- Dec 29, 2005
- 6,292
- 2,239
- 136
The "long" WUs give a double points bonus so those are the only ones I crunch. In your GPUGrid account, open up the GPUGrid preferences and under the "run only the selected applications", you'll see the following options:
ACEMD short runs (2-3 hours on fastest card): no
ACEMD beta: no
ACEMD long runs (8-12 hours on fastest GPU): yes
Docking CPU only app: no
I chose yes for ACEMD long runs and no for the others.
On top of double points for the long runs, if you return the WU within 24 hours, you will receive another 50% in bonus points and if you return the WU within 24-48 hours you will receive a 25% bonus.
My GTX570 finishes the long WUs well before the 24 h deadline.
With the latest Nvidia drivers, the GPU app uses less CPU time, but to be sure I free up a CPU for GPUGrid use. Also you'll want to upload results as soon as the WU finishes and also keep a short WU cache to prevent download of a WU until just before you need it. The time is calculated from Download time until the WU is fully uploaded.
I use a cc_config.xml file to report results immediately and to free up a core/logical core. This is the one for my 3930K which has 12 logical cores (I set number of cpus (ncpus) to 11):
<cc_config>
<options>
<report_results_immediately>1</report_results_immediately>
<ncpus>11</ncpus>
</options>
</cc_config>
ACEMD short runs (2-3 hours on fastest card): no
ACEMD beta: no
ACEMD long runs (8-12 hours on fastest GPU): yes
Docking CPU only app: no
I chose yes for ACEMD long runs and no for the others.
On top of double points for the long runs, if you return the WU within 24 hours, you will receive another 50% in bonus points and if you return the WU within 24-48 hours you will receive a 25% bonus.
My GTX570 finishes the long WUs well before the 24 h deadline.
With the latest Nvidia drivers, the GPU app uses less CPU time, but to be sure I free up a CPU for GPUGrid use. Also you'll want to upload results as soon as the WU finishes and also keep a short WU cache to prevent download of a WU until just before you need it. The time is calculated from Download time until the WU is fully uploaded.
I use a cc_config.xml file to report results immediately and to free up a core/logical core. This is the one for my 3930K which has 12 logical cores (I set number of cpus (ncpus) to 11):
<cc_config>
<options>
<report_results_immediately>1</report_results_immediately>
<ncpus>11</ncpus>
</options>
</cc_config>