The limiting factor in your comaprisons should be - the price of the most expensive system [Zedboard]. DMIPS are nice, but DMIPS per dollar appeal to finance sections. If you can buy more than 3 Pi's for the cost of the Zedboard, the performance factor you noted for the Zedboard becomes irrelevant.
If you need the FPGA section to totally thrash the competition, what happens when you give the competition an FPGA too?
So, I'd like to see a Beaglebone Black [BBB] thrown into the comparison mix. Then I'd like to see some kind of FPGA add-on for the Pi and BBB, such as the Logi-bone. The comparison of a Zynq system to a BBB [or Pi ] with a humble Spartan-6 "tacked on" would be... interesting.
@Adam: I think it might be useful for you to post a comment giving a brief explanation of what the whole "mining" thing is about. A lot of folks who aren't familiar with this area may be asking themselves "What does this actually mean?"
I think the concept is that - not unlike gold - bitcoins need to be "discovered" through a process that (intentionally) involves some intensive effort. This process is more like mining than simply manufacturing coins. Maybe someone will come up with a virtual mint to produce virtual representations of bitcoins. :)
Doesn't the Parallela include the FPGA chip from the Zedboard? Have they not yet arrived?
For benchmarking, take the Zedboard data as a Parallela baseline. When the FPGA code gets added, the Parallela may well be able to use that, too. The big difference comes when [if] you add code specifically targeting the Parallela CPUs.
Of course, if you were planning on doing this For Real Money, other questions are: how much better does the Parallela need to be? For every day you delay starting [undelivered Dev board] how far behind do you fall? How long will you need to use the expanded capabilities of the Parallela, before you catch up with those who started TODAY with a standard Zedboard?