Inconsistent numbers for floating point speed?
21 Nov 2012 20:36:29 UTC
Topic 196638
(moderation:
Hi,
I am wondering why I read completely different numbers for the floating point speed of E@H on different web pages. For example on the server status page I read at the moment a total speed of 831.9 TFLOPS, while I read 357.628 TFLOPS on boincstats.com and 1,082.616 TFLOPS on allprojectstats.com. This is really a large discrepancy! Can someone explain to me what that means?
Cheers
Inconsistent numbers for floating point speed?
)
Probably different ways of averaging. Einstein recently got a big boost from a large team; presumably E@H and allprojectstats have relatively short windows for reporting current flops levels, while boincstats has a longer one possibly tied to RAC (which is very slow to respond to change).
It's not as big as the discrepency you saw, but E@H jumped by ~25% in the last week and a half. If the former two sites are looking at short enough windows the big lump of WUs whose return was reported by the latter on the 28th could've triggered a much larger spike in their numbers.
http://boincstats.com/en/stats/5/project/detail/credit