Sluggish Response Times

[Crazybob]

I know it's in here somewhere, but what do you need to set so that your screen isn't dying trying to reload? This thing is toooo slooooow.

[joker]

Your computer or your internet?

[Maxwell]

Are you running the defaul app, or are you working with an app_info file?

[Crazybob]

This is the entire machine as far as screen response time.

Running with app_info file to run 2 WU's at a time. I know I saw a post regarding a cmd line you can add to give a little more priority to running programs. I just couldn't remember which post it was under.

[Bryan]

I too was facing a little slow down in response and there is some polling and something else dealing with frequency (size of wu packet sent to the GPU). I run the app_info w/ 2 wu per card.

All my systems have 2 GPUs each on MW. When they 1st started with the new app I turned off the CPU crunching and never experienced errors on MW. I finally decided to crunch AQUA and QMC and when I did that I started having some "validate errors". If I turned off the CPU then all errors went away. I played with different BOINC # cpus etc and couldn't find a good solution.

So:

I now run the app_info and set the "avg_cpu" so that 2 full threads are devoted to GPU scheduling. That means if I run something like QMC on my I7 then only 6 cpu wu are crunching.

Then:
Install Process Lasso (really cool program) and you can set MW to run at high priority. More importantly I set whatever CPU project I'm crunching to only run on cores/thread 0 to 6 and set the affinity for MW to 0-7. That reserves a full thread to MW for exclusive use by the GPU.

All errors went away when I did this. With running a single GPU then you can probably get away with only devoting 1 thread to the GPU and crunch 7 CPU wu.