Amazon has added a new Instance server to satisfy the needs of customers who require high-performance, graphics processing capabilities. The Cluster GPU Instance server is the eleventh instance offering from Amazon and is clearly not targeted at casual users. It comprises of a pair of NVIDIA Tesla M2050 "Fermi" GPU, a pair of quad-core Intel X5570 processors, 22GB of RAM, 1,690GB of local instance storage, and 10Gbps Ethernet. In a blog post, Jeff Barr, Amazon's lead web services evangelist noted that it is possible to get slightly more than a trillion FLOPS per Cluster GPU instance.

Of course, the programming chops to actually tap into the huge number of processing cores--448 of them per M2050--in an efficient manner is another matter entirely. By default, customers can use up to eight of these new Cluster GPU instances, with more accessible by contacting Amazon. Needless to say, it is also the most expensive instance available on EC2 at the moment, costing $2.10 per hour.

For more on this story:
- check out this article at CNET News
- check out this article at VentureBeat
- check out this article at Amazon Web Services Blog

Related Articles:
Amazon unveils new high-memory instance for EC2 
Amazon to offer a year of EC2 usage for free
Embarcadero's new RAD Studio XE suite in September 
Citrix unveils pilot of XenApp in Amazon cloud