Author |
Message |
|
Hi!
My pc crunching the two first WUs since 12/21, that's ok...
When i take a look in the BOINC manager, i can see the first WU need more 60 hours to give results, and the second before start need just 1h30 to be finished (that's wrong,..)
My G.C. is an ASUS EN8600GTS 256MB.
To test and see, i let run my pc all time, 24/24 since the 21, and it can give the first of two result just before the time limit.
Question : is my PC not strong enought or the WUs have too short time limit ? (I need to abandoned the second cause time limit over
just 10-15 min. after start)
Thx all et merry christmas ! |
|
|
123bobSend message
Joined: 21 Dec 08 Posts: 7 Credit: 251,750,735 RAC: 0 Level
Scientific publications
|
Same question here. I have an eVGA 8600GTS that I'm trying to use. Either the card is not capable of doing this, the WUs are too big, or the scheduler needs to take into account the slow cards.
I set the WU work queue buffer to 0 days of work. It still gave me 3 other WUs that won't have a prayer at finishing beore the 12/28 due time.
I have one WU that was due at 8 this morning. It still has 15 hours to go. Should I abort it? If so, I wonder if it's worth my time to run this on the 8600GTS, or if I should wait until I get a stronger card....
Regards,
Bob |
|
|
DestiSend message
Joined: 10 Jul 07 Posts: 19 Credit: 1,272,950 RAC: 0 Level
Scientific publications
|
The 8600 is old and not so fast, but it is ok if you can finish at least one wu within the time limit. You should cancel the other workunits so they could be distributed to other clients again.
The GPU support of BOINC is still in the beginning and it does not calculate the needed time right now, so slow systems got always too much work.
____________
Linux Users Everywhere @ BOINC |
|
|
|
Hi !
Right, my G.C. is too old... I cushioned the blow ! ;)
Maybe add a card... I do not prefer to be compelled to leave my PC work just to finish a WU !...
In your opinion, wich solution is better for nature ?
(retire me from gpugrid ?)
Thanks a lot.
--
http://www.boincstats.com/signature/user_1470037.gif |
|
|
|
Hi !
Now i don't receive Wu(s)... Somebody knows why ?
thx |
|
|
|
After searching more, I think that my 8600GTS is definitely not powerful enough ...
I will therefore, I hope, to wait to change the graphics card or WUs more suitable for my type of machine ...
I close the discussion.
Bye! |
|
|
|
Hi all,
I see we have a thread with this problem so I'll add my 2 cents here...
I've been aborting most of the WUs sent to my machine at the advice of the messages appearing in the BOINC manager, i.e., the WU is 1 or 2 days overdue and I likely won't get credit for it. I am assuming that, even if I let it run to completion, that the effort will be wasted since the WU would be sent to someone else and they MIGHT BE ABLE to run in the allocated time.
So, it seems to me that the report deadlines are unrealisticly early for my NVidia Quadro FX 1700 running with my Intel Xeon E5410 2.33 GHz CPU.
If this continues to be the case, I'll just detach from the project since my CPU cycles are being wasted.
Is anybody else, particularly those running the WU server, aware of this problem?
Scott Howard |
|
|
|
So, it seems to me that the report deadlines are unrealisticly early for my NVidia Quadro FX 1700 ...
You find a Quadro FX-overview concerning shaders and CUDA-capability in the link
http://www.nvidia.com/object/IO_11761.html
http://www.nvidia.com/object/cuda_learn_products.html - list of all CUDA enabled cards (1.0, 1.1. 2.0)
which shows us: the FX 1700 offers 32 shaders, but ...
Is anybody else, particularly those running the WU server, aware of this problem?
It was answered in the FAQ (Posted 27 Aug 2008) ...
* (...) Unfortunately Cuda 2.0 only supports Cards that have the compute capability of 1.1 and higher.
Cards that offer only compute capabilities at 1.0 level are lacking a hardware features (support for Atomic functions) that Cuda 2.0 relies on, hence Cuda 2.0 compiled apps won't run.
64 high clocked or 96 moderately clocked stream processors should be fine to participate and give results within a reasonable amount of time. (...)
... and complemented with the link (Posted 4 Jan 2009):
http://developer.download.nvidia.com/compute/cuda/2_0/docs/NVIDIA_CUDA_Programming_Guide_2.0.pdf
See page 85/86 in the pdf, FX1700 supports only CUDA 1.1, with these informations you should not wonder how long it took to finish that one task-ID 200960.
Perhaps such concrete models like Quadro-1.1-cards should be edited in the first FAQ posting (Message 1788) to inform participants before they attempt to run GPUgrid with 1.0/1.1 supported cards. I know if everyone would read carefully ...
|
|
|