I used Slicehost and Linode in parallel until it was obvious that Linode's CPU performance was quite consistent and Slicehost's would waver from marginally better to significantly worse. This is probably because Linode is configured to use a fixed CPU time allocation method.
Amazon's CPU commitments are just as reliable as Linode's, but they seem to be less for the same dollar.
As for actual hardware, to go based on `/proc/cpuinfo`, some of the older instances I manage show up as Xeon L5420 CPUs at 2.5GHz, and the newer ones tend to be L5520 at 2.27GHz for what that's worth. They seem fairly consistent, and if you don't like what you get, just like AWS you can delete the instance and get a new one. You get refunded for your unused time.
Amazon's CPU commitments are just as reliable as Linode's, but they seem to be less for the same dollar.
As for actual hardware, to go based on `/proc/cpuinfo`, some of the older instances I manage show up as Xeon L5420 CPUs at 2.5GHz, and the newer ones tend to be L5520 at 2.27GHz for what that's worth. They seem fairly consistent, and if you don't like what you get, just like AWS you can delete the instance and get a new one. You get refunded for your unused time.