It's normal for the current OpenCl work units to pause at 89.99% while the GPU communicates with the CPU. Usually it takes between 2-4 minutes for this back and forth to complete, then the work units goes to 100% complete.
When progress reaches 89.997%, the data has been completely crunched and all that is left to do is assemble all the potential candidate signals that might have been found and from them, compute a 'toplist', which is the 10 most likely signal candidates to be returned to the project for further analysis. During this toplist production stage, the progress counter remains static until it finally jumps to 100% when the toplist is ready.
If you abort the task at this time, you are throwing away all the work your machine has done.
Related to this topic I have 2 boxes with a GTX1060 in each. They both run 2 wu at a time. One runs Einstein 24/7 and the start times are staggered so the GPU ends up being used always. The other box splits it's time 50/50 with Seti. It starts it's tasks when Seti finishes so they end at the same time. That results in several minutes of down time for the GPU. That results in a minute or so longer run times. Any ides on how to get that box to stagger the start time and when it's time to shift over just pause the unfinished wu? Not a big deal but it would result in a couple of more being done a day.
how long did you wait before
)
how long did you wait before aborting??
It's normal for the current OpenCl work units to pause at 89.99% while the GPU communicates with the CPU. Usually it takes between 2-4 minutes for this back and forth to complete, then the work units goes to 100% complete.
When progress reaches
)
When progress reaches 89.997%, the data has been completely crunched and all that is left to do is assemble all the potential candidate signals that might have been found and from them, compute a 'toplist', which is the 10 most likely signal candidates to be returned to the project for further analysis. During this toplist production stage, the progress counter remains static until it finally jumps to 100% when the toplist is ready.
If you abort the task at this time, you are throwing away all the work your machine has done.
Cheers,
Gary.
Thanks. Odd behavior that
)
Thanks. Odd behavior that should've been explained in the FAQs.
Related to this topic I have
)
Related to this topic I have 2 boxes with a GTX1060 in each. They both run 2 wu at a time. One runs Einstein 24/7 and the start times are staggered so the GPU ends up being used always. The other box splits it's time 50/50 with Seti. It starts it's tasks when Seti finishes so they end at the same time. That results in several minutes of down time for the GPU. That results in a minute or so longer run times. Any ides on how to get that box to stagger the start time and when it's time to shift over just pause the unfinished wu? Not a big deal but it would result in a couple of more being done a day.