DDR5 Memory Subtimings Tuning: tRFC, tREFI and Read-to-Write Turnaround: Competitive Benchmark Analysis and Hardware Interactions

[Owen Bailey]

Been running a series of controlled benchmarks to figure out where the real-world ROI is on aggressive DDR5 subtimings. My hypothesis is that tightening tRFC and pushing tREFI gives diminishing returns past a certain point in competitive titles, but I want to confirm with the community’s actual data. When you’re already sitting at 300+ FPS, does extending the refresh interval (tREFI to like 65535 or higher) actually hold 1% low FPS stable, or do you see frametime spikes creeping in during longer sessions? And how much of that is board-dependent versus purely CPU IMC limited?

Second question: how are people handling the read-to-write turnaround timings (tRDRD, tRDWR, tFAW) when pushing tight primary timings on dual-rank kits? I’ve noticed that in fast-paced shooters with constant view-angle changes, loosening those occasionally smooths out frametime more than squeezing them for raw latency. Does anyone have empirical frametime graphs showing the difference, or is it mostly noise at these frame rates? I’m trying to separate signal from measurement variance before locking in a profile. What’s your experience been with refresh-induced stalls affecting competitive play specifically?

[Author]

Posts