They were the first experimental model, tested in the laboratory (planet Earth), but not judged suitable for release in the Beta test project for the rest of the galaxy ;-)
I thought the HitchHiker's Guide would have made that clear.....
# We claim a full CPU core for the application because BOINC won't reset the task to idle priority when less than one core is claimed
there is a solution?
It's not just because of BOINC. It's because of BOINC and the fact that the current test release still runs 30%-50% on the CPU. Please have a look again at my previous post again.
System ran overnight on E@H only and returned this WU. It appears to be successful, but only the validator knows for sure... It validated against another CUDA task.
[edit]I'm going to unsuspend SETI and see if the two will play nice together.[/edit]
The 2 S5R5 WUs were still running and BOINC started the CUDA WU. The task-manager showed that 3 app were running: two S5R5 both at 25% and the CUDA at 50 cpu usage.
These tasks failed due to a lack of device memory. This could be related to the game you played right before you restarted the CUDA tasks. Maybe it didn't clean-up (free) all graphics memory...
I'm certain the developers are working on it. I don't know how it all works on other projects, like Seti.
On GPU Grid the current application takes about 1-3% of the CPU to run a task on the GPU. The main load is the transfer of data from the system to the GPU and then unloading of the results (If I understood the developers there).
Last December the load was as high as 25% per GPU running because they were trying to see how efficient they could get. Or to put it another way, if they wasted CPU time would the tasks run faster. In the end the CPU time was cut to the current level with maybe a 5-10% loss in run time efficiency of the GPU Grid tasks, but this does allow those of us that run other projects to be able to run other projects (like EaH) more effectively...
For myself I stopped running Aqua (for example) because they were having so many troubles and the run times of the CPU tasks were over 100 hours and the project though they did look at trickles were resistant to using them (for whatever reason).
But that is the nature of a new technology and technique.
For myself, if the EaH task uses 100% of the GPU and one core I am not at all sure I would make the investment in that I can run GPU grid 100% on the GPU and still run EaH 100% on the CPU (effectively)... not sure if the current load is worth the investment ... :)
Of course, for me this is moot as I am working on WCG badges anyway ... so I have some time to see what is the results of this beta testing ...
Sorry to be a little thick on graphics card drivers, but ...
Before updating the drivers for an nVidia card to a newer version, should the old driver be removed from the system.
It wasn't required when I updated from 185.85 to current 190.38. If you (or anybody) do want to uninstall, finish and report all existing CUDA tasks first or make sure BOINC won't start during the upgrade or at reboot. All CUDA tasks will be lost if BOINC starts with no CUDA capable driver installed.
Quote:
If so, is the method to use - /Control Panel/Add or Remove Programmes/NVIDIA Drivers/Change or Remove/
That should work.
Quote:
I presume the new drivers cannot simply be installed over the existing graphic card drivers, as used in BOINC (after a programme shutdown)?
BOINC has to be shut down during the upgrade if that's what you're asking.
Quote:
As the graphic card is central to the system, I also assume any driver update should avoid a PC reboot between changing drivers?
AFAIK the system uses a generic driver when nothing fancier is in place. This is easy to notice from the reduction in resolution (unless you use minimal resolution all the time). So rebooting is allowed (but why would you want to do it?)
Sorry to be a little thick on graphics card drivers, but ...
Before updating the drivers for an nVidia card to a newer version, should the old driver be removed from the system.
It wasn't required when I updated from 185.85 to current 190.38. If you (or anybody) do want to uninstall, finish and report all existing CUDA tasks first or make sure BOINC won't start during the upgrade or at reboot. All CUDA tasks will be lost if BOINC starts with no CUDA capable driver installed.
Quote:
If so, is the method to use - /Control Panel/Add or Remove Programmes/NVIDIA Drivers/Change or Remove/
That should work.
Quote:
I presume the new drivers cannot simply be installed over the existing graphic card drivers, as used in BOINC (after a programme shutdown)?
BOINC has to be shut down during the upgrade if that's what you're asking.
Quote:
As the graphic card is central to the system, I also assume any driver update should avoid a PC reboot between changing drivers?
AFAIK the system uses a generic driver when nothing fancier is in place. This is easy to notice from the reduction in resolution (unless you use minimal resolution all the time). So rebooting is allowed (but why would you want to do it?)
Thanks Samuel7, very clear.
1. Shutdown BOINC, when all Einstein WUs are complete (use NNT).
2. Go to the nVidia web site and download the 190.38 driver to a safe folder (maybe do this first).
3. Instal the 190.38 driver (or the 181.20 referred to earlier or Samuel7's 185.85 drivers) over the top of my version (169.21 for XP).
4. Reboot PC and recommence crunching.
I have 4 PCs on a Belkin KVM switch and all run at my 24 inch screen's native resolution of 1,680 x 1,050
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
RE: Hmmmmm ... where does
)
They were the first experimental model, tested in the laboratory (planet Earth), but not judged suitable for release in the Beta test project for the rest of the galaxy ;-)
I thought the HitchHiker's Guide would have made that clear.....
RE: RE: RE: I have a
)
It's not just because of BOINC. It's because of BOINC and the fact that the current test release still runs 30%-50% on the CPU. Please have a look again at my previous post again.
Thanks,
Oliver
Einstein@Home Project
RE: Hi RandyC, RE: I
)
System ran overnight on E@H only and returned this WU. It appears to be successful, but only the validator knows for sure... It validated against another CUDA task.
[edit]I'm going to unsuspend SETI and see if the two will play nice together.[/edit]
Seti Classic Final Total: 11446 WU.
Hi Bernd, My system (BOINC
)
Hi Bernd,
My system (BOINC 6.6.36, Geforce 8800 GTS (320)MB, NvidiaDriver 182.50, Win XP 32, Core2Duo E6750)
the first 2 WUs are pending and approximately 20% faster than the old just-CPU ones
http://einsteinathome.org/workunit/56901612
http://einsteinathome.org/workunit/56901607
For those two WUs the Box "Use GPU while computer is in use" was checked.
I unchecked it to play a game and the CPU core which crunched the Cuda app switched to a new S5R5.
After I finished the game I checked the Box "Use GPU while computer is in use" again, but then the CUDA WUs errored out immediately
http://einsteinathome.org/task/136060092
http://einsteinathome.org/task/136060116
http://einsteinathome.org/task/136060117
and some more ... http://einsteinathome.org/account/tasks&offset=0
The 2 S5R5 WUs were still running and BOINC started the CUDA WU. The task-manager showed that 3 app were running: two S5R5 both at 25% and the CUDA at 50 cpu usage.
Sascha
Sorry to be a little thick on
)
Sorry to be a little thick on graphics card drivers, but ...
Before updating the drivers for an nVidia card to a newer version, should the old driver be removed from the system.
If so, is the method to use - /Control Panel/Add or Remove Programmes/NVIDIA Drivers/Change or Remove/
I presume the new drivers cannot simply be installed over the existing graphic card drivers, as used in BOINC (after a programme shutdown)?
As the graphic card is central to the system, I also assume any driver update should avoid a PC reboot between changing drivers?
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!
Hi
)
Hi tolafoph,
These tasks failed due to a lack of device memory. This could be related to the game you played right before you restarted the CUDA tasks. Maybe it didn't clean-up (free) all graphics memory...
Oliver
Einstein@Home Project
RE: I'm certain the
)
On GPU Grid the current application takes about 1-3% of the CPU to run a task on the GPU. The main load is the transfer of data from the system to the GPU and then unloading of the results (If I understood the developers there).
Last December the load was as high as 25% per GPU running because they were trying to see how efficient they could get. Or to put it another way, if they wasted CPU time would the tasks run faster. In the end the CPU time was cut to the current level with maybe a 5-10% loss in run time efficiency of the GPU Grid tasks, but this does allow those of us that run other projects to be able to run other projects (like EaH) more effectively...
For myself I stopped running Aqua (for example) because they were having so many troubles and the run times of the CPU tasks were over 100 hours and the project though they did look at trickles were resistant to using them (for whatever reason).
But that is the nature of a new technology and technique.
For myself, if the EaH task uses 100% of the GPU and one core I am not at all sure I would make the investment in that I can run GPU grid 100% on the GPU and still run EaH 100% on the CPU (effectively)... not sure if the current load is worth the investment ... :)
Of course, for me this is moot as I am working on WCG badges anyway ... so I have some time to see what is the results of this beta testing ...
RE: Sorry to be a little
)
It wasn't required when I updated from 185.85 to current 190.38. If you (or anybody) do want to uninstall, finish and report all existing CUDA tasks first or make sure BOINC won't start during the upgrade or at reboot. All CUDA tasks will be lost if BOINC starts with no CUDA capable driver installed.
That should work.
BOINC has to be shut down during the upgrade if that's what you're asking.
AFAIK the system uses a generic driver when nothing fancier is in place. This is easy to notice from the reduction in resolution (unless you use minimal resolution all the time). So rebooting is allowed (but why would you want to do it?)
Deleted by myself. Answered
)
Deleted by myself. Answered by an other user.
RE: RE: Sorry to be a
)
Thanks Samuel7, very clear.
1. Shutdown BOINC, when all Einstein WUs are complete (use NNT).
2. Go to the nVidia web site and download the 190.38 driver to a safe folder (maybe do this first).
3. Instal the 190.38 driver (or the 181.20 referred to earlier or Samuel7's 185.85 drivers) over the top of my version (169.21 for XP).
4. Reboot PC and recommence crunching.
I have 4 PCs on a Belkin KVM switch and all run at my 24 inch screen's native resolution of 1,680 x 1,050
Shih-Tzu are clever, cuddly, playful and rule!! Jack Russell are feisty!