Author |
Message |
|
Hi
I tested F@H GPU with CUDA before coming on gpugrid, the display was fluid, no problem at all.
With gpugrid, the display are not fluid with video or gui's programm
Can we do anything to fluidify the system?
This problem will be resolved in the future?
Thanks |
|
|
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
Hi
I tested F@H GPU with CUDA before coming on gpugrid, the display was fluid, no problem at all.
With gpugrid, the display are not fluid with video or gui's programm
Can we do anything to fluidify the system?
This problem will be resolved in the future?
Thanks
They are more fluid because have very small number of atoms, so the kernel takes less to complete. We know how to improve it at some computational cost, but it will go away alone when faster cards are used.
gdf |
|
|
|
Ok thanks GDF |
|
|
koschi Send message
Joined: 14 Aug 08 Posts: 124 Credit: 792,979,198 RAC: 17,226 Level
Scientific publications
|
My first unit is about to finish in one hour and unfortunately I'm experiencing the same problem. X is lagging like hell, but even worse, connected with that sound is interrupted when I change windows or open menus.
As my card (9800GT, that I bought yesterday just for crunching) won't become faster by itself, there should be maybe another solution that upgrading to a GTX260/280.
Maybe you can modify the application to use n-2 shader or like that, so that some small resources are always available for the OS. Or make the application check for the presence of a file in directory projects/www.ps3grid.net. If, lets say file 'heyIamaheavilyusedworkstation' is present, then slow down the calculation. Otherwise, which would be the default, you can eat up the whole performance ;-)
I'm running Kubuntu Intrepid with kernel 2.6.26.5, BOINC 6.3.8 and driver 177.13 on an Intel Q6600 @ 2.4GHz ...
regards,
koschi |
|
|
TemujinSend message
Joined: 12 Jul 07 Posts: 100 Credit: 21,848,502 RAC: 0 Level
Scientific publications
|
koschi
Maybe you can modify the application to use n-2 shader
or
lets say file 'heyIamaheavilyusedworkstation' is present, then slow down the calculation. Both good ideas there
GDF
We know how to improve it at some computational cost Can you give us some idea of the computational cost?
I'm sure most of us suffering poor X response would welcome an improvement for a little increase in WU times. |
|
|
GDFVolunteer moderator Project administrator Project developer Project tester Volunteer developer Volunteer tester Project scientist Send message
Joined: 14 Mar 07 Posts: 1957 Credit: 629,356 RAC: 0 Level
Scientific publications
|
koschi
Maybe you can modify the application to use n-2 shader
or
lets say file 'heyIamaheavilyusedworkstation' is present, then slow down the calculation. Both good ideas there
GDF
We know how to improve it at some computational cost Can you give us some idea of the computational cost?
I'm sure most of us suffering poor X response would welcome an improvement for a little increase in WU times.
The first is not possible. We have to test to know how much we can improve it and what is the computational cost.
gdf |
|
|
|
i play on the same computer i use ps3grid but games dont run smoothly.
i noticed something else too : ps3grid crashes when boinc starts ps3grid when i'm playing :(
is it possible to have an option that enables ps3grid to run only when screensaver is on ? but i don't want to stop crunching cpu only projects. |
|
|
|
For gaming I suggest a somewhat cumbersome workaround: suspend the BOINC network activity (so that it doesn't decide to fetch new GPU-WUs) and pause the active GPU-WU. BOINC may start the next GPU-WUs immediately, so I'd try to pause them as well. I can't test this method though, because I just ordered my card :)
Ideally it should not be a problem to run CUDA besides DX or OGL (that's how physics is supposed to work on Nv hardware, after all), but practically I can imagine situations where it just doesn't work, e.g. the CUDA app crashes.
BOINC should be prepared for such cases. I'd suggest to have:
- an option in the BOINC GUI to manually pause the CUDA-apps but keep the CPUs crunching [should be easy to implement]
- an automatic mechanism, which disables CUDA-apps when DX or OGL activity is detected; the user should be able to toggle this on/off
Regards, MrS |
|
|
|
For gaming I suggest a somewhat cumbersome workaround: suspend the BOINC network activity (so that it doesn't decide to fetch new GPU-WUs) and pause the active GPU-WU. BOINC may start the next GPU-WUs immediately, so I'd try to pause them as well. I can't test this method though, because I just ordered my card :)
Ideally it should not be a problem to run CUDA besides DX or OGL (that's how physics is supposed to work on Nv hardware, after all), but practically I can imagine situations where it just doesn't work, e.g. the CUDA app crashes.
BOINC should be prepared for such cases. I'd suggest to have:
- an option in the BOINC GUI to manually pause the CUDA-apps but keep the CPUs crunching [should be easy to implement]
- an automatic mechanism, which disables CUDA-apps when DX or OGL activity is detected; the user should be able to toggle this on/off
Regards, MrS
I'm not sure about the cuda - DirectX - OpenGL switching, but it is easy to suspend any project. Just go to the projects tab, select the project and click suspend. All computation for that project will be suspended and if you have other projects they will pick up. So no harm in running a CPU project in the background ;)
____________
|
|
|
|
Oh, I never noticed that button. That's a much better solution!
MrS
____________
Scanning for our furry friends since Jan 2002 |
|
|