This is occurring on a machine with a single CUDA card - a GeForce 680.

I am trying to also allow the machine to run CPU tasks as I have a very fast 8 core CPU. Under project prefs, I enabled the CPU app, but when I go into BOINC and do a "update project", it won't download any new projects and gives me the message

this computer has reached a limit on tasks in progress

it will continue to do GPU tasks - only downloading a new GPU task when the current task is finished.

How do I adjust the settings so that it downloads GPU tasks. I don't know what I am missing here settings-wise.

So having a similar problem, and suddenly BOINC Manager assigns only one job for one video card .. Tasks for other cards are downloaded in a row but wait .. they just will not start count on ..

6.4.2014 18:38:06 | GPUGRID | Finished upload of I868-NATHAN_KIDc22_full3-8-10-RND5010_0_9
6.4.2014 18:38:07 | GPUGRID | Sending scheduler request: To report completed tasks.
6.4.2014 18:38:07 | GPUGRID | Reporting 1 completed tasks
6.4.2014 18:38:07 | GPUGRID | Requesting new tasks for CPU and NVIDIA
6.4.2014 18:38:09 | GPUGRID | Scheduler request completed: got 1 new tasks
6.4.2014 18:38:11 | GPUGRID | Started download of e5s7_e3s6f68-GIANNI_trypben1MC9-0-LICENSE
6.4.2014 18:38:11 | GPUGRID | Started download of e5s7_e3s6f68-GIANNI_trypben1MC9-0-COPYRIGHT
6.4.2014 18:38:12 | GPUGRID | Finished download of e5s7_e3s6f68-GIANNI_trypben1MC9-0-LICENSE
6.4.2014 18:38:12 | GPUGRID | Finished download of e5s7_e3s6f68-GIANNI_trypben1MC9-0-COPYRIGHT
6.4.2014 18:38:12 | GPUGRID | Started download of e5s7_e3s6f68-GIANNI_trypben1MC9-0-coor_file

6.4.2014 20:32:36 | GPUGRID | Sending scheduler request: Requested by user.
6.4.2014 20:32:36 | GPUGRID | Requesting new tasks for CPU and NVIDIA
6.4.2014 20:32:37 | GPUGRID | Scheduler request completed: got 0 new tasks
6.4.2014 20:32:37 | GPUGRID | No tasks sent
6.4.2014 20:32:37 | GPUGRID | No tasks are available for Long runs (8-12 hours on fastest card)
6.4.2014 20:32:37 | GPUGRID | This computer has reached a limit on tasks in progress
6.4.2014 20:36:06 | GPUGRID | update requested by user
6.4.2014 20:36:07 | GPUGRID | Sending scheduler request: Requested by user.
6.4.2014 20:36:07 | GPUGRID | Requesting new tasks for CPU and NVIDIA
6.4.2014 20:36:09 | GPUGRID | Scheduler request completed: got 0 new tasks
6.4.2014 20:36:09 | GPUGRID | No tasks sent
6.4.2014 20:36:09 | GPUGRID | No tasks are available for Long runs (8-12 hours on fastest card)
6.4.2014 20:36:09 | GPUGRID | This computer has reached a limit on tasks in progress

To help it connect and disconnect ins BOINC CPU only project .. How simple.
But why is this happening ..? Perhaps that is linked from cpu only-GPUGRID new tasks..

In my experience, the GPUGrid project will give out up-to-2-tasks-per-GPU for a given system. And lately, there are some times when some of the queues are out of work (your own log pasting indicated that there were no long tasks available).

The best thing to do would be to make sure your project settings (on the GPUGrid website) are setup to retrieve tasks from any of the queues... and then, if you ever want to know how many tasks are available, look at the "unsent" column in the "Tasks by application" table at the bottom right of this page: http://www.gpugrid.net/server_status.php

One other thing to check might be to look at your account here on the website, and look at the tasks assigned to your computer. The number of "in progress" tasks should match what you see in BOINC.