>Assuming your FPS was actually something like 358, not 358,596, that's still an extremely high FPS. When your FPS is so high, it becomes difficult to accurately measure the reasons for worst-case performance. A sudden spike within a single frame could be for any of a dozen reasons, such as CPU cache eviction, a GPU pipeline stall, or some other program on your computer taking CPU resources.
As an aside, programmers need to stop using FPS and other inverse units of measurement when performance tuning. It's much easier to reason about 2.79~ milliseconds than it is to reason about 358 FPS. For instance, assuming a single threaded application, let's say I know my game simulation alone can be updated at 125FPS, while my rendering code can be updated at 30FPS. How many FPS does my game run at when I run both together? You'd be hard pressed to figure that out without converting the two to 8 and 33⅓ milliseconds per tick, respectively, making the combination of the two obviously 41⅓ milliseconds, and thus about 24.2FPS.
Yes, your comment about performance measurements under lights loads not accurately reflecting performance under greater loads is true, but 2.79 milliseconds really aren't so far from the standard target of 16.66 milliseconds per frame than the 358 vs. 60FPS comparison would make one guess.
As an aside, programmers need to stop using FPS and other inverse units of measurement when performance tuning. It's much easier to reason about 2.79~ milliseconds than it is to reason about 358 FPS. For instance, assuming a single threaded application, let's say I know my game simulation alone can be updated at 125FPS, while my rendering code can be updated at 30FPS. How many FPS does my game run at when I run both together? You'd be hard pressed to figure that out without converting the two to 8 and 33⅓ milliseconds per tick, respectively, making the combination of the two obviously 41⅓ milliseconds, and thus about 24.2FPS.
Yes, your comment about performance measurements under lights loads not accurately reflecting performance under greater loads is true, but 2.79 milliseconds really aren't so far from the standard target of 16.66 milliseconds per frame than the 358 vs. 60FPS comparison would make one guess.