Sheesh I did it again!
This isn't a punitive action. I have a number of projects on the go and I try out different combos. Like moving deck chairs on the Titanic. This one seems to be better placed elsewhere for now. It'll be back. Who really cares where the old girl goes? I know it doesn't contribute very much wherever it is. It's either DC or the rubbish tip. that's why no more Ram. I don't even know if it'll run 256MB sticks even if PC100 wasn't so expensive.

Cheers. ozylynx

???? What value is that Sek?

Bring 'em on carl.... I've also got a GX1 with 4 Dimms so Ram's no problem there, it's 2 Mhz too slow though.... Plenty of 128 and 64 lying about. Still got 16 unused ports in the switch too.

Cheers. ozylynx

The one discussed 1/2/3 posts up in this thread:

<duration_correction_factor>0.795994</duration_correction_factor>

sits in the client_State.xml file. If > 1, your machine is slower than the benchmark, if < 1, your machine is faster. With FA@H only on your machine it's an indicator how fast the effective processing takes place, which could indicate why the claims are higher on a WU than the average.

cheers.

okay, here goes.... your box is a dedicated drone that only can and does FA@H. On the Quorum 3 sample you gave, the machine send in a claim of: 34.75 105 / 22. The value 105 is excess of the 43 by a factor of 244%, thus a far outlier, which under the new rule gets half of the median of the 2 others in quorum i.e the 22.

Based on the benchmark your machine should approximately have claimed a value of 105 / 1.60271 = 65.51 or in different terms, it should have finished the job in 34.75 / 1.60271 = 21.68 hours, at which time it would have claimed the 65.5. Surprise surprise that would have been the expected value for a FA@H.

Now the other machines claiming 43 in that quorum, assuming they only done FA@H for a while, have a DCF of less than 1 (carl.h's list of 10 or so in a row would set an enough stable number). I Suppose Carl.h can look it up his DCF.... effectively his boxes are completing the WU faster than the benchmark suggests..... we knew that already, but he needs to give us his DCF to confirm....the present Benchmark ignores the memory size.

If running in 256mb ram, there will be substantially more swapping than on a 512mb. Why.... well we know that FA@H on a high rammed machine uses 285 peak ram.

I'm getting the picture..... suppose others long figured it out (of course i could always have it entirely by the wrong stick).

cheers

I've looked at a couple of my machines on that basis and each comes much closer to anticipated results. Clever devil you are!!

So it would seem that the fault may be that BOINC isn't applying this factor?
No doubt it can be edited by those who know how and ......
Please, please tell me I'm wrong about that!!!

Cheers ozylynx