Non-homogenous redundency.
11 Jan 2006 12:38:51 UTC
Topic 190583
(moderation:
From the FAQ...
Quote:
Each machine that does work will claim a slightly different amount of credit. For fairness and uniformity, we have adopted the BOINC/SETI standard way to determine how much credit to award. Since each unit of work is done by more than one machine, we use the credit claims of all these machines to determine how much credit to award. Every valid result for a given unit of work is then awarded exactly the same amount of credit.
... yes, there can be a slight difference between machines, however, I was investigating why my pending was going up and up and happened to look at the machines that are also receiving the same units as me. Oddly my system seems to be being paired up with PowerMacs, which we all know fly through E@H, and to a lesser extent, AMD's. Net result when I do eventually get credit, I will get a lot less then I claimed, not the "slight difference" stated in the FAQ.
Is an example.
AMD chips and Macs turn these things out much faster than Intel's, perhaps Intel owners should look to other projects which do use HR, or do not use redundency?
If it were to be fair, then you'd expect to see as many results "rounded up" as "rounded down" when in fact, every result I have on my page at the moment is rounded down.
Wave upon wave of demented avengers march cheerfully out of obscurity into the dream.
Non-homogenous redundency.
)
Which is why I asked Bruce to look into changing Einstein@Home over to the new FLOPS counting. He said he would, but, no indication if this is going to become something on their to-do list.
We can but hope ... but you are correct. This is an issue that is of fairly high significance to many of the participants. But, to this point I have not seen much interest from the projects. I know it means more work for them. But ...
I also seem to suffering
)
I also seem to suffering recently, from this problem, and now albert is here the problem only seems to have got worse. When I started here I got what I claimed within 10% or so now I always seem to get granted about 15% below claim.
RE: I also seem to
)
I seem to benefit :) [example]
(I'am the one with the lowest claim.)
Michael
Team Linux Users Everywhere
RE: RE: I also seem to
)
Michael,
Just curious - have you ever tried an optimized Linux BOINC client? I use one of the Windows optimized BOINC clients mainly because it "normalizes" my Seti "Claimed credits" (since I also run an optimized Seti app). A side effect is that my Einstein "Claimed credits" are dramatically skewed to the upside. Since the standard Linux BOINC seems to yield lower "Claimed credits" than Windows BOINC's, an optimised client help might even things up.
Stick
RE: Just curious - have
)
I consider this cheating, because it inflates claimed credit.
Michael
PS: Before I get flamed. I'am not accusing anyone using an optimised BOINC
client of cheating. I just don't like it. I decided to use only stock
clients. This includes the SETI optimised science application. And yes, I'am
aware of all the pro's and con's.
Team Linux Users Everywhere
RE: I consider this
)
I certainly understand your point of view. I am not particularly "into" counting credits anyway. I am more interested in crunching WU's - which is why I run the optimised Seti application. (And I run the opt BOINC client just because I don't like being the one to "bring down" my Seti quorum.)