What would you conclude from this? This setup maximizes the number of elements which can influences progress.
What is the bottleneck for each device? How many times has this been tried? Are the websites chosen to favour accelerators in one device which are absent in another.
Setups like this are very easy to influence depending on which outcome you want them to have. For a fair comparison, you need to pin down as many variables as possible and then measure, repeatedly, the metrics you want to compare.
There is nothing wrong with doing some "real-life" tests as well, but without more information you cannot conclude anything here.