ImSpartacus
Regular
Can't edit the above post, but after watching GN's documentation above and the Wasson interview below, I noticed something interesting.
At 7:23, after talking about 99th %tile frame times, Wasson says, "You've converted them back, the frame time, back into FPS, I think?" and Burke responds, "Yes, that's right."
But that's not how I interpret what GN is doing. In their documentation video (which is fucking fantastic to even have in the first place, by the way), at 4:28, Burke mercifully walks through their methodology. He says, "I pull from that [per-frame data] the slowest 1% of frames and, again, the slowest 0.1% of frames." As he says that, he's scrolling through many frame time records. It sounds to me that he's not pulling one singular percentile frame time and then converting it to an equivalent FPS (i.e. 1000/[Frame Time in ms] = [Equivalent FPS]) as Wasson said, but instead calculating a standard FPS using only the bottom 1% (or 0.1%) of frames (as opposed to using 100% of frames like you would in a "standard" average FPS).
Both seem reasonable to me (and I think Wasson would "approve" of both), but it's important to note that 1%tile frame time converted into equivalent FPS will always be a higher FPS than a "1% Low FPS" that uses the 1% worst frame times to calculate FPS (or technically, it will never be lower). If anyone gives a shit, I went full aspie and made myself a quick spreadsheet with pseudo data to illustrate an example of how a converted 1%tile frame time-derived FPS will be different from a "1% Low FPS". I find it much easier to just provide examples rather than try to explain these kinds of concepts in walls of text. As with any creation of this nature, I appreciate any & all feedback and validation.
This all sounds like stupid minutiae, but I think it's an important example of why crystal clear documentation is essential to understanding what in the world you're looking at (and why I didn't assume that I knew what "minimum fps" meant).
