Sometimes I see bunches of work units being processed by my GPU then nothing for a while even when I still have many work units waiting in the work queue. SETI has been using GPU even when I watch videos.

Workunits are assigned to a resource once requested by BOINC. So that means work sitting in your queue will be assigned to either the CPU or GPU and will wait for that resource to become available.

So therefore if the work you have waiting in the queue is for your CPU, they will not be re-assigned to your GPU and processed. They will continue to wait their turn to be processed by your CPU.

Not assigning work units to a specific resource would result in a greater thruput.
My idle GPU is many times faster than any of my CPU cores. Is there a chance of
changes to the resource assignment method to allow the use of any free resource?

Which machine are you talking about?

The one with the gtx 750?

Have you considered using the lunatics optimized applications?

Just a thought

Yes, that is the machine. Will look into the lunatics thing. Thank you for the suggestion.

Not assigning work units to a specific resource would result in a greater thruput.
My idle GPU is many times faster than any of my CPU cores. Is there a chance of
changes to the resource assignment method to allow the use of any free resource?

There is no way to switch to a general "resource pool" without a significant re-write of the BOINC code.

The idea to change from assigned resource to a general pool has been discussed. I don't recall what the outcome of that discussion was, however.

Yes, that is the machine. Will look into the lunatics thing. Thank you for the suggestion.

Switching to the Lunatics third-party application will only make better use of your existing resources (optimized). They will not resolve the assigned resource vs. resource pool issue.

Found alternate solution: I have increased my thruput (cobblestones earned) by allocating percent of cpu cores cores to one full cpu core plus about 3 to 4 more percent to feed the GPU. Option to use the GPU all the time is left on. This way, my downloaded work units are mostly assigned to the GPU and each gets processed very quickly relative to the slower cpu core. This WITHOUT switching to Lunatics 3rd party applications.

This WITHOUT switching to Lunatics 3rd party applications.

Why 'WITHOUT' is all-CAPS?
In fact the stock GPU apps used since v7 apps are from Lunatics
(but the CPU apps are not so are slower than Lunatics)

Look the 2 Screenshots here:
http://setiathome.berkeley.edu/forum_thread.php?id=75115&postid=1541308#1541308

---

 


- ALF - "Find out what you don't do well ..... then don't do it!" :)
 

Because my original concern was that my gpu was not being used for long periods of time while all 4 cpu cores were kept very busy.

Looking at the processing times of work unit om my cpu vs the gpu, for every one work unit done by the cpu (5 to 6 hours each) the gpu processed processed about 10 on this machine but only if the gpu was actually processing.

My desire was to keep my gpu very busy with SETI for increased work unit thruput.
I got the desired results by changing only one existing parameter and freeing 3 cpu cores for other work with low or no gpu requirements.

Glancing at Lunatics info gave me an impression that the switch was an more complex one than reassigning the percentage of cores used to one plus a few percent.