Tried with 30/60/90

Oddly enough both 30 and 90 were more accurate, 60 was lower again (488 this time). I believe it's because it's refreshing the pools/miners at the 60 second mark before it wraps up the benchmark, throwing it off a little bit.