It's sort of weird: The progress counter of WUs for 7.03 openCL is reset for every WU - after 20-22 seconds the counter goes back to 0% and after that everything goes on without a problem.
This phenomenon does not affect the result but is confusing nevertheless. Anyone knows why this happens?

---

Om mani padme hum.

It takes about that time to load the GPU.
What are your CPU cores doing ?

---

With each crime and every kindness we birth our future.

It takes about that time to load the GPU.
What are your CPU cores doing ?

Interesting and it makes sense. Just did not think about the GPU-loading before.
CPU cores are busy with crunching on CPU based WUs: 4/6 cores doing normals WUs, 1/6 is for GPU-based WU support and 1/6 is idle or used by other tasks.
Everything within normal parameters.

Edit: Observed the CPU load: When a new WU is started, CPU load showes a peak probably indicating the GPU loading phase. Learning is fun :)

---

Om mani padme hum.

I`m certain your GPU doesn`t get enough resources on start up.

I only utilize 50% of the cores on my 8350.

You can either live with that or free another CPU core.
I would bet your CPU times improve from this as well.
Mine did.

---

With each crime and every kindness we birth our future.

It's sort of weird: The progress counter of WUs for 7.03 openCL is reset for every WU - after 20-22 seconds the counter goes back to 0% and after that everything goes on without a problem.
This phenomenon does not affect the result but is confusing nevertheless. Anyone knows why this happens?

Simple, introduced in Boinc 7.2.38 was this changeset:

http://boinc.berkeley.edu/gitweb/?p=boinc-v2.git;a=commitdiff;h=9136a369d4e15cc727c06b55b50c833e184bf9fc

client: if app doesn't report fraction done, estimate it

This means before the app has reported progress (fftw might be doing it's planning), Boinc will estimate it, when the app does report progress, it'll then count up (from start, or last checkpoint).

It also means if you get a driver restart, and the app basically stalls, Boinc will continue estimating progress even through there is none, it'll approach 100% but never get quite get there:

http://boinc.berkeley.edu/gitweb/?p=boinc-v2.git;a=commitdiff;h=34f252870310b18c7cbe3e71573daff6b01e768c;hp=9136a369d4e15cc727c06b55b50c833e184bf9fc

client: if app doesn't report fraction done, estimate fraction done in a way that converges to but never reaches 100%.

Claggy

It's sort of weird: The progress counter of WUs for 7.03 openCL is reset for every WU - after 20-22 seconds the counter goes back to 0% and after that everything goes on without a problem.
This phenomenon does not affect the result but is confusing nevertheless. Anyone knows why this happens?

Simple, introduced in Boinc 7.2.38 was this changeset:

http://boinc.berkeley.edu/gitweb/?p=boinc-v2.git;a=commitdiff;h=9136a369d4e15cc727c06b55b50c833e184bf9fc

client: if app doesn't report fraction done, estimate it

This means before the app has reported progress (fftw might be doing it's planning), Boinc will estimate it, when the app does report progress, it'll then count up (from start, or last checkpoint).

It also means if you get a driver restart, and the app basically stalls, Boinc will continue estimating progress even through there is none, it'll approach 100% but never get quite get there:

http://boinc.berkeley.edu/gitweb/?p=boinc-v2.git;a=commitdiff;h=34f252870310b18c7cbe3e71573daff6b01e768c;hp=9136a369d4e15cc727c06b55b50c833e184bf9fc

client: if app doesn't report fraction done, estimate fraction done in a way that converges to but never reaches 100%.

Claggy

Indeed it appears to be that way. I tried several CPU configurations with no measurable result, so it has to be the altered Bionic 7.2.38 settings. Broadened my knowledge once more, thanks.

---

Om mani padme hum.