Indeed, this is related to our (experimental) addition of a performance rating for the hardware environment we're testing on. This is rendered as the big number on the composite scores page. Theoretically, our benchmarks covering a diverse spectrum of technologies could be useful for measuring hardware and network performance. But a full run of all test implementations takes 111 hours, and I suspect few people who want to measure hardware performance would be willing to wait that long. So we selected a subset to use for hardware measurement, hence the icon you see.