TFLOPS is still a valid metric for theoretical performance, and when you're comparing the same base architecture, it's quite useful. It's just cores times clockspeed times 2 (because fused multiply add operations give maximum throughput), and you're right that it's hard to compare across architectures. We do normally try to provide some additional information about how TFLOPS isn't the only important metric, but as a base comparison point it at least gives us something.Can we stop using Tflops as a measuring tool in the articles on a tech site? Tflops doesn't mean all that much between architectures, and its very misleading. Vega 64 could hit 12 tflops but a 5700 xt is 25% faster on avg in games while only doing a bit more than 9 tflops. And there are even more drastic examples. It just sounds extremely ignorant, especially from a tech site.