Well I haven't had the time to update all of my 6 GeForce cards yet but the one I did seems to run the same but that is just my GeForce 610M in this laptop so I don't expect much but when I get some time I will see how it does on my 660Ti and a couple 650Ti's and a couple 550Ti's (all OC'd)
Now the MAIN thing you want to do is suspend your tasks when these drivers are being installed since that is what will cause the error on those tasks.
I actually sit and watch it all happen and since I hate to ever suspend these tasks I let them run while the driver *downloads* and then make sure I *suspend* the tasks as the driver is *installed* and that way I can *resume* the project tasks without getting that error.
(and I never play games since I am from the pong days )
Oh and I usually find that you need to reboot before you see the driver update on your host list here.
Now the MAIN thing you want to do is suspend your tasks when these drivers are being installed since that is what will cause the error on those tasks.
What I do is exit BOINC completely (and prevent it from starting up after the reboot) and then update the drivers; it is a clean install that takes several reboots right there. Finally, after the drivers have been installed and rebooted to ensure they are enabled, I re-enable the BOINC startup and reboot the PC.
It may have been just a bad batch of work units; they are Betas, and it happened once before (and once since), but the timing seems very suspicious to me.
(I haven't tried other applications yet, because I like FGRP3-opencl being suboptimal enough to not overheat this stupid laptop.)
Khangollo, those cards in the laptops do run a bit warmer but the card you have is a bit newer than my 610M and so is your CPU and I have been running the GPU tasks here 24/7 for 2 years now.
I do have a little 4in fan blowing on the vent all the time and leave it in the coldest room at night and check the temp with the GPU-Z when I am using it (like now) and I usually keep it running between 79C and 86C which is safe temp for our laptops.
This is running the BRP PAS tasks and since it is 8-core I also run the T4T -VB tasks and once in a while use 6 cores to run LHC or Pogs all at the same time.
Now the MAIN thing you want to do is suspend your tasks when these drivers are being installed since that is what will cause the error on those tasks.
What I do is exit BOINC completely (and prevent it from starting up after the reboot) and then update the drivers; it is a clean install that takes several reboots right there. Finally, after the drivers have been installed and rebooted to ensure they are enabled, I re-enable the BOINC startup and reboot the PC.
It may have been just a bad batch of work units; they are Betas, and it happened once before (and once since), but the timing seems very suspicious to me.
Yeah I never exit Boinc myself when I update the drivers since I have other tasks running (and I rather not stop a T4T with VB task)
I had that error happen when first started running GPU's a few years ago so after that I just suspended the Einstein tasks when the driver is being installed and then the reboot and then resume the tasks and haven't had any erros since then. (of course when I reboot I do have to suspend that T4T task and the VB for a couple minutes)
Couple of days ago I've updated BIOS on my motherboard and after that found that BRP5 tasks are taking much more time to finish - up to 30% more.
After some musing I've found the reason - the RAM memory was overclocked a bit to 1600 MHz and was reset to default value in a process of BIOS updating.
I've overclocked the memory one more time and now it's back to previous time for one WUs to process.
However, I was always sure that the difference between OC and non OC memory will be couple of percents but not couple of dozens.
Can anybody confirm it or it is just my imagination?
I got a simular slow down when I was running BRP4 tasks and BRP5 tasks at the same time while running 3 at a time on my GTX660. Now I only run BRP5 tasks on the GPU.
Sid you are 100% correct for BRP tasks. The better the PCI bandwidth and
corresponding main memory bandwith the faster a powerful GPU will perform.
Einstein is almost alone in this respect. Milkyway and SETI are GPU limited
Any other projects in BOINC limited by PCI/memory bandwidth like BRP in Einstein?
Alot of people use Einstein as a backup project for SETI some with very powerful
GPU systems like three GTX690's which perform very well in SETI but bog the PCI memory bus so much here in Einstein they thought their systems were broken.
.....on another note........I am tempted to get a couple of these for backups and new hosts GTX GeForce 550Ti
I have 2 from another company and they work good especially at this price.
You might want to check around and see what information you can find on a GTX 750, also.
There's a stock Gigabyte model available for like $110.
I would compare my overclocked 550Ti to my overclocked 750Ti for you, but both have been gone from Einstein for long enough that there isn't much evidence left in the statistics.
I think you are going to discover that the 750 is faster than a 550Ti, but the reason I mention it is because a "stock" 750 burns so much less electricity and produces so much less heat that I suspect the $20 difference would be worth spending.
I don't *know* times for these cards anymore since mine have aged-out of my view, so this doesn't rise to the level of a purchase-recommendation; I'm just thinking it might be worth your investigation if you make one.
I can see in the statistics that a 750 (no-Ti) has turned in times of 18,000 on BRP5, but I have no idea if that's two-at-a-time or not. Even if the times were the same, I really think you might like a 750, non-Ti, to replace a 550Ti.
EDIT - I'll be back to doing some serious Einstein crunching September 1, if that isn't too late. Then I'll have a better basis for comparisons. In the meantime Phil is running some 750s.
Well I haven't had the time
)
Well I haven't had the time to update all of my 6 GeForce cards yet but the one I did seems to run the same but that is just my GeForce 610M in this laptop so I don't expect much but when I get some time I will see how it does on my 660Ti and a couple 650Ti's and a couple 550Ti's (all OC'd)
Now the MAIN thing you want to do is suspend your tasks when these drivers are being installed since that is what will cause the error on those tasks.
I actually sit and watch it all happen and since I hate to ever suspend these tasks I let them run while the driver *downloads* and then make sure I *suspend* the tasks as the driver is *installed* and that way I can *resume* the project tasks without getting that error.
(and I never play games since I am from the pong days )
Oh and I usually find that you need to reboot before you see the driver update on your host list here.
RE: Now the MAIN thing you
)
What I do is exit BOINC completely (and prevent it from starting up after the reboot) and then update the drivers; it is a clean install that takes several reboots right there. Finally, after the drivers have been installed and rebooted to ensure they are enabled, I re-enable the BOINC startup and reboot the PC.
It may have been just a bad batch of work units; they are Betas, and it happened once before (and once since), but the timing seems very suspicious to me.
RE: (I haven't tried other
)
Khangollo, those cards in the laptops do run a bit warmer but the card you have is a bit newer than my 610M and so is your CPU and I have been running the GPU tasks here 24/7 for 2 years now.
I do have a little 4in fan blowing on the vent all the time and leave it in the coldest room at night and check the temp with the GPU-Z when I am using it (like now) and I usually keep it running between 79C and 86C which is safe temp for our laptops.
This is running the BRP PAS tasks and since it is 8-core I also run the T4T -VB tasks and once in a while use 6 cores to run LHC or Pogs all at the same time.
So after 2 years I haven't killed this laptop
RE: RE: Now the MAIN
)
Yeah I never exit Boinc myself when I update the drivers since I have other tasks running (and I rather not stop a T4T with VB task)
I had that error happen when first started running GPU's a few years ago so after that I just suspended the Einstein tasks when the driver is being installed and then the reboot and then resume the tasks and haven't had any erros since then. (of course when I reboot I do have to suspend that T4T task and the VB for a couple minutes)
(they don't give you long to
)
(they don't give you long to edit here)
This driver on one of my 550Ti seems to be running the same as the previous driver
But I wonder more about the 660Ti and the 2 I have running with Phenoms since they have ran faster than the Athlons using the same cards.
Couple of days ago I've
)
Couple of days ago I've updated BIOS on my motherboard and after that found that BRP5 tasks are taking much more time to finish - up to 30% more.
After some musing I've found the reason - the RAM memory was overclocked a bit to 1600 MHz and was reset to default value in a process of BIOS updating.
I've overclocked the memory one more time and now it's back to previous time for one WUs to process.
However, I was always sure that the difference between OC and non OC memory will be couple of percents but not couple of dozens.
Can anybody confirm it or it is just my imagination?
I got a simular slow down
)
I got a simular slow down when I was running BRP4 tasks and BRP5 tasks at the same time while running 3 at a time on my GTX660. Now I only run BRP5 tasks on the GPU.
Sid you are 100% correct for
)
Sid you are 100% correct for BRP tasks. The better the PCI bandwidth and
corresponding main memory bandwith the faster a powerful GPU will perform.
Einstein is almost alone in this respect. Milkyway and SETI are GPU limited
Any other projects in BOINC limited by PCI/memory bandwidth like BRP in Einstein?
Alot of people use Einstein as a backup project for SETI some with very powerful
GPU systems like three GTX690's which perform very well in SETI but bog the PCI memory bus so much here in Einstein they thought their systems were broken.
.....on another note........I
)
.....on another note........I am tempted to get a couple of these for backups and new hosts GTX GeForce 550Ti
I have 2 from another company and they work good especially at this price.
RE: .....on another
)
You might want to check around and see what information you can find on a GTX 750, also.
There's a stock Gigabyte model available for like $110.
I would compare my overclocked 550Ti to my overclocked 750Ti for you, but both have been gone from Einstein for long enough that there isn't much evidence left in the statistics.
I think you are going to discover that the 750 is faster than a 550Ti, but the reason I mention it is because a "stock" 750 burns so much less electricity and produces so much less heat that I suspect the $20 difference would be worth spending.
I don't *know* times for these cards anymore since mine have aged-out of my view, so this doesn't rise to the level of a purchase-recommendation; I'm just thinking it might be worth your investigation if you make one.
I can see in the statistics that a 750 (no-Ti) has turned in times of 18,000 on BRP5, but I have no idea if that's two-at-a-time or not. Even if the times were the same, I really think you might like a 750, non-Ti, to replace a 550Ti.
EDIT - I'll be back to doing some serious Einstein crunching September 1, if that isn't too late. Then I'll have a better basis for comparisons. In the meantime Phil is running some 750s.