Improved?

Main source of the news, that possible changes many things in cloud environment so far, is this.

Now, its obvious this could have huge impact on nowadays high-permormance calculations, as well as a first step towards real cloud computing for ordinary people and sole developers.

But let’s start over again and remember few things about cloud computing.

What we can find in wikipedia about it:
Generally, cloud computing customers do not own the physical infrastructure, instead avoiding capital expenditure by renting usage from a third-party provider. They consume resources as a service and pay only for resources that they use. Many cloud-computing offerings employ the utility computing model, which is analogous to how traditional utility services (such as electricity) are consumed, whereas others bill on a subscription basis.

So, how should we live with that? Although, this business model is still under development, many people can enjoy recent developments in this sphere already now.

Improved!

Back to the news.
Previously, only cluster based images for high-performance calculations were available (i.e. multi-core-based CPUs), but from today along with them, people can have GPU calculations on the cloud. This opens many doors to sole developers and small enterprises, as soon as they don’t have easy access to huge data-centers, on the one hand, and spent much money for that on the other. Moreover, it is possible to run (unix-based systems only for now) OS in hardware-virtualized machine (HVM) mode, where you have guest operating system, which aware that it’s virtual, and can communicate with hardware directly (unlike paravirtual mode, where all guest OS calls are process by middle system and passed to the host OS, and only after that it reaches hardware).

Here is specs for such kind of instance with GPU:
The Instances provide 22 GB of memory, 33.5 EC2 compute units, and utilize the EC2 cluster network, which provides high throughput and low latency for HPC and data intensive applications. The company informed that, each GPU instance features two Nvidia Tesla M2050 GPUs, delivering peak performance of more than one trillion double-precision Flops.