How to raise CPU Utilization

There are two different GPU searches available.  Make sure you have just one selected in your preferences whilst you are getting used to how Einstein works.  The gamma-ray pulsar GPU search will have estimates that are too high.  The GW GPU search will have estimates that are too low.  Since Einstein uses duration correction factor (DCF) to adjust estimates and since there is only one of these, the two searches will tend to conflict with each other and cause estimates that can fluctuate wildly.  Until you get used to how this plays out, you will find it easier to stick to just one GPU search.

Once you have a GPU search selected, the main reason you wouldn't be getting GPU tasks would be that BOINC hasn't detected the OpenCL capabilities of your GPU.  Check the first 30-50 lines in the event log when BOINC starts up to see that the OpenCL capability of your GPU is clearly detected and listed.  If not, you don't have a proper driver installation.

Another thing to check is the scheduler contact log.  If you go to the website and look at your list of hosts attached to the project, the far right hand column is a link to the details of your host's most recent contact with the scheduler.  Click that link and you can see the full details.  Don't be put off if there happens to be a lot of strange stuff.  Just look in the early part for what is recorded about the work your host asked for.  Here is an example of what I'm talking about, taken from your machine just now.  There are not many lines since the machine was just returning a task and not asking for anything new.

  ... <excess snipped> ...
2020-03-04 03:03:27.4746 [PID=18468] [send] effective_ncpus 7 max_jobs_on_host_cpu 999999 max_jobs_on_host 999999
2020-03-04 03:03:27.4746 [PID=18468] [send] effective_ngpus 0 max_jobs_on_host_gpu 999999
2020-03-04 03:03:27.4747 [PID=18468] [send] Not using matchmaker scheduling; Not using EDF sim
2020-03-04 03:03:27.4747 [PID=18468] [send] CPU: req 0.00 sec, 0.00 instances; est delay 0.00
2020-03-04 03:03:27.4747 [PID=18468] [send] work_req_seconds: 0.00 secs
  ... <excess snipped> ...

The scheduler is being told that effective_ncpus is 7 and effective_ngpus is zero.  It would appear that your GPU is not being detected -OR- your preferences have excluded it.  You need to check that.

You can also see the CPU work request is 0.00 sec.  There is no entry for a GPU work request.  Your BOINC client isn't asking for GPU work.  It's either that preferences don't allow it or you are missing the OpenCL libs in your driver installation.  I'm guessing the latter since I think there should be an entry for 0.00 secs for the GPU if it had been detected but didn't need any work - just like for the CPU cores.

I am curious about your problem with the 5gb limit.  The reason I ask is that I have three P102-100s which are "5gb limited" versions of the gtx-1080ti.  I have been wondering if I can unlock the extra memory on these "mining only" cards.  They can co-exist with other gtx boards and are cheaper than the RTX though they require 8+6 power.