• We’re currently investigating an issue related to the forum theme and styling that is impacting page layout and visual formatting. The problem has been identified, and we are actively working on a resolution. There is no impact to user data or functionality, this is strictly a front-end display issue. We’ll post an update once the fix has been deployed. Thanks for your patience while we get this sorted.

Milkyway@home running multiple WU at once

waffleironhead

Diamond Member
With the switch over at milkyway to much shorter WU based of of the modified fit program I was able to run the project on my HD6950 GPU now.

Since running the WU's I noticed that my gpu was not being fully loaded, so I decided to try out an app_config.xml file to see if running more at once would better use my gpu. I decided to share my results with you all.

These are the wu times based off of different configs.
1/1 is 1 gpu wu running while using a single cpu core
.5/.5 is 2 gpu wu while each using .5 cpu core
.33/.33 is 3 on the gpu each using 1/3 of a cpu core
.25/.25 is 4 on the gpu each using 1/4 of a cpu core
.125 is 8
.0625 is 16

Running 1 unit per GPU with 1 CPU : 23 seconds a WU
Read as
1/1 : 23 seconds each
.5/.5 :36 seconds/2 =18 seconds each
.33/.33 :49 seconds/3 =16.33 seconds each
.25/.25 :66 seconds/4 = 16.5 seconds each
.125/.125 :132 seconds/8 = 16.5 seconds each
.0625/.0625 : (getting anywhere from 180 seconds to 250 seconds) averaged to 225 seconds/16 = 14.065 seconds each.
I tried doing 32 units at once, but the system became unresponsive and I had to do a hard reset.

Seems strange, but running 16 wu at the same time netted the best results.
I do have some errors showing up now and a lot of inconclusive wu, so will have to wait and see if things are actually better running 16 at once over time. They system is pretty slow to use as well, so not very good for usability.

app_config.xml file contents
<app_config>
<app>
<name>milkyway</name>
<gpu_versions>
<gpu_usage>.0625</gpu_usage>
<cpu_usage>.0625</cpu_usage>
</gpu_versions>
</app>
</app_config>
 
I usually only ran 3 at a time. The biggest problem with running too many at once (on a dedicated machine where responsiveness doesn't matter) is that Milkyway doesn't give you enough tasks between updates.

This was ridiculously the case when I had 4 280X's on one box, each running 3 at a time. 😀
 
Yeah, I can see how the 60 WU limit may be way too small when running multiple GPU and
running multiple wu at once. Add in the 60 seconds between asking for more WU and I could see running dry quite easily.
 
Not sure what happened. I turned off Milkyway for the primegrid race. Now that I turned it back on WU times are now 6-8 minutes for the 16 at once settings. That totally blows the efficiency out of the water. Maybe these are some different type WU? Will have to do some looking.
 
Back
Top