At VMworld, GRID 2.0 Powered “Tower of Power” Drives Billions of Pixels

by Brian Caulfield

For VMworld this week in San Francisco, we designed and built an enormous “Tower of Power” that serves as a centerpiece for our presence at the show.

Supported by our DesignWorks developer suite and state-of-the-art demo engine technology, our 16-foot tall, four-sided creation showcases how our GRID 2.0 technology can put any application on any device.

The “Tower of Power” — 336 Christie MicroTiles grouped into 56 desk-sized displays — does more than just display a host of advanced, remotely-hosted apps. It makes them dance.

The tower is an imposing presence. Its four walls are 14 tiles tall and six wide. Lining up all 336 of the tower’s 16-inch by 10-inch rear projection tiles would stretch 450 feet. That’s the length of one and a half football fields.

Erecting our "Tower of Power," took nerves, GPUs, and a really tall ladder.
Erecting our “Tower of Power,” took nerves, GPUs and a really tall ladder.

But despite its size, the images on the tower’s screens seem to flit effortlessly across it. Our engineers have figured out ways to send a wave of ripples across these virtual desktops. Or twist these apps into a whirling storm of pixels for a virtual 3D tornado. All across the surface of a display generating more than 7.4 billion pixels a second.

Step One: Putting GRID 2.0 to Work

Our goal: to show how NVIDIA GRID 2.0  can accelerate powerful visual computing apps that can be served up to any display. NVIDIA GRID accelerates virtual desktops and applications, giving enterprises the power to deliver powerful graphics to any user, on any device. Even one the size of a building.

As a result, our “Tower of Power” is a machine that connects to real apps, with real capabilities.  Even as it was being assembled during a quick two week sprint, our “Tower of Power” was getting work done.

In fact, Michael Thompson, part of the small team of engineers who helped assemble and test the tower in a loading dock at our Silicon Valley campus, would use the huge display to beam into his desktop PC on the other side of our campus to update the demo software running the display.

“No way we could get this kind of resolution in our cubes,” the tall, t-shirt clad engineer said last week as he updated the code powering it all while sitting at a folding table just in front of the half-finished display.

The story behind the story: NVIDIA GRID 2.0. All the apps on the display are running on VMWare Horizon virtual machines powered by four HP blades. Each of the four blade servers runs four of our new Tesla M6 GPUs. And each blade server can run 16 virtual machines. That gives us the power to support a total of 64 different virtual desktops.

Step Two: Using Quadro to Put GRID 2.0 on Display

While GRID 2.0 is the engine that makes these apps scream, our Quadro GPUs pours all this content into a remarkable custom display. The four walls of displays are powered by four NVIDIA Quadro GPUs. Another four help take the virtual desktops generated by NVIDIA GRID 2.0 and — using our new DesignWorks developer suite — turn them into pixels we can pick up and play with.

The message behind the monolith: if our technology can support 7.4 billion pixels of worth of virtualized desktops, just imagine what it can do for your enterprise.