Does anyone understand why the arm64 instance types are showing roughly half the performance versus the comparable x64 instances in this? Other benchmarks I've seen put them within 20% of each other. Is this just a hyperthreading vs full core thing?