BOINC/Primegrid is grossly overestimating my CPU's ability to complete tasks. I've been watching my account over the last couple of months I've found that I either have to abort the tasks or BOINC will eventually do it.

The problem is the BOINC/Primegrid insists on sending more work with very short TAT and I'm aborting more tasks.

I've got my system to have 0.25 days of extra work which shouldn't allow that much.

I aborted several tasks as they wouldn't be started by the deadline. BOINC immediately sent me 34 more tasks that expired on december 1. These tasks wont be started until the 29th of this month. I've done the best case scenario. thats 275 hours of work over 7 cores. thats about 39 1/4 hours of work in 48 hours. That only works if I don't play any video games (which I do) during that time. BOINC should see that it only has so many hours a day to run tasks and not attempt to fill every single moment with a WU.

A simple easy solution is to move back the TAT date by a few days. Thats for every project on Primegrid. I'm getting a bit tired of having so many abandoned WU's when its just so unnecessary. You guys are clearly wasting so much bandwidth as I've seen so many WU's just sit. I can't believe that everyone is just quitting the project and not completing the work. This appears to me to be a problem with attempts to overfeed CPU's and wasting both of our time.

while its nice to know that the system will eventually report back the work. However, This prevents me from running other projects that have long TAT. Primegrid/BOINC will ignore those tasks and fill(overfill) my Que to the point that once again I am forced to abort tasks.

I was mistaken. I actually received 49 tasks. The second grouping of tasks is slated for 2 days after the first grouping. The huge problem once again is the limited time to return tasks and the quantity of tasks.

your point 3. is not relevant. BOINC knows after so many runs how long a task takes to run. Primegrid WU's always seem to have the incorrect time assigned to them when I get them. Some either complete well before time others complete several hours after. My CPU isn't in control of how BOINC and Primegrid check times.

I don't OC and Don't have a heat problem. 45 C on water cooling is generally considered good. You assume that the problem is with my system and not the other computer involved.

My PrimeGrid cache ceased to be a problem once I set BOINC manager to store '0' days of work and '0' days of additional work. The only thing that works consistently (EDIT: for me) is BOINC automatically downloading new work only as needed. This setting worked when I used my old AMD FX-8350 as well as my current Intel cpu.

I would only maintain a cache of work if I had crappy or intermittent internet access.

Since you are running multiple applications at PG, and BOINC gives a score (duration correction factor) only to a single project, not each application, the problem is BOINC not being able to provide an accurate number for your machine. You also run both CPU and GPU projects, which skews that number even more. BOINC is notorious for having issues with projects that have multiple applications, especially if they are of different styles, as PG has, with Sieve, LLR, GFN and AP.

Many people have asked the BOINC people to fix this. but they have not found a good way to provide that sort of ability. So, since each project gets one DFC, it's hard for it to estimate how much work you need.

____________
My lucky numbers are 121*2^4553899-1 and 3756801695685*2^666669±1
My movie https://vimeo.com/manage/videos/502242