It’s how long it takes the system to render the next frame. High frame times are no good. Equates to lower average fps, and poor player experience. You also want stable frame times. This equates to smooth gameplay and less “stuttering”. Anything under 20ms is considered good. 10ms and less is great. Anything over 50ms will be perceived by the player in a negative way.

I interpret it as the time taken to render a frame. Unlike FPS which is basically a moving average (or rather 1 divided by the average frame time), frame time is a single data point. Collecting frame times allows you to do things like compute the median or, in this case, the lowest 1% of the frame times. That can give you a better idea of how smooth performance appears to the player, and what the worst-case performance is like.

I’m not surprised at the confusion, because they’re using it… not wrong, but very confusingly.

Frame time is literally the time to render a frame. So you’d expect that to be a number of miliseconds per frame and so for lower to be better.

But they’re not looking at frametimes, they’re looking at 1% lows and expressing that in fps, not in frametimes. So yeah, confusing.

For the record, the reson why the term is becoming popular is that there are now widespread visualizations that will give you a line of your frametimes in a graph so you can see if the line is flat or spiky. You’ve probably seen it on the Steam Deck or performance analysis videos or whatever. The idea is that all frametimes being consistent is better than high fps but low 1% or 0.1% low. So stable 60fps can look better than spiky 90fps and so on.