Please visit https://einsteinathome.org/account/prefs/privacy and set "Should Einstein@Home show your computers on its website?: " to YES and save changes. That will help to see what's going on with your computer (last contact with project server etc.).
GTX 650 is a little bit older technology and may just be incompatible with the Gravitation Wave GPU application. There have been problems with other GTX 650's too, but I can't remember how they were able to run these GW GPU tasks in the end. Currently a 2GB card should be enough for many of the GW GPU tasks if the system is compatible with them at all.
There are two things you could try:
Stop Boinc from running any work. Set Einstein to "no new tasks". Exit Boinc.
1. Your GPU driver is currently very old and that can be a significant part of the problem. Visit Nvidia website and download the newest driver for your system. Then install the driver and reboot computer.
2. If you already haven't enabled 'Beta' applications you should try also that.
GTX 650 is a little bit older technology and may just be incompatible with the Gravitation Wave GPU application. There have been problems with other GTX 650's too, but I can't remember how they were able to run these GW GPU tasks in the end. Currently a 2GB card should be enough for many of the GW GPU tasks if the system is compatible with them at all.
There are two things you could try:
Stop Boinc from running any work. Set Einstein to "no new tasks". Exit Boinc.
1. Your GPU driver is currently very old and that can be a significant part of the problem. Visit Nvidia website and download the newest driver for your system. Then install the driver and reboot computer.
2. If you already haven't enabled 'Beta' applications you should try also that.
Start Boinc. See if your computer will then get GW GPU tasks v2.09 instead of the v2.07 .
The memory requirement for the current GW GPU tasks varies systematically with the DF value.
So far observed values of DF for work actually issued have ranged from 0.10 up to 0.50 with the higher values needing more GPU RAM space for success.
The one stderr file I examined for your errors indeed shows the expected symptom for not enough VRAM to run a particular Einstein GW GPU task:
CL_MEM_OBJECT_ALLOCATION_FAILURE
Unfortunately for your chance of success, the GW GPU work issued to your machine so far happens to be predominately (perhaps entirely, I'm too lazy to check them all) very high DF, so no chance at all that they will work with 2 GB GPU VRAM. I'd have hoped the project's attempts to protect lower capability GPU machines from tasks they clearly can't run would have stopped these from coming to you, but that seems not to be working properly (at the project server level, not you) in this case.
I doubt very much that a new driver or other such adjustments will make your GPU run these tasks. If VRAM capacity is the only problem, then very likely your machine will successfully run GW tasks with DF of 0.20 or less.
You may prefer to specialize in Gamma-Ray Pulsar GPU tasks, which have a very much lower VRAM capacity requirement.
I was prepared to see problems and that there would've been plenty of errors because of the 2GB on both of those cards. But there hasn't been... which was surprising and delightfull to see. After a few days now this host has currently 390 valids and 100 pending. There has been about 16 errors. All of the errored tasks have had DF of 0.40 or 0.45.
But... there have been also succesfull tasks wih DF of up to 0.40 (at least). For example:
From what I've seen I agree that yes... tasks with higher DF tend to require more memory, but to my surprise... not always! I doesn't seem to be completely systematic. Tasks with DF of 0.40 used to fail in thes cards but it looks like sometimes they may now run even succesfully.
Also, there seems to be very much tasks available with lower DF values. No lack of them. For this host the project server has succeeded now very well in sending tasks that these 2GB cards are actually able to run. I think it's also important to mention that this host has been running only v2.09 tasks. I believe v2.07 would not have this result with these cards.
So, this is why I had hopes also for the GTX 650 2GB and mentioned about the driver also. It would be interesting to see the current situation for that card.
Quote:
You may prefer to specialize in Gamma-Ray Pulsar GPU tasks, which have a very much lower VRAM capacity requirement.
Please visit
)
Please visit https://einsteinathome.org/account/prefs/privacy and set "Should Einstein@Home show your computers on its website?: " to YES and save changes. That will help to see what's going on with your computer (last contact with project server etc.).
GPU > 2GB !!!!
)
GPU > 2GB !!!! impèrativ / zwingend / necessary / настоятельный
Hello, thank Well done !
)
Hello, thank you, Well done !
GTX 650 is a little bit older
)
GTX 650 is a little bit older technology and may just be incompatible with the Gravitation Wave GPU application. There have been problems with other GTX 650's too, but I can't remember how they were able to run these GW GPU tasks in the end. Currently a 2GB card should be enough for many of the GW GPU tasks if the system is compatible with them at all.
There are two things you could try:
Stop Boinc from running any work. Set Einstein to "no new tasks". Exit Boinc.
1. Your GPU driver is currently very old and that can be a significant part of the problem. Visit Nvidia website and download the newest driver for your system. Then install the driver and reboot computer.
2. If you already haven't enabled 'Beta' applications you should try also that.
Please visit https://einsteinathome.org/account/prefs/project . Under the "Beta settings" set "Run test applications?: " to YES and save changes.
Start Boinc. See if your computer will then get GW GPU tasks v2.09 instead of the v2.07 .
Richie wrote: GTX 650 is a
)
The memory requirement for the current GW GPU tasks varies systematically with the DF value.
So far observed values of DF for work actually issued have ranged from 0.10 up to 0.50 with the higher values needing more GPU RAM space for success.
The one stderr file I examined for your errors indeed shows the expected symptom for not enough VRAM to run a particular Einstein GW GPU task:
CL_MEM_OBJECT_ALLOCATION_FAILURE
Unfortunately for your chance of success, the GW GPU work issued to your machine so far happens to be predominately (perhaps entirely, I'm too lazy to check them all) very high DF, so no chance at all that they will work with 2 GB GPU VRAM. I'd have hoped the project's attempts to protect lower capability GPU machines from tasks they clearly can't run would have stopped these from coming to you, but that seems not to be working properly (at the project server level, not you) in this case.
I doubt very much that a new driver or other such adjustments will make your GPU run these tasks. If VRAM capacity is the only problem, then very likely your machine will successfully run GW tasks with DF of 0.20 or less.
You may prefer to specialize in Gamma-Ray Pulsar GPU tasks, which have a very much lower VRAM capacity requirement.
Good luck.
archae86 wrote:The memory
)
My recent observations are different on that DF. A few days ago I fired up again this host that has 2 x GTX 960 2GB: https://einsteinathome.org/host/12768123
I was prepared to see problems and that there would've been plenty of errors because of the 2GB on both of those cards. But there hasn't been... which was surprising and delightfull to see. After a few days now this host has currently 390 valids and 100 pending. There has been about 16 errors. All of the errored tasks have had DF of 0.40 or 0.45.
But... there have been also succesfull tasks wih DF of up to 0.40 (at least). For example:
https://einsteinathome.org/task/1014579939
https://einsteinathome.org/task/1014579938
https://einsteinathome.org/task/1014572443
From what I've seen I agree that yes... tasks with higher DF tend to require more memory, but to my surprise... not always! I doesn't seem to be completely systematic. Tasks with DF of 0.40 used to fail in thes cards but it looks like sometimes they may now run even succesfully.
Also, there seems to be very much tasks available with lower DF values. No lack of them. For this host the project server has succeeded now very well in sending tasks that these 2GB cards are actually able to run. I think it's also important to mention that this host has been running only v2.09 tasks. I believe v2.07 would not have this result with these cards.
So, this is why I had hopes also for the GTX 650 2GB and mentioned about the driver also. It would be interesting to see the current situation for that card.
Yep, that is a good option, no doubt about that.