Frametime Analysis

We present a more in-depth analysis than just average FPS to show how the framerate changes over time, which helps highlight FPS drops. Minimum FPS at both the 95th and 99th percentile are reported in these charts, too. A second chart, a histogram, shows shape and spread for the frametime data—how tightly grouped the measurements are. The "IQR" result is called "Interquartile Range," which is an outlier-resistant statistical value that tells us the range in the middle of the frametime distribution.

In the following charts, we are comparing two retail memory kits. The sub-timings are based on the individual XMP profiles and are not adjusted further. By doing so, these charts represent a direct comparison between 6600 MT/s and 8000 MT/s.

Tests are conducted with the following components:
CPU: Intel Core i9-13900K (Locked P-core 5.5 GHz, E-cores 4.3 GHz)
GPU: PNY GeForce RTX 4090 XLR8 VERTO
Memory (1): Colorful CVN ICICLE DDR5-6600 CV16G6600D5VP34S (34-44-44-108)
Memory (2): Patriot Xtreme 5 RGB PVXR532G80C38K (38-48-48-84)




When the game is CPU bound as seen here with low graphic settings, system memory also becomes a sort of bottleneck as well. These charts helps shows how the sub-timings can affect the frame-rate in the 1% lows.




Increase the graphical settings, depending on the game, this may be enough that we are once again GPU bound, meaning each frame rendered is waiting on the graphics card as the CPU tasks are completed faster than the calculations required for the GPU to render the next frame . Here there is a still a small gap between the two memory kits.




Raising the game's resolution to 4K and Cyberpunk 2077 becomes completely GPU bound, leaving the CPU waiting on the graphics card to finish each calculation. With a margin of error of 5%, these two memory kits are indistinguishable at this resolution using a RTX 4090.