If I understand well I get an instance or one thread at 2.17, for 157.68 US$/year.

To have equivalent processing power to one of my 980X rigs running at 4 Ghz I would need to rent 48 threads on Amazon. That equals to a yearly cost of 7'568 US$. If you consider that such top rig like mine can easily run 4 years you get a total of 30'274 US$. Every additional year is another 7'568 US$ on top.

My rig costed around 2'000 US$.

The Amazon approach does financially make sense if you rent no more than three threads, and have no possibility to use a computer.

I took a different approach to the cost justification for running at AWS. I have 3 PC's at home that run 24X7 during the winter months. Best estimates are that they are using ~$50 (US) in electricity per month. In the winter, the extra heat in the house is nice. In the summer, there is an additional charge for air conditioning to keep the house/computers cool. When the outside temperature gets up to 110 degrees farenheit, it gets too hot in my home office so I shut them down. I don't even want to think about what it costs to run the air conditioning.

AWS has a High CPU Extra Large instance that has 7 GiB memory, 8 CPU cores @ 2.5-3.0 ghz per core, 1690 GB disk storage and a 64-bit Ubuntu server operating system. The current spot price is $0.07 per hour. If the instance runs 24X7 for a 30 day month, the cost is $50 per month.

For the cost of what I am spending on electricity, I can keep 8 cores running in the summer when I have to shut down because of the heat. That doesn't replace the capacity that I have at home, but at least it allows me to keep making a contribution.

If I figure in the cost of computers at $2,000 per computer ($6,000) spread over 4 years ($1,500 per year), I could keep another 20 cores running at AWS.

And AWS gets to fix everything that breaks and pay for the air conditioning.

I am currently testing on AWS with a Micro instance. I'm used to working with Ubuntu desktop, but it took me a while to figure out how to work with Ubuntu server. It is all command line interface (no mousing around). The one thing I can say for certain about the Micro instance is that it is slow. I think my wrist watch would be faster. An HCC job ran in about 13 hours. The last job I ran was DSFL and it ran for 88 hours. While it's slow, it is good for getting used to the interface and the operating system.

Now that the weather in the Northern hemisphere is warming up, I'll be starting up a High CPU Extra Large spot instance pretty soon.

I have read that The Roadrunner Supercomputer was taken offline on March 31.

Could this supercomputer be adopted to number crunch WCG projects?

Clive