Bonus Credit? Did I miss an announcement?
20 Apr 2016 3:34:44 UTC
Topic 198577
(moderation:
This morning (before about 1400Z) one of my lesser PC's was credited with approximatly 1,000,000 more Cobblestones than it had previously held, resulting in an overall RAC bump from about 530,912 to about 633,373. There was no change in rank among EaH crunchers, though, as it seems other members received some similar bump.
Did I miss a big bonus announcement about dedicated participants getting freebies? :)
I've spent the day expecting another "adjustment", but, so far the "bonus" remains.
Anybody tell me what is going on with this?
Stan
Bonus Credit? Did I miss an announcement?
)
Not sure what happened but I ended up with double my normal points than usual.
RE: Anybody tell me what is
)
Seeing as all your computers seem to have GTX 970 GPUs, I suspect you may have benefited from a sudden influx of previously denied credit for BRP6 tasks due to a validator problem as discussed in this thread. If you had a lot of BRP6 tasks that had previously been declared invalid, the sudden reversal of that would give quite a bump in credit. Other than that, I don't know of any other 'bonuses' floating around :-).
Cheers,
Gary.
RE: I don't know of any
)
I keep a daily spreadsheet entry which among other things tabulates awarded credit with an adjustment for pending credit for my account. I have years of observation experience.
While there are some genuine causes of moderate variation, my last 24 hour Delta includes about 500,000 Extra Credit not attributable to changes in pending work.
In my case, the system or systems reporting an anomalously high increase in credit are currently exclusively running Parkes PMPS work with the application reporting on the webpage as "Binary Radio Pulsar Search (Parkes PMPS XT) v1.57 (BRP6-Beta-cuda55)windows_intelx86".
I am quite confident that there has been some sort of credit accumulation anomaly here. In my case, among three systems, to first-order it appears that one may have been affected not at all, one modestly affected, and one greatly affected. I am less confident in the system-level conclusion than in the account level conclusion because I do not maintain a pending adjusted score for each individual system.
The 'bonus' Gary describes
)
The 'bonus' Gary describes would have manifested as a transfer from invalid to valid, rather than the usual pending to valid - if you weren't expecting that, it might not show up in the particular allowance you made for pendings.
Since I'm running exclusively Parkes PMPS BRP6 work too (on intel_gpus here, at a very regular rate), I'll have a look after lunch and see if I was affected too.
Edit - I've had a look around, and I don't seem to have had any extra credit recently. But my stats are still distorted by doing some extra Einstein work as backup for another project which was offline at the end of March.
Yes, it is possible that the
)
Yes, it is possible that the adjustment I had to make to the BRP validators and a necessary revalidation has given some users some extra credits.
It's also possible that some of you will see a lower invalid rate as the validator is now a bit more relaxed on what is in the result files.
RE: The 'bonus' Gary
)
While that is true, and an error in my original post, I happen to recall the (unusually high) invalid count just before the adjustment. It was seven. That would give 28,000 credits, whereas my anomaly is in the neighborhood of 500,000 credits. So the great majority of the anomaly is from something else.
Following up on Christian's musing, I'd guess that the revalidation cycle had the effect in at least some cases of granting credit more than once for the same work, but that this blip is behind us.
In my personal case, the undeserved credit bump is just a bit more than a day's production, so not a huge issue if it is not a continuing problem.
Thank you, gentlemen! Your
)
Thank you, gentlemen! Your info is most helpful.
If the revalidation "look back" were deep enough, I can understand the bump. In my case, the bump gave me credit for around 250 previously invalid BRP6 tasks. Dating back to the my early days of deploying GTX970's, I could easily have that many reassessed.
My current "invalid" issues revolve around one specific GPU which seems more prone than the rest to produce errors and invalids, even though its mem clock is dialed a couple hundred MHz below stock. (I should probably swap that card out!)
Stan
Report in
)
Report in https://einsteinathome.org/node/198576&postid=156681 indicates that revalidating only supposed to affect a few recent tasks. A bug apparently sprinkled significant unearned credits among the deserving.
Stan