Very true; as Kipz suggested, it has a lot to do with the kernel timings (this can also be said with Windows NT; the internal timings of the o/s may differ from Win2k (see below)). At least we know that the League Servers (4&5) will be ok with the final changes made to the minigun (i.e. they will reflect the true values); but this cannot be said for any non-league linux UT servers that run 136.Psychotic said:If you scroll back a couple of pages, Kippenizjer gives you a hint: something to do with system time under linux differing from the UT time. Result is Linux based servers are not so accurate at getting the tickrate correctly.
In most cases a tickrate of 30 results in a "real tickrate" of 20, but not always. It depends on the Linux/kernel version and the hardware I believe. f.ex. 2v2 Server #4 is set to TR 30 but the "real tickrate" on it is 28-29.
Regarding the kernel / o/s timings - UT is coded to round the timings it receives then sleep for a bit if it isn't the correct time to do something about it; i.e. An event happens at Time Frame 100ms... The UT server will process time until it reaches the next time frame as set by the tickrate; so if the tickrate was 45 the next Time Frame the server would process would be 122.2ms (1000/45=22.2). Now; with Linux the default kernel timing is 10ms, and NOTHING in the world will tell Linux any different, which means that the 22.2ms "deltatime" will be rounded up to 30ms - which explains Psy's table with the Real Timings compared to the Set Tickrate. (Tickrate of 30 returns deltatime of 33.33', which Linux rounds up to 40.... Divide by 2 to get Real Timing Value).
The Windows NT pattern doesn't add up at all though... Even if the timings were rounded down (which they aren't, since it returns an absolute figure, rather than a rounded one); there must be an outside factor affecting the timings.
BTW Are you sure Server3 is still WinNT?