> The first benchmarks indicated the plain Archimedes had higher FP throughput than a 16MHz 386 with a 387.
After some poking about, it is clear this statement is not true for any definition of "first" or "benchmark" or even considering confusion between 386 and 387.
In FP terms, the A3000 got 76 kWIPS, compared to a 387/40's 5.7 MWIPS. Scaling to earlier 16 MHz parts would suggest ~1.5 MWIPS, still almost 20 times as fast as the ARM. Even a bare 386DX40 gets 316 kWIPS, scaling suggests very similar performance overall.
It's plausible that an ARM2 would outperform a 386*SX*/16, but that's without the 387SX. No other combination of 38x parts would not outperform it.
After some poking about, it is clear this statement is not true for any definition of "first" or "benchmark" or even considering confusion between 386 and 387.
In FP terms, the A3000 got 76 kWIPS, compared to a 387/40's 5.7 MWIPS. Scaling to earlier 16 MHz parts would suggest ~1.5 MWIPS, still almost 20 times as fast as the ARM. Even a bare 386DX40 gets 316 kWIPS, scaling suggests very similar performance overall.
It's plausible that an ARM2 would outperform a 386*SX*/16, but that's without the 387SX. No other combination of 38x parts would not outperform it.