Nvidia CEO Jen-Hsun Huang is kicking off the 2012 GPU Technology Conference (GTC) conference right now, and he just announced that Nvidia has developed the world's first GPU for cloud computing, based on Nvidia's Kepler GPU. He's calling it the Nvidia VGX, and described it as "a technology that virtualizes the computing environment such that irrespective of your computing device, we can provide access to the corporate technologies and data that you need." Nvidia envisions deploying the VGX in a data center so that employees can access the power of the virtual machine from any device. It's a pretty scalable solution, as well — a single VGX node can serve up to 100 users simultaneously, and Jen-Hsun Huang demonstrated that live on stage.
In a demo, Nvidia showed off an iPad running Citrix Receiver connected to the VGX virtual GPU and displayed a 3D simulation running in Autodesk on Windows that ran in real time. A second demo showed Grady Cofer (visual effects supervisor at Industrial Light & Magic) using a Macbook Air to remotely access the ILM render farm to make changes to scenes from The Avengers and Battleship in real-time.
The key parts to the VGX that Huang highlighted were the virtualized GPU itself, a low-latency remote display (which allows for the GPU to power a device without having to connect directly), and energy efficiency, so as not to burn through the batteries on mobile devices like the iPad. Each VGX board itself contains four GPUs, each with 192 CUDA cores and 4GB of frame buffer — it also has 16GB of onboard memory and uses the standard PCI Express interface. While Nvidia didn't give any exact dates on when VGX board would be deployed, the company is shooting to have them available for enterprise deployment through Nvidia's hardware partners later this year. Also, if Citrix Receiver isn't your virtual desktop application of choice, you'll be happy to hear that Nvidia's also working with VMware and Microsoft. While this is clearly an enterprise-focused product, Nvidia is also planning to bring gamers cloud-based GPU technology, too: the same low-latency remote display feature is key to the company's new GeForce Grid GPU.
Sean Hollister contributed to this report.