I did do that when I ran a Neoscrypt miner test because one of the miners can't tolerate nearly as much overclocking, but for this test what I did was run one of the miners on both machines for a few minutes then tweak the MSI AB settings so that reported hashrate was the same for that miner regardless of which machine it ran on.

*If* we can all agree on a *practical* test methodology here (ie - one that doesn't involve me buying lots more GPUs at 2x MSRP) and the difference in earnings/shares/whatever is less than 1% after 24 hours then for the next iteration I will flip the machines the miners are running on at the 12 hour mark so that any remaining difference in GPU speed will get cancelled out.

But like I said, we need to be realistic about test conditions here. I don't have a mining farm, but I do have a willingness to test in as scientific a manner as possible and that seems to be strangely absent here.