> Do you only have one ClimatePrediction model running on your machine? I
> noticed the same thing on my dual-cpu machine. But then BOINC decided to
> download another CPDN unit and both CPU's are now going 50-50. I decided that
> perhaps one CPU was splitting itself between according to my preferences, and
> the other was running exclusively SETI since it didn't have a long one to
> run.
>
> Just a guess on my part.
>

As it turns out, I have 2 CP work units on the machine. Given how long they take to run, there is apparently no need to download more. The machine has 5 S@H workunits though - given how quickly these things finish, the machine tends to download another one as soon as one finishes.

At this moment in time, 4 of rhe S@H workunits are running, and both CP workunits have been preempted. There is one S@H workunit queued up in the ready to run status.


It does sound like the problem has been resolved - we just need to wait for an updated release.

4.05 has an apparent bug that counted WUs instead of projects when determining share. For example if your shares were set up 50-50, and you have 2 S@H WUs and 1 CPDN WU (for simplicities sake) the actual share would be 66 S@H and 33 CPDN.

The good news is that this has apparently been fixed in 4.08. The bad news is that 4.08 has apparently broken modem connections (worse than they were) so it will probably not be put into production.

Do you only have one ClimatePrediction model running on your machine? I noticed the same thing on my dual-cpu machine. But then BOINC decided to download another CPDN unit and both CPU's are now going 50-50. I decided that perhaps one CPU was splitting itself between according to my preferences, and the other was running exclusively SETI since it didn't have a long one to run.

Just a guess on my part.

I am running BOINC on a quad-cpu machine, and I have both S@H and ClimatePredictor running on it. Supposedly they have equal weighting, but what I notice is that whenever a S@H workunit is completted, it tends to preempt whatever work might be going on with ClimatePredictor, and put the thing to work on a S@H workunit. Given that each S@H workunit completes in about 6 hours, this preemption takes place every 2-3 hours. The net result is that I am tending to see 3-4 CPUs being used for S@H, and 0-1 CPUs being used for ClimatePredictor.

I suppose the problem is that ClimatePredictor has workunits that take ~1000 hours to complete, so it doesn't get the benefit of this same pre-emption behavior.

I suppose what I could do is goose up the priority of CP to try and manually balance the thing.