I already compared the clocks, temperatures, and power consumption of two samples of the same model, the Gigabyte AMD Radeon RX 9070 XT Gaming OC. They differed in memory—the first had Hynix GDDR6 chips and the second Samsung. However, the two tests were months apart. Now I have both units on hand, so I can test performance under identical conditions. We’ll also look at the impact of enabling the fast memory timing option.
Default vs. Fast Memory Timing
Before presenting the results on how memory timing affect performance, we’ll first check how much GPU clock speed and other factors affect card behaviour when switching from the default timing to Fast Memory Timing under a sustained load in F1 24. If the GPU clocks differed significantly, it wouldn’t be fair to attribute performance differences to memory alone.
This is our standard operating-behavior test at 3840 × 2160 (Ultra), five laps on the Singapore track in wet conditions. There’s no drop in load between loops, and the card stays fully warmed up throughout.
Based on the measurements, switching from the default settings to Fast Memory Timing doesn’t meaningfully change the card’s operating behavior. So the measured performance differences can largely be attributed to the timing change.
The differences in the measured operating characteristics are within the usual measurement error.
⠀
⠀
