Anyone else getting have the credit for WU's taking twice as long?
CPU time (sec) claimed credit granted credit task
24,700.91 58.71 171.44 _S5R4a_3
14,415.92 237.42 237.42 _S5R3b_0
Using MacPro 8core 2.8g, stock app 6.01
Copyright © 2024 Einstein@Home. All rights reserved.
Why Such Low Credit On New WU's ?
)
Another WU: 42233443 ~ 1/7 of S5R3. :-)
Same here:
)
Same here: http://einsteinathome.org/workunit/42203520
Twice as long and the credit went down...I wouldn't mind if the credit stayed the same as before, decreasing it is a bit of a bummer. But I'll still crunch...
Human Stupidity Is Infinite...
S5R4 search progress Total
)
100 Years?
Woah!
By the way: It was announced the credit would go down in S5R4, 80% with approximately doubled runtime is okay in my opinion.
Optimization of apps - though not as much as in S5R3 - must be taken into account also as much as faster CPUs in the course of the S5R4 run.
RE: S5R4 search
)
That does look suspiciously like a factor of ten, or two, out on some code line, but we'll see .... :-)
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: RE: S5R4 search
)
May be: (317 [from S5R3] + 1.5 [from S5R4]) / 0.00865 = 36820
:-)
Has looked other results (with claimed and granted credit) - "Claimed" != "Granted" and granted near 192 CS. It's OK.
Yes, up to now only a few
)
Yes, up to now only a few S5R4 are being distributed, crunched and returned.
But with the same clients and development of CPU- and App-Performance this run should last nearly thrice as long as S5R3.
My latest achievement: 50776 s on Pentium M 1600 under linux.
S5R3 was around 29000 - 30000 s with the linux-power-apps.
Wow. I'm glad I cut down my
)
Wow. I'm glad I cut down my quota before switching the new app_info in, S5R4's are estimating 1/3rd as long as the R3s.
RE: Wow. I'm glad I cut
)
My estimates ( in seconds ) for one machine are rather varied:
S5R3 ( 4.26 app, 1167.80 freq, 406 seq ) 76980 ( actual, has completed )
S5R4 ( 6.04 app, 228.4 freq, 45 seq ) 94080 ( estimate, when completed )
S5R4 ( 6.04 app, 228.4 freq, 22 seq ) 34560 ( estimate, when completed )
Cheers, Mike.
I have made this letter longer than usual because I lack the time to make it shorter ...
... and my other CPU is a Ryzen 5950X :-) Blaise Pascal
RE: S5R4 search
)
Hopefully a factor of 10 is off on that estimation!
I haven't been keeping up to date with credits and the S5R4 run but if they announced it beforehand that's fine.
Seems to be a general trend to cut back on the credits lately on projects and increase WU's, gearing it more towards the science! :)
And for the record, looks like I've had one complete so far, I keep getting old R3...
S5R4 32,543 seconds (Core 2 Duo 2.2 Mac Intel) 6.01 app...
Human Stupidity Is Infinite...
My first new S5R4 WUs had an
)
My first new S5R4 WUs had an estimated processing time of 4 hours. But in fact they needed over 21 hours on my "Intel(R) Pentium(R) D CPU 3.00GHz".
Now they claim 64 credits (which would be fine for 4 hours but not for more than 20 hours)!
See these WUs: 42177544 and 42199540
Please adjust the timings / credits...
[EDIT]
I also just noticed that misscalculating the estimated run time per WU also leads to an overload of the WU chache...
An other host of mine, with a work chache of 7 days, loaded new WUs based on the estimated run time of 4.5 hours...
but as the WUs are running for more than 20 hours some of the WUs will miss the deadline!
I first noticed this behaviour on an wingman of the host I mentioned first in this post... so this will be a general problem of hosts with larger WU caches...
[/EDIT]
Udo