Is frame time analysis flawed?

[MyName13]

0.1% and 1% minimums only tell what the worst stuttering will look like, so if two CPUs have same or similar minimums, one is weaker and one is stronger and the weaker one has more stuttering, how will we know if one is worse than another without wave chart (if that's its name)?Wouldn't it make more sense to use 95/97/99 percentiles instead (or along with 1% minimums)?

[manikyath]

the 0.1 and 1% lows were there to replace "a single lowest frame", and i guess these were just the most interesting ones.

 

there's a perfect spot somewhere between just having max/min/low, and having so many bars you're better off just putting up fps graphs, which to the untrained eye is just "a mess"

[MyName13]

7 minutes ago, manikyath said:

the 0.1 and 1% lows were there to replace "a single lowest frame", and i guess these were just the most interesting ones.

 

there's a perfect spot somewhere between just having max/min/low, and having so many bars you're better off just putting up fps graphs, which to the untrained eye is just "a mess"

I understand what you're trying to say with your first sentence, but the problem with average framerates is that they are the same as minimum framerate.Minimum framerate doesn't say anything about frame times, neither does average framerate which means that we need average frametimes (anandtech and linus seem to be using 97 percentiles), why aren't we using average frametimes then?The whole story about 1% lows is that they are far more reliable than minimum framerates and yet nobody talks about average framerates (it's probably possible to have same averages and minimums but not the same average frametimes, for example pentium g4560 performs just like ryzen 1200 but we don't know what their average frametimes look like).

[manikyath]

Just now, MyName13 said:

but the problem with average framerates is that they are the same as minimum framerate.Minimum framerate doesn't say anything about frame times

average and minimum are *way* different cases.

average framerate is all frametimes added up, divided by number of frames.

minimum framerate is the *one* lowest frame, which is more often than not a completely useless number, hence it was replaces by 0.1% and 1% lows.

[MyName13]

6 minutes ago, manikyath said:

average and minimum are *way* different cases.

average framerate is all frametimes added up, divided by number of frames.

minimum framerate is the *one* lowest frame, which is more often than not a completely useless number, hence it was replaces by 0.1% and 1% lows.

Did you even read my post?Average framerate doesn't say anything about frametimes (which means it's equally flawed as minimum framerate) which means that if two CPUs have the same or similar minimum frametimes and average framerates that there might be difference in average frametimes, one might stutter more than the other even though they have the same minimums, do you get it now?

[MyName13]

bump