Quote Originally Posted by pirogue View Post
Gotcha. Thanks.

I'm trying some new things today.
2600K + 7970 #1:
<count>.041</count>
<avg_ncpus>.33</avg_ncpus>
<max_ncpus>.33</max_ncpus>

2600K + 7970 #2:
<count>.0625</count>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>

980X + 7970:
<count>.041</count>
<avg_ncpus>.5</avg_ncpus>
<max_ncpus>.5</max_ncpus>

2600K #1 is averaging 24 every 7-8 minutes.
2600K #2 is averaging 16 every 5-6 minutes.
980X is averaging 24 every 6.5-7.5 minutes.

It looks like I've reached my max at around 1 per ~19-20 seconds or 3 per minute, which works out to ~998K WCG PPD each. (3 x 60 x 24 x 33 (BOINC points average over 125K WUs) * 7 (WCG conversion))

I'm hoping that the longer individual times will help with the PV backlog. Only time will tell.

With PVs getting validated, I should be able to sustain ~3MM WCG PPD until I go broke paying the electric bill.

Aside: I wish they reported the time in seconds. Trying to keep track of exact times is tough when you have to look at boinc manager at exactly the right instant.

Suspend network for a while to get a bunch waiting to go then highlight each wu in turn and use "properties" to see the total and cpu time at your leisure.

Spent a little time playing with the 7950 and things like:

<avg_ncpus>0.8</avg_ncpus>
<max_ncpus>0.8</max_ncpus>
<coproc>
<type>ATI</type>
<count>.1</count>

At first glance The advantage is slim. I will probably need to let it run for a day to be sure.