Joe42 said:
Not if we're only interested in p4's.
Add up the clock speed for dual cores to get a total, and that ought to act as a fairly good measure of folding performance, as long at its all the same core/architecture.
Then we can use that to calculate a £/mhz number for each processor to find the best one.
Wouldn't it be easier to keep cool without the rackmount? That way you would have more space, and could have each pc sitting on a shelf in the open air. I know its a bit untidy and such but it ought to be cooler and cheaper...
1. calculating MHz/£ is pointless, it would be better to get an idea of how quickly it does a unit for your chosen project. In essence a two step process of calculating the time per WU and then the WUtime/£.
That is a static captial cost for each node you add but allows the best price per WU.
The reason I say this is that new hardware/nodes will need to have a compariable performance rating.
Next calculate the power consumption cost for a each hour of running for a single node - that is both the power (KW) and the heat output (BTU) per hour. Calculate the cost of electricity per hour and the cost of cooling the heat output per hour.
Then look at the take the time for the average WU in hours (or part of) and then multiply the cost of electricity and cooling to give the cost of a WU in terms of electricity and cooling.
This is your ongoing cost per WU that you will incur for processing it. If you have more than one node then multiply that figure by the number of nodes.
2. Cooling without a rack is ok as long as it's design right, however you will still have the same heat output in BTU/hr being soaked into the room which will need extraction.
In additional you'll need to factor in a cost of replacement parts (hard discs are the norm) and spread that over each WU processing cost or set aside a pool of money to replace dead bits.
You could make a webpage for public display of progress and a paypal donation system for paying for the hosting/electricity.
You'd need to start with a single node, and gain financial backing through the paypal.
At the end of each month you will need to make a judgement call if the project continuation is a go/no go or you (the individual running it) will incur the costs and possible debit (even after recovering money from the sale of the nodes).
Then comes all the complications in that individual's tax... but that would need exploration.