Amazon Web Services has added an Nvidia Grid GPU instance to its cloud infrastructure-as-a-service portfolio, making GPU acceleration available to cloud developers building graphics-hungry 3-D apps and games.
Amazon's new G2 instance, launched Tuesday, features 1,536 CUDA cores and 4 GB of video RAM. It also includes an Intel Sandy Bridge processor running at 2.6GHz with Turbo Boost enabled, 8 virtual CPUs, 15 GB of RAM and 60 GB of solid state storage, Jeff Barr, chief evangelist at AWS, said in a blog post.
AWS already offers a GPU instance called CG1, which Barr described as "a great fit" for high-performance computing workloads. The new G2 instances include a feature that works well for building cloud-based applications, he said in the blog post.
Nvidia's Grid GPUs, which are based on the vendor's Kepler architecture, come with an H.264 encoding engine that compresses high-resolution graphics and lets them be streamed to any Internet-connected device in real time.
The H.264 encoding engine, combined with the CPU processing power of the G2 instance and AWS storage, messaging and database services, is ideal for building cloud based apps, Barr said in the blog post.
Grid GPUs are widely available from OEMs and resellers, but this is the first time the technology is being made available on a public cloud IaaS, Jeff Brown, vice president and general manager of Nvidia's professional visualization business, said in an interview.
Nvidia launched Grid technology last year and has worked with Citrix, VMware and Microsoft to make it compatible with their hypervisors. Now VMs running on all three can drill through the hypervisor to tap into GPU capacity to speed virtual desktop infrastructure performance, he said.
Nvidia's Grid GPUs are also found in the vendor's Grid Visual Computing Appliance, launched in March, which makes high-end GPU-based graphics workstation performance available to any Windows, Linux or Mac users over a corporate network.
Meanwhile, AWS on Nov. 1 cut pricing on EC2's M3 (Second Generation Standard) instances, which are geared for media encoding, batch processing, caching and Web serving. It's chopping 10 percent off the price of On-Demand M3 instances and 15 percent off Reserved Instance pricing.
PUBLISHED NOV. 5, 2013