Weird SuperPI calculation

[tiborrr]

Please check this result:

http://hwbot.org/result.do?resultId=606222

 

Note the time lenght of Loops 4-7. A big discrepancies occur, the run looks like it is bugged somehow (perhaps CPU is overclocked a tad too high) therefore should be removed from the database.

 

[K404]

LOL VERY weird loop times. From 0.7 to 2.0 secs (ignoring initial value, Loop1 and loop2)

[knopflerbruce]

Certainly a bug.

[TheKarmakazi]

notice how his watermark blocks the validation too?

[Warrior_oF_Byte]

and it's 1.4 version...

[Massman]

The end result doesn't seem bugged; completely in line with what we'd expect from that frequency. Score is also 2y old, so there are more reasons not to block than to block the score.

 

The bugged loops could be caused by instability or even VGA problems ...

[tiborrr]

I dissagree - the result should be removed, it's not our fault the result got away with it for two years. The best comparison is always given by the Loop 10 (roughly half of the run) which is way off on this calculation.

 

But I will respect hwbot's decision and not cause more coup d'etat

[Massman]

Oh, wait ... now I see the problem. It's not a few bugged loops, but the end time is too fast if you compare it to the other loop results. At loop 10, he's already 1 second behind you whereas in loop 15 he's one second ahead.

 

Blocking the result may indeed be more appropriate given the large variations in loop time.

[TASOS]

Below it's an old score of mine with the same cpu.

 

Notice that the time between loops is not always @2 sec each.

[Hyperhorn]

You didn`t use Super Pi Mod, so this just shows the effect of rounding. Comparison between Mod and Non-Mod-Super-Pi is always questionable IMO.