See that's the entire problem - the ratio here is completely uninteresting. The *difference* in speed is what is relevant especially since you're just talking about small overheads at these frame rates and not anything fundamentally scalable (for which ratios might, and I repeat *might* be relevant). .9ms differences could come from almost anything, but they're pretty irrelevant. At "normal" frame rates for actually relevant workloads they go away in the noise.
As I explained above, this is incorrect. You're making a false assumption that the .9 ms from the 1000->9999 fps difference is some scalable part of the system that gets more expensive with stuff that's "more complex". This is almost certainly untrue when you're talking about numbers so small that you're just measuring driver overhead, etc.
But of course, you can't difference FPS numbers because they are reciprocals (which is why they get stupid in upper ranges), but you *can* difference ms/frame meaningfully.
And yes, my point was that overall Win7 vs XP was basically a wash... i.e. your numbers above indicating that Win7 was drastically slower seem odd. Seeing as they are from a forum post though I'd tend to trust the article numbers somewhat more.
Last edited by a moderator:
