The problem in HPC is less often pure CPU horsepower though, it is often cache or memory bandwidth, or in the interconnects.
I guess you might be able to build a system in the cloud to provide TOP500 level of performance, but it would be pretty hard even with the fancy EC2 HPC instances (http://aws.amazon.com/ec2/hpc-applications/).
Thanks for pointing out the HPC instances that Amazon has. A few commenters were saying that it's not really a supercomputer without a fast interconnect. Yes, they have that! You just pay more for those instances.
In my experience Amazon did a pretty good job setting things up. It's fun to play around with HPC instances, you can get some sweet performance.
Amazon can. That has no information about how the nodes were allocated. They could have hand picked X rack of nodes that were all connected via the same switch, etc. You don't get that guarantee from AWS.
Fair enough, I guess.... they could have done many things.
Although they do not provide an answer, here are some links to additional info - I spent some time searching for additional info on the Top500 setup, but found little:
I guess you might be able to build a system in the cloud to provide TOP500 level of performance, but it would be pretty hard even with the fancy EC2 HPC instances (http://aws.amazon.com/ec2/hpc-applications/).