Ok, I installed BOINC 6.10.4, setted in 1.o all on_frac values

Only 2 projects active: SETI and SETI beta.
Both have equal shares.
All debts zeroed.

both use app_info.
SETI beta use application that needs only part of CPU but uses accelerator (otherwise idle in current config).

So, I sen avg_cpu for this app to 0.4
Thjat is, 4 SETI tasks + some SETI beta tasks can go simultaneously.

But BOINC runs 4 SETI tasks in high priority mode leaving no room for SETI beta tasks (and also leaving accelerator idle).

Why in such config BOINC use high priority mode???????

---

That same problem I told you about yesterday. Your <on_frac> is very low, as BOINC thinks your system was off since February 2009. Exit BOINC, edit <on_frac> in client_state.xml to 0.990000, save changes, restart BOINC.

---

That same problem I told you about yesterday. Your <on_frac> is very low, as BOINC thinks your system was off since February 2009. Exit BOINC, edit <on_frac> in client_state.xml to 0.990000, save changes, restart BOINC.

Hm, not sure.
I recived your message and manually setted all frac values to 1 in the beginning of client_state.xml ...
Described behavior was after that...
BTW, my frac values were not so low as we were afraid. Lowest (BOINC on time) ~90% activity almost 1 work allowed almost 1 too ...

---

That same problem I told you about yesterday. Your <on_frac> is very low, as BOINC thinks your system was off since February 2009. Exit BOINC, edit <on_frac> in client_state.xml to 0.990000, save changes, restart BOINC.

I tried that (the 1.0), rebooted, didnt help. Every cuda type app I am running (einstein, seti, collatz, milkyway especially) gets a timeslice as soon as it downloads. At one time I had 24 milkway "waiting to run" with another 6 stuck uploading. Think I had 4 seti "waiting to run" I am starting to see computation errors after 2 or so seconds of runtime (a seti cuda job) There must be some type of problem with 30+ cuda apps (milkyway, seti, collatz, einstein) having run only a few seconds before being swapped out.

I noticed the 24 task limit on milkyway was exceeded when tasks are stuck uploading. More discussion here of the milkyway problem with 6.10.4

Currently I have 13 cuda type apps that have received a time slice and are waiting to run and I suspect I will have more when collatz is back.

This problem is not unique to 6.10.4 as I recall seeing it when I went from 6.6.36 to 6.10.1 on all 3 of my cuda systems, but it was not as bad. Only the "very next" cuda app was started leaving only 1 orphaned "waiting to run" on each of 3 systems. Starting with 6.10.4 all cuda apps get orphaned with "waiting to run"