Message boards : Number crunching : Discussion of Boinc 7 Configurations
Author | Message |
---|---|
I'm starting this thread for anyone to discuss/ask questions about Boinc 7 (still alpha) configurations/settings. Boinc version 7 is still undergoing Alpha testing and development. It is therefore not yet recommended, however there are some changes worth noting at this stage. Stipulating which GPU project is allowed to run Some projects offer both GPU and CPU tasks. You can now specify to include or exclude GPU tasks by project. On multiple GPU systems, you can also stipulate which GPU runs which project, using new Boinc cc_config.xml options. See the <exclude_gpu> configuration options at http://boinc.berkeley.edu/wiki/Client_configuration. For example, here's part of Jacob Klein's cc_config.xml <options> section. The GTX 460 GPU only runs GPUGrid.Net, while the 2 slower 9800 GT GPUs (which cannot handle GPUGrid.Net) run other projects: <use_all_gpus>1</use_all_gpus> <!-- ======= SETUP GPU 0: GeForce GTX 460 ======= --> <exclude_gpu> <url>http://einstein.phys.uwm.edu/</url> <device_num>0</device_num> </exclude_gpu> <exclude_gpu> <url>http://setiathome.berkeley.edu/</url> <device_num>0</device_num> </exclude_gpu> <!-- ======= SETUP GPU 1: GeForce 9800 GT ======= --> <exclude_gpu> <url>http://www.gpugrid.net/</url> <device_num>1</device_num> </exclude_gpu> <!-- ======= SETUP GPU 2: GeForce 9800 GT ======= --> <exclude_gpu> <url>http://www.gpugrid.net/</url> <device_num>2</device_num> </exclude_gpu> The Network Usage work buffer options have changed They are now: "Minimum work buffer: Try to maintain enough tasks to keep busy for this many days" and "Max additional work buffer: In addition, maintain enough tasks for up to this many days." Here's how it works: When the resource's work level falls below [Minimum work buffer], the client will request enough work for the resource to stay busy for [Minimum work buffer + Max additional work buffer]. When it requests work, typically it will receive that [Minimum work buffer + Max additional work buffer] amount of work from only 1 project. Note: Using an extremely low minimum work buffer, like 0 or 0.005, discourages project variety on the client, since it will only request work when it is essentially completely out of work, and then usually just gets work from 1 project. For clients always connected to the Internet, to maximize bonus credit without putting undue stress on the servers while also ensuring variety of projects on the client, ideal settings are: Minimum work buffer: 0.20 days Max additional work buffer: 0.20 days For clients always connected to the Internet, to maximize credit bonuses without putting undo stress on the servers while not caring about variety, ideal settings are: Minimum work buffer: 0 days Max additional work buffer: 0.40 days - Credit to Jacob Klein
| |
ID: 24030 | Rating: 0 | rate: / Reply Quote | |
Apart from what you have above it also only does scheduling checks once a minute, so may take up to a minute before it will do a schedule request. Earlier clients such as 6.12.34 would do one almost immediately. | |
ID: 24033 | Rating: 0 | rate: / Reply Quote | |
I've been interested in the functionality now provided by <exclude_gpu> ever since my old GT 240's became outdated here at GPUGrid. I decided to jump in and play around with BOINC 7.0.20 and the new functionality. | |
ID: 24071 | Rating: 0 | rate: / Reply Quote | |
I've been interested in the functionality now provided by <exclude_gpu> ever since my old GT 240's became outdated here at GPUGrid. I decided to jump in and play around with BOINC 7.0.20 and the new functionality. I have a couple of machines where the GPU is dedicated to GPUgrid work. I have been using the <exclude_gpu> on one of them running 7.x with some success. It also solved an annoyance with CPDN. My cc_config looks like: <cc_config>
<options>
<http_1_0>1</http_1_0>
<exclude_gpu>
<url>http://climateprediction.net</url>
</exclude_gpu>
<exclude_gpu>
<url>http://einstein.phys.uwm.edu</url>
</exclude_gpu>
</options>
<log_flags>
<sched_op_debug>0</sched_op_debug>
<work_fetch_debug>0</work_fetch_debug>
</log_flags>
</cc_config> The machine runs 3 projects, namely CPDN and Einstein on the CPU. The GPU (and a single core) are dedicated to GPUgrid. ____________ BOINC blog | |
ID: 24074 | Rating: 0 | rate: / Reply Quote | |
Message boards : Number crunching : Discussion of Boinc 7 Configurations