We do not currently have a published SPEC benchmark.
CPU usage is very dependent on the model of CS-Series array, and the workload, but it's safe to say we run very well with database workloads, which also tend to benefit highly from our in-line compression.
I don't think we have any specific white papers on this topic, however, as there is a great deal of variation based on the workload. Your Nimble sales team would be able to provide more specific information on the expected performance of an array with a specific workload.
To expound a little bit. A lot of questions arise as a result of how efficiently we lay out the writes to our file system. In most instanced the reported IO on a Nimble array is substantially less than that of a traditional Disk array because we coalesce our writes in to a single stripe, effectively reducing the workload of spinning disks to a single sequential write stripe. So where a traditional array may generate thousands of IO/s we generally see in the hundreds or less. Don't be alarmed, its just a well written file system at work.