Thanks Rob, much appreciated. After reading your description and the other threads on app_config I now understand at least this part of it.

When the ati_hcc1 betas started I was running 7.0.42 and immediately started getting errors after a few seconds that said I should reset the project. I did but before I let it run I read what branjo said about not getting any client higher than 7.0.31 to work so I immediately downgraded to it and the units started running good taking about 5 min a unit on my HD 5770 which rates low in power for the GPUs.

Everything was going great, my monitor was lagging a little but nothing to worry about, when on the 30th I started getting WUs that were due the next day and were running High Priority. Thats when my monitor started seriously lagging, too much. I started looking at app_config thinking that it might be a way to throttle back on the GPU a little, but after looking at all the info determined that the newer clients and especially 7.0.45 will run the GPU with the same preferences as the CPU, which may be the answer I am looking for. If not then I will have to abort those GPU WUs that run High Priority.

I have a suggestion for the developers about how the order WUs are run because I will get WUs that are due sooner than WUs I already have so they sit there until they are due the next day and then run High Priority which gives me a little trouble. Seems the project app or the client runs the WUs in the order of when they are received instead of when they are due, which to me makes more sense, wouldn't have the High Priority problem unless the job is rated that way.

I just upgraded my 2009 Mac Pro cpu from 4 core to 6 core Westmere, about 4 months ago, which runs 12 threads at 3.33ghz and does it fly through WUs. I have climbed about 4000 in Points Generated Ranking in that little time so it doesn't seem to have any problem with running any of the projects and I have the preferences set to 50% usage to keep the cpu temp in line (I also use SMC FanControl to help with that) so I am hoping that running the GPU at 50% will do what I need.

Sorry for the length of this post but I wanted to get my sometimes frustrating experience of trying to participate in the WCG project out there in hopes that it might help someone else.

Thanks again for your help.

Branjo, did you ever get a version higher than 7.0.31 to run on your Mac?

I just noticed that I didn't get any betas for FAAH, wonder why?

David

David, missed your feb 1 post till now. Seen a few posts that state their downgrading from a 7.0.4x to 7.0.31 resolved failures, where 7.0.31 is supposed to be the first to install proper on OSX Lion. Will be interested to learn if 7.0.4x will process the next batch of beta properly as else this could hint at some bug in the client [cant see app_config.xml being part of this, it merely restricting the concurrent processes of same sciences, unless it's too slow. We know from past testing that some sciences can be crashed if they're interrupted in the first few seconds when the model is being set up... therefor LAIM on, it seemed to reduce that issue, but LAIM does not work for GPU, only for CPU tasks]

My sample app_config.xml showed that the HCC tasks for the GPU side are limited to just 1 allocating 1 full core. Setting <gpu_usage>0.5</gpu_usage> for the GPU control and the CPU control at <cpu_usage>1.0</cpu_usage> will give an info point if that's stable, then slowly build this up to see what remains stable. (Regrettable the client has to be restarted each time, hopeful a future client will get a "read app config" function on the Advanced menu or be integrated into already existing "read config". Not seen it in GIT yet [but lots of Android work being put in these last weeks].

Is there any official word on the status of this? I assume it was sidelined because of FAAH beta, but could we get an update?