Hi Friends of Seti,

i hope my question is easy to answer ...

On my FX6300 HexaCore some WU´s are finished in about 1,5h and other one´s needs more than 4h .... only on wu´s that the cpu alone is computing. on gpu to wu´s need always the same time.

ty Michael

Workunits have different angle ranges.
That means some will finnish faster than others.
But thats similar for GPU.

---

With each crime and every kindness we birth our future.

Also.. That is a Bulldozer CPU. Shared FPU for two cores. I run my FX6100 on 50% of CPUs and get stable consistent run times for tasks. If I run all the cores, the run times increase by about 30% and have a variance of about 15% on their run times. Technically, you do get more work done in a 24 hour period by running all the cores, even if the individual tasks take longer, but I like consistency, so I run on half the cores.



But yes, for MB, it all depends on the angle-range (AR) for each WU for how long it will take anyway.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Ty for the answers. everything ok for me now :-)

@Cosmic_Ocean, i know that my CPU only has 3 FPU´s, so my settings for seti are set as yours. Only 3 Cores have to be used for computing.

Is it possible to see the AR off the WU anywhere ?

greetings

Ty for the answers. everything ok for me now :-)

@Cosmic_Ocean, i know that my CPU only has 3 FPU´s, so my settings for seti are set as yours. Only 3 Cores have to be used for computing.

Is it possible to see the AR off the WU anywhere ?

greetings

You can find the AR for completed MultiBeam workunits in the Stderr output of each workunit.

Cheers.

Just to help you out.. I've taken a look at some of the CPU tasks you've completed.

CPU time: 43,527.55 seconds
WU true angle range is : 0.432710

CPU time: 13,723.06 seconds
WU true angle range is : 0.011160

And then there is this one..
CPU time: 6,081.61 seconds
WU true angle range is : 11.959736


I thought VHAR was supposed to take a lot longer than the ~0.43 AR ones. That one that I picked didn't have a -9 overflow, either.. all zeroes except for one triplet.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Just to help you out.. I've taken a look at some of the CPU tasks you've completed.

CPU time: 43,527.55 seconds
WU true angle range is : 0.432710

CPU time: 13,723.06 seconds
WU true angle range is : 0.011160

And then there is this one..
CPU time: 6,081.61 seconds
WU true angle range is : 11.959736


I thought VHAR was supposed to take a lot longer than the ~0.43 AR ones. That one that I picked didn't have a -9 overflow, either.. all zeroes except for one triplet.

No VLAR`s takes longer.

---

With each crime and every kindness we birth our future.

Now I'm just confused. Because in the random samples I picked, both very low and very high angle ranges took less than half the time as a normal one. So which is supposed to take longer?


Just looking in the task list for my single core machine, I see the obvious differences in the ETA on tasks that haven't started yet:

09:17:17
09:17:36
07:26:54
09:17:31
09:36:15
09:36:46
03:32:09
11:56:59
09:11:35

~0.43 are the ones that are in the mid 9-hour range, and then there's the 3h 32m one (shorty, VHAR?) and 11h 56m one (VLAR?). I always ended up getting L/H mixed up and confused. I've been reading and participating on these forums for the better part of a decade and I thought I had a grasp on it a long time ago..

Sorry about the thread hijack though.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Now I'm just confused. Because in the random samples I picked, both very low and very high angle ranges took less than half the time as a normal one. So which is supposed to take longer?


Just looking in the task list for my single core machine, I see the obvious differences in the ETA on tasks that haven't started yet:

09:17:17
09:17:36
07:26:54
09:17:31
09:36:15
09:36:46
03:32:09
11:56:59
09:11:35

~0.43 are the ones that are in the mid 9-hour range, and then there's the 3h 32m one (shorty, VHAR?) and 11h 56m one (VLAR?). I always ended up getting L/H mixed up and confused. I've been reading and participating on these forums for the better part of a decade and I thought I had a grasp on it a long time ago..

Sorry about the thread hijack though.

The table in http://seticlassic.ssl.berkeley.edu/faq.html#q125 is still a good indication of how much processing is done at various angle ranges. The first two data rows are VLAR, those rows with G entries are midrange, and the last ten rows are the beginning of VHAR (it actually extends indefinitely to higher angle ranges, though IMO anything above about 15 degrees indicates a problem with the way true_angle_range is calculated).

Basically, VHAR means the telescope pointing was being driven fast enough that the time period a star system was within the telescope beam was short and insufficient data was gathered to do analysis at the tightest band widths. That makes them shorties.

--------------------------------------------

Back to Crash Override's system, the "CPU time: 43,527.55 seconds WU true angle range is : 0.432710" case on Task 3411030688 is clearly an anomaly, there are several other CPU tasks at nearly identical angle range with times around 15000 seconds. There isn't enough information to deduce what happened, though.

                                                                  Joe

As always.. thanks for the insight, Joe.

---

Linux laptop:
record uptime: 1511d 20h 19m (ended due to the power brick giving-up)

Thank you Guys for the aswers.

Some WU´s have to be anomaly, because they where stopped and re started often, while my homeserver has get a new setup and hardware configuration. Now my calculate times are different again. the new gpu (ati 7870) calculates much faster then the old gtx610 :-) btw what i´m also don´t understand is the difference between the cpu calculating. i´ve got an amd fx6300, as i said. this is an "6 Six Core" with 3 FPU´s. So i have setup my calculating configuration to use 100% CPU Power by limiting Multicores to 3 ... my idea was that boinc will use only the real 3 Cores and so my stats & computing time will be boosted. but this was a fault ... every thread needs about 4h in this configuration. Yesterday i changed the config back to use 80% CPU Power and fully 6 Cores and voila there comes back the threads with around 2h. now i´m a little bit confused.

Ok the theme about the WU Range is clear, but why does boinc have no small wu´s for computing about 2h in Config with 100% / 3 Cores, but it has in Config with 80% / 6 Cores ?

I think that boinc doesn´t really take the real FPU Cores in the 100% / 3 Core config. Can that be ?

(Again, sry for my bad english, but i didn´t have use it for many years)

greetings Mike