Yes, you paid less for one month, but you had to pay a full month. One could say Joe's datacenter is a ripoff because they don't let you pay just for a couple of days, or hours.
The whole point of EC2 is that there's no minimum fee, you pay only what you use and you even have an API to control that, so you can request 100 machines to deal with a peak that only lasts a couple of hours.
Is it a bad deal for a dedicated machine? Sure, but for the same reason you wouldn't buy a Prius to do the work of a large pickup, not because it's a "ripoff".
Testing methodology and interpretation of the results was pretty sloppy: Only one test run per platform for single threaded performance, comparisons are very round numbers that exaggerate differences.
If you look into the test results, you'll see that I/O differences seem to be responsible for the majority of the differences in test results between the dedicated box and the ec2 instances. CPU does have an impact though, with the dedicated box showing ~2.5x the CPU performance of an amazon instance.
In the end though its the same old story, if you compare one EC2 instance to one dedicated server, the dedicated server is a better by most measures: RAM, CPU, I/O, bandwidth...
Yes, you paid less for one month, but you had to pay a full month. One could say Joe's datacenter is a ripoff because they don't let you pay just for a couple of days, or hours.
The whole point of EC2 is that there's no minimum fee, you pay only what you use and you even have an API to control that, so you can request 100 machines to deal with a peak that only lasts a couple of hours.
Is it a bad deal for a dedicated machine? Sure, but for the same reason you wouldn't buy a Prius to do the work of a large pickup, not because it's a "ripoff".