It’s indeed been about 100ns for a long time. Part of this is that memory is larger though, so there may be more decisions to make to look up a line (and those are made faster). And throughput has improved. Some consumer high-end desktop hardware (think gaming rather than workstation) can have lower latency ram.