You don't need a sample app_info, if you want to run multiple GPU tasks at once use the 'GPU utilization factor of BRP apps' preference in the Einstein project preferences.
I have GTX 960 (GV-N960IXOC-2GD) and I wish to improve its performance for the project.
In the beginning, I didn't tune anything and currently I have a lot of Einstein tasks and all of them work as 1-count (only 1 count/thread concurrently).
Yesterday, I changed all 1 count-values to 0.5 (Einstein project preferences). BOINC is processing them still one by one (1-count). Should I wait new tasks to have 0.5-count processing?
Likewise, there are some additional options - FPS (20), 3D-acc (low), width+height (800x600). Are there some ways to improve performance (according to my video-card)? Because my video-card can manage 1920x1080 processing with high-video-rates, and FPS is higher then 20. I suppose there should be a possibility to improve processing in times.
Yesterday, I changed all 1 count-values to 0.5 (Einstein project preferences). BOINC is processing them still one by one (1-count). Should I wait new tasks to have 0.5-count processing?
Your change in preference for this particular attribute will take effect on your host only when it is assigned and successfully downloads new work. The change will take effect then, not only on the new work, but also on previously downloaded work.
Quote:
Likewise, there are some additional options - FPS (20), 3D-acc (low), width+height (800x600). Are there some ways to improve performance (according to my video-card)? Because my video-card can manage 1920x1080 processing with high-video-rates, and FPS is higher then 20. I suppose there should be a possibility to improve processing in times.
That stuff has only to do with running the screensaver function that gives you some eye candy to watch. It has nothing to do with how work is processed, but running the screensaver does consume system resources--so turning off the screensaver (if you have it running) is the way to get (a little) performance increase in computing Einstein work.
I tried to force BOINC to get new tasks but it didn't. There is some kind of communication, but no new tasks.
I still have ~20 Einstein-tasks for GPU, and I think they will be finished soon (the pace is about one in hour). I suppose after these tasks I will get some new and it will be clear if Einstein gets new options (the processing is still one-by-one).
Although, my intel gpu works 0.5+0.5 for SETI. It means that BOINC can manage 0.5-processing, and currently only question is about GTX and its option-handling.
I still have ~20 Einstein-tasks for GPU, and I think they will be finished soon (the pace is about one in hour).
You currently have a mix of Arecibo and Parkes work for you GTX 960. The Parkes work gets 4.4 times the credit, and will take much longer to run than the Arecibo you have been doing lately. Keeping track of this distinction will be important in assessing the effects of changing how many GPU tasks run at once.
The mind of BOINC is indeed mysterious, but sometimes project imbalance has been the reason for seemingly odd failures to get new work, at least on my machines in the past.
You currently have a mix of Arecibo and Parkes work for you GTX 960. The Parkes work gets 4.4 times the credit, and will take much longer to run than the Arecibo you have been doing lately. Keeping track of this distinction will be important in assessing the effects of changing how many GPU tasks run at once.
The mind of BOINC is indeed mysterious, but sometimes project imbalance has been the reason for seemingly odd failures to get new work, at least on my machines in the past.
Ok, thank you, I will pay attention to it.
And BOINC just started concurrent work for Einstein GPU:
You currently have a mix of Arecibo and Parkes work for you GTX 960. The Parkes work gets 4.4 times the credit, and will take much longer to run than the Arecibo you have been doing lately. Keeping track of this distinction will be important in assessing the effects of changing how many GPU tasks run at once.
The mind of BOINC is indeed mysterious, but sometimes project imbalance has been the reason for seemingly odd failures to get new work, at least on my machines in the past.
Ok, thank you, I will pay attention to it.
And BOINC just started concurrent work for Einstein GPU:
Presumably, it takes time to sync.
I think Einstein only does the multiple units thing once it gets NEW units after the change is made, not immediately.
As the remaining last BFB i
)
As the remaining last BFB i allways try to do the best for the cause... Realy miss the burning bird company.
Just imagine Pretyhot Battleship and her scort fleet sailing with us here... We could be easely break the 2.5MM mark.
RE: You don't need a sample
)
I have GTX 960 (GV-N960IXOC-2GD) and I wish to improve its performance for the project.
In the beginning, I didn't tune anything and currently I have a lot of Einstein tasks and all of them work as 1-count (only 1 count/thread concurrently).
Yesterday, I changed all 1 count-values to 0.5 (Einstein project preferences). BOINC is processing them still one by one (1-count). Should I wait new tasks to have 0.5-count processing?
Likewise, there are some additional options - FPS (20), 3D-acc (low), width+height (800x600). Are there some ways to improve performance (according to my video-card)? Because my video-card can manage 1920x1080 processing with high-video-rates, and FPS is higher then 20. I suppose there should be a possibility to improve processing in times.
bsivko wrote:Yesterday, I
)
Your change in preference for this particular attribute will take effect on your host only when it is assigned and successfully downloads new work. The change will take effect then, not only on the new work, but also on previously downloaded work.
That stuff has only to do with running the screensaver function that gives you some eye candy to watch. It has nothing to do with how work is processed, but running the screensaver does consume system resources--so turning off the screensaver (if you have it running) is the way to get (a little) performance increase in computing Einstein work.
You need to slow down a bit
)
You need to slow down a bit archae.......hard for me to keep up with you on the World stats page
archae86, thank you for the
)
archae86, thank you for the clarification!
I tried to force BOINC to get new tasks but it didn't. There is some kind of communication, but no new tasks.
I still have ~20 Einstein-tasks for GPU, and I think they will be finished soon (the pace is about one in hour). I suppose after these tasks I will get some new and it will be clear if Einstein gets new options (the processing is still one-by-one).
Although, my intel gpu works 0.5+0.5 for SETI. It means that BOINC can manage 0.5-processing, and currently only question is about GTX and its option-handling.
RE: I still have ~20
)
You currently have a mix of Arecibo and Parkes work for you GTX 960. The Parkes work gets 4.4 times the credit, and will take much longer to run than the Arecibo you have been doing lately. Keeping track of this distinction will be important in assessing the effects of changing how many GPU tasks run at once.
The mind of BOINC is indeed mysterious, but sometimes project imbalance has been the reason for seemingly odd failures to get new work, at least on my machines in the past.
RE: You currently have a
)
Ok, thank you, I will pay attention to it.
And BOINC just started concurrent work for Einstein GPU:
Presumably, it takes time to sync.
RE: RE: You currently
)
I think Einstein only does the multiple units thing once it gets NEW units after the change is made, not immediately.