I find it a little bit weird that:
My AMD 2500+ (1650/2808) only reaches an average of 44 credits per day and has an average turnaround time of 1,45 days and my
AMD 2600+ (1937/3291) gets an average credit of almost 400. Am I doing anything wrong ?
Copyright © 2024 Einstein@Home. All rights reserved.
WU performance
)
What else is running on your system? Do you run other OBINC projects? Check your process list to see what other apliations are taking CPU time. Does it run 24/7?
BOINC WIKI
Well part of it is that the
)
Well part of it is that the 2600 is doing the WUs about 13000 sec faster than the 2500. The RAC depends on when the WU is validated, if the 2600s WUs are validated faster the RAC goes up faster.
You show 39 WUs for the 2600 and 8 for the 2500, big difference...If you look at the view computer in your preferences you will see that the 2500 has WUs that are validated by themselves and the 2600 has numerous WUs that are validated at the same time, so a higher RAC>>>>
Link to Unofficial Wiki for BOINC, by Paul and Friends
I most probably found the
)
I most probably found the issue
Even when my screen was blanking out, it still seemed to process a lot of GUI commands so the task manager showed three times as much as BOINC did.
Anyway, I would like to run BOINC as a service, but I did not find an option to install as service afterwards.
I tried to reinstall but just got a Repair or Remove option. Maybe it would be a good idea to get a change option as well.
I just checked the results and I'm now down to 25000 seconds instead of 35000 per WU before.
RE: Does it run
)
The statistics for the computer is showing % of time BOINC client is running 95.6152 %
which should be quite good,
I was not worried about the low RAC in fact it was more the high calculation time per WU.
RE: I most probably found
)
If you have an updated version to install you can install the later version (such as 4.45 over 4.30). Otherwise, do an uninstall and then install again. The uninstall will not damage (well, should not) your work in progress. There are complete instructions in the Wiki to do this, look in the "How-To" Guides. Pictures and everything ... :)
Another (a new issue) on my
)
Another (a new issue) on my Notebook:
This is a Desktop P4 CPU with HT (3GHz), and it is all crunching Einstein at both "CPU's", the machine is having 512MB RAM and is not doing anything else (more or less)
http://einsteinathome.org/host/354822
What worries me is the really low benchmarks of the CPU:
560.87 floating Point
701.01 integer
I'm runnning BOINC (same version) crunching Einstein on a second machine with a similar configuration:
http://einsteinathome.org/host/388627
P4 3.2GHz HT + 1Gig RAM and the benchmarks on those machines is almost double.
1495.35 floating
1831.82 integer
can anyone give me a hint where to look for reference machines and their performance ?
Oh, by the way, both BOINC's are running as services.
RE: Another (a new issue)
)
Motherboards design's chipset/memory bus etc. can have an effect on performance, and also that you have 1Gig of memory on one computer and 0.5 on the other.
Intel CPU's will also effectively thottle back if they overheat, not uncommon for notebooks.
Also check in taskmanager that einstein is running at 98% or more, and that no other applications are taking cpu resourses.
Hope this helps
Andy
Is it better to return a WU
)
Is it better to return a WU every 8 hours with my Pentium 3.0GHz running Windows XP Professional with 512 Meg of RAM and Hypertreading turned off or with hypertreading on and returning 2 WU every 13.5 hours?
My CPU time has dropped and that is good, but so has my Credits Claimed.
http://einsteinathome.org/account/tasks
Hank
RE: Is it better to return
)
As it is granted credit, which is the 'average', see Paul D Bucks Wiki Manual Credit that is of any significance it is better to have HT switched on as you will crunch more units overall.
Andy
RE: Is it better to return
)
Hank,
This is the question of throughput vs. speed. If you look at your numbers you can do one work unit in 8 hours or 2 in 13.5 hours; which is an *EFFECTIVE* rate of just under 7 hours per work unit.
So, with HT, you take advantage of HT and get a little payoff from it. I have a mix on all my machines and I THINK (no proof as I think of it) that I do a little better than that as I am not always running two Einstein@Home work units in contention with each other. In any case, HT works, and you get more done for the same unit time.
Oh, and the cost of using your computer vs. Einstein@Home will also be less when run with HT on as what you do and what the science applications will be doing are so different ... but with HT off, when you use the CPU you do no EAH work in the back-ground (so to speak).