Your WU 165131651 has status 'unknown' - that can't be right.
Here's another one
It's already sent to two new hosts.
I also have one of these, in addition to four successful validations and eleven pendings.
For my "unknown": as two additional copies have been assigned task IDs but are marked "unsent" this may generate extra work. Of course extra work must be performed if there was a validate error to be resolved. But "unknown" status does not seem normal
Your WU 165131651 has status 'unknown' - that can't be right.
Here's another one
It's already sent to two new hosts.
I also have one of these, in addition to four successful validations and eleven pendings.
For my "unknown": as two additional copies have been assigned task IDs but are marked "unsent" this may generate extra work. Of course extra work must be performed if there was a validate error to be resolved. But "unknown" status does not seem normal
Should be fixed now.
Here just in time before the additional instances were sent out.
I also have one of these, in addition to four successful validations and eleven pendings.
For my "unknown": as two additional copies have been assigned task IDs but are marked "unsent" this may generate extra work.
Should be fixed now.
Here just in time before the additional instances were sent out.
Yes, indeed--that one has changed status on the two original tasks to "completed and validated", and the two newly generated tasks originally tagged as "unsent" now show "Didn't need".
I believe credit would have to be 5000 at least to keep with BRP4.
Not sure why we were even asked for comparisons with other projects. Time wasted methinks. Looks like for a lot of GPUs credit has gone down, and that's even with much longer WUs. I'd say the very small but very vocal "Let's keep our credits as low as seti" crowd has won it's only victory ever. They'll have to break out the champagne tonight. At least we can disable running BRP5 for now. Most other projects credits have gone up, here they're going down. Good idea.
Yes, credit is 'disappointing'. Is this an attempt to take GPU crunching credit to CPU levels ? Obviously each project has to follow it's own thoughts, but not really worth the investment in expensive equipment. I make more than this in 1 hour on Milkyway with 1 GPU, and Milkyway doesn't pay the very high credit levels of certain other projects.Just my personal opinion obviously.
First, the most iportand of all: Best wishes for good, stable health!!!
Quote:
from the very short tests on Albert.
I think, it wasn´t to short, but too little chance to catch one of these BRP5 tasks. I tried it for while and got BRP4 tasks for CPU instead of BRP5 tasks for GPU.
I´m crunching with a GT440 and a GTX550Ti, running 2 task at a time.
Taken from 16 successfully crunched tasks at now, I obeserve a very wide spread of run times for the GTX550Ti of 3.4 (the shortest 7.13 , the longest 24.2 [h]). The ratio in run times between the different cards corresdonds to the inverse ratio in crunching power 311/730 [GFLOPS] as awaited. But that would mean at the GT440 a maximum run time of about 2.5 days. That seems to me a bit long.
Forthermore I got 5 succesfuly running tasks, that didn´t show any CPU activity and no Progress bar in BOINC Manager/Tasks. And after I restarted the PC, the Run Time was also restarted, beginning at zero. Looking into the task, I found the hint .
Quote:
... Can't set up shared mem: -1. Will run in standalone mode.
There is still plenty of free memory on the GPU and in RAM at the CPU. That happens with both cards. There is no remarkable longer running time, but there are only a few observations now. I had always 1 task of 4 (=2x2) running this way. Why this? Going to 3 tasks/GPU reduced the GPU load ververely and cluttered it much more. For test I switched back to crunching single tasks. The GPU load increased and the theire variation drastically reduced.Let´s see further.
From MW@H I got 817 +/- 63 [Cobbleston/h], tested on 2780 tasks. The 2 task of BRP5 validated now gave 457 and 333 [Cobbleston/h]. So, for inter project comparison BRP5 are too low.
Well, also for the sake of other parts of the system we had to get the validation going somehow. My focus was mainly on some technical problems that luckily enough none of you actually noticed (what happened is that the first results were uploaded to the wrong server at the "wrong" side of the Atlantic, and all of this while we are frantically working on the next GW search).
Anyway, we'll keep as much of an eye as possible to this credit thing and will probably adjust it as soon as time allows.
All my pending BRP5 tasks
)
All my pending BRP5 tasks (15) were successfully validated, no errors.
I believe credit would have to be 5000 at least to keep with BRP4.
-----
RE: RE: RE: Your WU
)
Too late.
Hm - the "unknown" tasks have sever_state "over" but outcome "initial". This indeed looks odd. Certainly needs investigation.
BM
BM
RE: RE: Your WU 165131651
)
I also have one of these, in addition to four successful validations and eleven pendings.
For my "unknown": as two additional copies have been assigned task IDs but are marked "unsent" this may generate extra work. Of course extra work must be performed if there was a validate error to be resolved. But "unknown" status does not seem normal
RE: RE: RE: Your WU
)
Should be fixed now.
Here just in time before the additional instances were sent out.
BM
BM
RE: RE: I also have one
)
Yes, indeed--that one has changed status on the two original tasks to "completed and validated", and the two newly generated tasks originally tagged as "unsent" now show "Didn't need".
RE: I believe credit would
)
Not sure why we were even asked for comparisons with other projects. Time wasted methinks. Looks like for a lot of GPUs credit has gone down, and that's even with much longer WUs. I'd say the very small but very vocal "Let's keep our credits as low as seti" crowd has won it's only victory ever. They'll have to break out the champagne tonight. At least we can disable running BRP5 for now. Most other projects credits have gone up, here they're going down. Good idea.
Yes, credit is
)
Yes, credit is 'disappointing'. Is this an attempt to take GPU crunching credit to CPU levels ? Obviously each project has to follow it's own thoughts, but not really worth the investment in expensive equipment. I make more than this in 1 hour on Milkyway with 1 GPU, and Milkyway doesn't pay the very high credit levels of certain other projects.Just my personal opinion obviously.
Hallo BM! RE: ... me
)
Hallo BM!
First, the most iportand of all: Best wishes for good, stable health!!!
I think, it wasn´t to short, but too little chance to catch one of these BRP5 tasks. I tried it for while and got BRP4 tasks for CPU instead of BRP5 tasks for GPU.
I´m crunching with a GT440 and a GTX550Ti, running 2 task at a time.
Taken from 16 successfully crunched tasks at now, I obeserve a very wide spread of run times for the GTX550Ti of 3.4 (the shortest 7.13 , the longest 24.2 [h]). The ratio in run times between the different cards corresdonds to the inverse ratio in crunching power 311/730 [GFLOPS] as awaited. But that would mean at the GT440 a maximum run time of about 2.5 days. That seems to me a bit long.
Forthermore I got 5 succesfuly running tasks, that didn´t show any CPU activity and no Progress bar in BOINC Manager/Tasks. And after I restarted the PC, the Run Time was also restarted, beginning at zero. Looking into the task, I found the hint .
There is still plenty of free memory on the GPU and in RAM at the CPU. That happens with both cards. There is no remarkable longer running time, but there are only a few observations now. I had always 1 task of 4 (=2x2) running this way. Why this? Going to 3 tasks/GPU reduced the GPU load ververely and cluttered it much more. For test I switched back to crunching single tasks. The GPU load increased and the theire variation drastically reduced.Let´s see further.
From MW@H I got 817 +/- 63 [Cobbleston/h], tested on 2780 tasks. The 2 task of BRP5 validated now gave 457 and 333 [Cobbleston/h]. So, for inter project comparison BRP5 are too low.
Kind regards and happy crunching
Martin
Well, also for the sake of
)
Well, also for the sake of other parts of the system we had to get the validation going somehow. My focus was mainly on some technical problems that luckily enough none of you actually noticed (what happened is that the first results were uploaded to the wrong server at the "wrong" side of the Atlantic, and all of this while we are frantically working on the next GW search).
Anyway, we'll keep as much of an eye as possible to this credit thing and will probably adjust it as soon as time allows.
BM
BM
On my slightly overclocked
)
On my slightly overclocked Radeon 7850@1050Mhz runtime of BRP5 compared to BRP4 is factor 10 longer.