Blog Post

And Now Playing, GPU Clouds

Stay on Top of Enterprise Technology Trends

Get updates impacting your industry from our GigaOm Research Community
Join the Community!

Hollywood has always been one of the biggest consumers of large-scale computing systems. Back in the day, it was Silicon Graphics’ Oxygen servers. And now, in keeping up with the times, it is cloud-based computing systems. And not just any cloud. Instead, PEER 1 Hosting has created a specialty cloud using graphics processing units (GPUs) from Nvidia (s nvda).

The new effort was announced at the 37th Annual Siggraph International Conference. The cloud, which is managed by PEER 1 Hosting, runs Nvidia’s RealityServer 3D web application service platform. It uses NVIDIA Tesla GPUs and 3D web services software developed by Mental Images, a division of the chip company. These include graphics rendering, complex quantitative processing, video compression, and large-model 3D web services for access by mobile clients.

While PEER1 and Nvidia talk about this offering a good way for web application developers to add interactivity and 3D capabilities to their services, the fact is that the big consumer of this cloud will be Hollywood. The timing of the cloud is certainly right. Hollywood, like many other industries, is grappling with two major trends — extreme digitization and increasing demand for content via wireless/edge devices. The on-demand nature of this GPU cloud helps Hollywood lower its cap-ex spending on hardware.

This is one of the many speciality clouds we expect are going to be coming to market. Stacey had previously reported about a gaming-oriented cloud developed by AMD (s AMD) and Otoy. AMD owns ATI Technologies, a graphics chips company that competes with Nvidia.

Related GigaOM Pro Content (sub req’d):

One Response to “And Now Playing, GPU Clouds”

  1. Anyone have a clue how these businesses deal with the ‘in and out’ costs? For anything other than a very complex rendering run I would THINK you might be paying more to get your data into and out of the cloud than for the compute cycling… (that is, you have to have really big pipes in your own house and you have to by per Gig to move the data into or out of most clouds…).