-
MilkyWay@Home requires double-precision, so it performs well if you have a higher-end GPU (like Radeon HD 79xx). On such GPUs you can complete a WU under 1 minute. Lower models are running DP at a considerably reduced clock, so the performance is relatively weak.
Einstein@Home uses by default the GPU up-to ~75% when running a single task, but you can easily configure multiple WUs/GPU in their control panel (no need to use app_config). BRP4 tasks support AMD via OpenCL, but nVidia tasks run currently using CUDA 3.2, which isn't optimal performance for later (Kepler) family. Apparently, CUDA 4.x/5.x has a bug which causes BRP4 tasks to fail, but nVidia didn't fix this yet.
GPUGrid tasks run GPU at high load, on Kepler GPUs they also require a full CPU core for feeding the GPU. There are long tasks, which can run several hours (more than 12h on slower GPUs) and short tasks (few hours). They have CUDA 3.x and 4.x tasks for older and newer families. They give credit bonus if you complete a task within 24h after download, so it's required to set a low WU cache and set <report_results_immediately>1</report_results_immediately> in cc_config.xml
It seems to me, that there are currently no tasks in queue...
Last edited by Mumak; 04-24-2013 at 11:56 PM.
Posting Permissions
- You may not post new threads
- You may not post replies
- You may not post attachments
- You may not edit your posts
-
Forum Rules
Bookmarks