Look at these wu's low credits?

http://moowrap.net/result.php?resultid=375770

DD

Well I found the reason. For some reason my cpu yes was checked.

DD

exactly, the CPU is enable by default (this will maybe change btw).

If you look for your tasks with Distributed.net Client v1.02 (ati14) app, this takes much less cpu time and give much more credits.

Not so, I'm getting the same (low) credit for both CPU and GPU tasks. Issue still persists after detach/reattach from project..........

http://moowrap.net/workunit.php?wuid=410582

http://moowrap.net/workunit.php?wuid=323091

Not so, I'm getting the same (low) credit for both CPU and GPU tasks. Issue still persists after detach/reattach from project..........

http://moowrap.net/workunit.php?wuid=410582

http://moowrap.net/workunit.php?wuid=323091

Your credits are correct. The first is a (small) GPU WU of 12 Stat Units, x8.5 per stat unit = 102

The second is a CPU WU, again the credits are correct, same calculation.

Your focus should be on what the server is sending you. At present you are getting small WUs for a 58XX class card, you should be getting bigger size WUs. You dont lose credits as such because the app scales well, x8.5 credits per Stat Unit. I've seen this happen recently with someone else. Run the WUs out for an hour or so, see if the smaller ones run out and start being replaced by larger ones - however, you are not losing many credits against the norm - a few for more load/unloads, but not that many.

Your other issue is by running CPU WUs against all CPU cores, you will strangle the GPU application. Keep one CPU core free for each GPU thats running - for 5970 owners reading this, that includes 5970s where two CPU cores need keeping free to release the card to full use. With the latter though, watch the VRM temperatures, which will increase markedly if an overclock is in place, its likely the GPU will need downclocking a fair bit to bring all back in balance.

Lastly be wary of the GPU timing shown in the Account Task Page - its incorrect for many people. The correct timing is shown inside the stderr file at the bottom (the summary line).

Regards
Zy

Not so, I'm getting the same (low) credit for both CPU and GPU tasks. Issue still persists after detach/reattach from project..........

http://moowrap.net/workunit.php?wuid=410582

http://moowrap.net/workunit.php?wuid=323091

Hey Clod, looks like Blackheeler has been given CPU credit on a GPU work unit, both work units got 102 credits. Validator issue I believe.

Conan

---

No its not validator - look inside his stderr's - he is getting 12 Stat Unit sized WUs, so credits are correct

Regards
Zy

Not so, I'm getting the same (low) credit for both CPU and GPU tasks. Issue still persists after detach/reattach from project..........

Looks like another PFC scaling issue, where scheduler is thinking your host is slow and so is sending tiny work units to it. This should clear up as soon as you return enough work units to determine it's actual speed. As such, detach/reattach is something that shouldn't be done because that resets the history statistics.

I'll see if I can improve this PFC scaling (maybe I'll just disable it as it's just giving more problems than fixing things).

-w

..... I'll see if I can improve this PFC scaling (maybe I'll just disable it as it's just giving more problems than fixing things)...

Would be a shame to lose it, its a welcome flexible facility - maybe a Sticky Post "Titled 102 Credit WUs" would help for now until you nail the bugs.

Regards
Zy

Looks like another PFC scaling issue, where scheduler is thinking your host is slow and so is sending tiny work units to it. This should clear up as soon as you return enough work units to determine it's actual speed.

I'm guessing the 500+ tiny WU's I have received and returned in the 2 days since my first post still don't qualify as "enough" to deserve bigger WU's?

I'm guessing the 500+ tiny WU's I have received and returned in the 2 days since my first post still don't qualify as "enough" to deserve bigger WU's?

I just noticed that upstream (BOINC devs) has fixed a bug in this PFC scaling and I'm about to push that update live. Hopefully this means scheduler will finally believe you have fast host and starts sending huge work units for you too.

-w

Thank you Teemu, the larger WU's are coming through now :-)