to rely on access to main system memory. The systems will also be bundled with Nvidia’s new CUDA-X AI SDK for GPU-accelerated data science applications, with TensorRT for inferencing and cuDNN for Distributed Deep Neural Networks support. Nvidia notes, “Together, these libraries accelerate every step in a typical AI workflow, whether it involves using deep learning to train speech and image recognition systems or data analytics to assess the risk profile of a mortgage portfolio.”
Nvidia also noted that its Data Science Workstation platform will be adopted for production systems by Dell, HP and Lenovo, with smaller boutique and regional system builders like Boxx, Microway and Colfax offering products based on the technology as well.
“When you’re designing a building or bringing a character to life for your movie, the ability to visualize your creation with correct lighting and accurate materials, rather than a low-resolution approximation, dramatically improves your workflow,” said Greg Estes, vice president of Developer Programs at Nvidia.
On the GTC 2019 expo show floor, a number of systems were already on display, including offerings from Dell and Lenovo, live and running a myriad of machine learning and rendering workloads. Nvidia notes systems from their various OEM partners and system integrators will be available in market immediately.
