SETI@Home Sprint

  • Thread starter Deleted member 88227
  • Start date
D

Deleted member 88227

Guest
I've got all my rigs (CPUs and GPUs) rocking on the project. All but three of my rigs are offline until around this time tomorrow. Gonna dump my load just before I go to bed. They're wingman projects so we're probably gonna leave a lot on the table after the sprint. I've already got 281 pending results at the time of this post. I had around 220 before I started bunkering this afternoon as some of the units have been in queue for a week or more now.

Hopefully applejacks can get his issues solved before tomorrow.

Also hoping none of my boxes are experiencing any issues as they're all offline working.

Our current biggest competitor is Team Art Bell and they're averaging around 500k points per day.

The [H]orde on the other hand seems to be doing around 100k points per day. So lets hope we can multiple our production by a good bit.
 
My current status since attaching last night. Didn't try to do any kind of bunkering as I, perhaps incorrectly, thought that was pointless. Whoops. :( That said, if only WUs validated after the Sprint starts are the ones that count, then I'm sitting on a decent amount.

State: All (4204) · In progress (1184) · Validation pending (1906) · Validation inconclusive (18) · Valid (1080) · Invalid (0) · Error (16)
​
 
Well I think it's important for projects that require a wingman. Simply because you have a higher chance of the first batch of work units getting validated before the project is over. So all WU's that are validating right now are probably from a wingman's work days/weeks ago that are immediately validating.

That's a crap ton of WU's though. I think I grabbed about 2100 before I took my main rigs offline. Three rigs require net access 24/7 (well technically only two) because of web services I host on them. My main rig (what I am posting from) could be disconnected tonight before I go to bed since I wont be getting on it until I get home from work tomorrow. So I think I'll have about 1000 - 1500 WU's to dump tomorrow when I grant them net access again.

State: All (2799) · In progress (2155) · Validation pending (294) · Validation inconclusive (1) · Valid (46) · Invalid (0) · Error (303)
 
Well, you know me... I typically focus on one project and nothing else. ;) Guess we'll see if my spacing on trying to bunker was a mistake or not!
 
Well, so far, so good. Looks like TAB dropped their bunker in the second hour; but wasn't enough. Right now we're sitting pretty in 1st by 140,579 points.

BTW, don't know who AEPZ is, but they're really churning out the WU's and have been for quite some time.
 
Okay, dumped my load.

I had around 60 valid results before I dumped it. Now I am sitting at 1113 valid results and 787 pending with 10 inconclusive. Took my score from around 630k points to almost 710k points.

Should give us a good boost. All rigs are currently crunching 302 tasks with plenty in queue. (~1500)
 
I've being playing around with some optimisations for Seti with limited success. After the Collatz gravy train, this project has really brought me down to Earth with a bang. Some observations:

  • Firstly, a statement of the obvious. Fast recent nVidia cards want opencl rather than cuda tasks. It takes a while for the server to figure out that opencl is better for your card but it seems to have managed it for 4 out of my 5 gpus. I think it was encouraged by me running only 1 opencl at a time versus 2 whenever cuda tasks popped up. The app_config I used was:
<app_config>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda42</plan_class>
<avg_ncpus>0.5</avg_ncpus>
<ngpus>0.5</ngpus>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>cuda50</plan_class>
<avg_ncpus>0.5</avg_ncpus>
<ngpus>0.5</ngpus>
</app_version>
<app_version>
<app_name>setiathome_v8</app_name>
<plan_class>opencl_nvidia_SoG</plan_class>
<avg_ncpus>1.0</avg_ncpus>
<ngpus>1.0</ngpus>
</app_version>
</app_config>​

This is a quite generous with the cpu allowance but I like to err on the side of caution.​
  • When you finally get a steady stream of opencl tasks, additions to the seti mb_cmdline-8.22_windows_intel__opencl_nvidia_SoG.txt file seem to speed things up a bit. Given a lack of familiarity with the sub-projects, it's hard to tell by how much as the run times are damn inconsistent and the whole thing could well be a figment of my imagination. There are numerous flags you can enter but I kept it simple and for the 1080 Tis (plenty of memory) just added:
-sbs 2048 -period_iterations_num 1 -high_perf
​
  • The optimisations app was a bit of a disaster for me. Maybe it was released prior the appearance of the MB V8 opencl app which would explain why it deleted that app and all it's tasks just leaving me with cuda tasks. This is more than likely due to user error and maybe I should have installed it prior to any wus being downloaded but I'll be leaving this one alone for now.

We seem to be dumping our way to another 25 FB points! Nice going guys.
 
Seeing as how I failed horribly with the R@H sprint I aint touching shit. I'm getting tasks so I'm gonna leave it at that.
 
lol, always best not to touch shit after you've dumped you're load.
 
We are straight up crushing it.

seti3.jpg
 
And we're still killing it with a 938K lead and only a few hours to go. Big congratz to everyone involved with a special mention to fastgeek for contributing so much to the unassailable lead we built on day 1.
 
I've already submitted 'no new tasks' for my rig. After this - I suppose I'll just get back to the marathon (err - start on it, any specific projects you would suggest that I fetch after the sprint comes to a conclusion?

Thanks :)
 
Last edited:
You can stay on SETI@Home if you want. We're 9th place in the Marathon.
 
Nah, I'm done with trying to help listen for little green men. :p Pulled in ~735K over the sprint and still have that much again pending. Honestly not sure if this project as a Marathon is worth pursuing. The leaders are heavily entrenched and, chances are, everyone above us is throwing a good amount at this on a regular basis. That and, honestly, while I think it's absurd to think that nothing else is out there, I'm honestly not sure that this projects approach has a snowballs chance in hell of finding anything worthwhile. (Note this is based on an old opinion from what I read they were doing at the time; maybe it has changed.) To each their own, of course! :)
 
Tend to agree. I think Seti should be somewhere near the bottom of our FB things to do pile. Especially so, given the current dearth of Astropulse work.
 
Last edited:
You must log in or register to reply here.
Back
Top