• Guest, The rules for the P & N subforum have been updated to prohibit "ad hominem" or personal attacks against other posters. See the full details in the post "Politics and News Rules & Guidelines."
  • Community Question: What makes a good motherboard?

Goodnight SETI - Final data is in the splitter queue.

Page 2 - Seeking answers? Join the AnandTech community: where nearly half-a-million members share solutions and discuss the latest tech.

Wiz

Diamond Member
Feb 5, 2000
6,453
14
81
Had it set to get the max work and am running multi multi cores and decent Cuda cards. But I will run out too, just be a bit longer for me.
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
I ran out a couple of days ago, thought I had a high cache set, but I've just looked and it was only at 2-4 days. What's the maximum out of curosity?
 

Wiz

Diamond Member
Feb 5, 2000
6,453
14
81
I believe it's 10 days. Have not checked but from memory that is what I had.
 

StefanR5R

Diamond Member
Dec 10, 2016
3,728
4,025
106
There are several limits in play. They are applied in combination, such that the maximum work buffer depth results from the minimum out of these different limits.
  1. Built into the client: "Store at least __ days of work" and "Store up to an additional __ days of work" can be set to 10 and 10 at most.
    These are not 10+10 24hour days though. Instead, the client keeps tabs on how much of the time of day the computer is on, and how much during this uptime the client is up and has computation enabled, and then reduces those 10+10 days in its internal calculation of how much work is going to fill those buffers.
  2. Built into the client: The client stops requesting new work when it has got 1000 or more tasks ready to run.
  3. Not sure if built into the client or into the server: The client stops receiving new work when, according to the client's projection based on former uptime (see item 1) it determines that it will be impossible to complete any more work before the task deadlines.
    This limit is ineffective at SETI@home because the project server sets task deadlines many weeks into the future.
  4. Configured at the project server: SETI@home's scheduler is giving at most 100 CPU tasks in progress to a client, and at most 100 GPU tasks in progress per each GPU on the client.
    Sorry that I made you read this far, but this 4th limit, not any of the earlier ones, is the most limiting one on any computer with somewhat recent hardware.
There are workarounds:
  • Use more than one client on a host.
  • Configure the client to "detect" more GPUs on the host than there actually are (e.g. 10 GPUs, such that up to 1000 GPU tasks can be had), but configure also that it then only starts as many GPU tasks in parallel as there are actual GPUs on the host.
  • Build the client from patched sources, such that it does not only "detect" more GPUs than there really are, but that it also accepts more than 1000 tasks ready to run in the queue.
    Here is an example of a host which, at the time when I checked, showed 33 GPUs to the server and had almost 5000 tasks in progress.
    This modification is very common among members of team GPU Users Group, who implemented it for their team members with the intention to buffer enough work to last during an entire Maintenance Tuesday despite modern GPU hardware and performance-optimized 3rd party application versions.
    A few users outside of team GPUUG implemented this modification independently too.
 
Last edited:
  • Like
Reactions: Assimilator1

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
Thanks for a detailed answer! :)

I would've upped the cache if I'd actually thought about it! :oops:
Too late now! Anyway, by shear fluke I ran out at a nice round ~8.5 million :).
 

QuietDad

Senior member
Dec 18, 2005
497
60
91
And just like that it's over. Last PC finished the last task and SETI gets retired for the time being. I sure hope they come up with a phase two. Now to waste the next three days disassembling 4 PCs, probably retiring two and deciding on what's next for them. For the first time since BIONC has been introduced, I have to research what to crunch next.
 

Markfw

CPU Moderator, VC&G Moderator, Elite Member
Super Moderator
May 16, 2002
20,885
9,080
136
And just like that it's over. Last PC finished the last task and SETI gets retired for the time being. I sure hope they come up with a phase two. Now to waste the next three days disassembling 4 PCs, probably retiring two and deciding on what's next for them. For the first time since BIONC has been introduced, I have to research what to crunch next.
Easy.... covid-19 research, either Rosetta@home or WCG (coming soon), or if you have video cards, Folding@home.
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
You've only ever run SETI? Wow! :openmouth: That's project loyalty! :).
Me, I'm interested in more projects than I have CPU cores for ;) (well, to give a decent output anyway).
It really depends what floats your boat, but with Covid 19 floating about causing 'issues' :p, a lot of us have focused on Rosetta@H & Folding@H.
I've switched to Rosetta & Folding for the time being (aside from a little GPU stability testing, as I've recently under volted my RX 580 & cut it's power draw by ~25w!).
 

QuietDad

Senior member
Dec 18, 2005
497
60
91
I occasionally would get into some of the team races and events in other things, but they tended to be on other machines I had built as burn int tests. These 4 where 90% SETI forever. One is my main desktop that will me used as that still, and I won't know what to do with "response time" or suspending BIONC when I needed to to anything and one is still the home web server but now that machine is way overkill to just run Apache. I may take one on my older Q6XX machines retiring and move the web server to that. We'll see. Electric Company will be complaining as the fleet was power hungry (1 machine 2x Xeon(R) CPU X5355 , a AMD FX(tm)-6300 Six-Core Processor , a Intel(R) Core(TM)2 Quad CPU Q9400 (overclocked) and if it wasn't in a proprietary SFF case so I could put a real video card and power supply in it (240 watts wont power anything) a Intel(R) Core(TM) i5-2400 . I am thinking of selling them off, or some of them anyway, But knowing the cost of building them,knowing I wouldn't pay that for them and not wanting to deal with today's world in selling something (too many people trying to scam or lowball you) I may look for a needy family or nursing home to donate them to with everything going on.
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
Response time?
As for suspending BOINC, I never found I needed to, even when playing games, I do though suspend any GPU crunching & reduce computing preferences to allow 3 real CPU cores for Elite Dangerous, seems to work fine, although I find when doing heavy combat I need to up that to 4 cores (it's officially required specs). I also suspend GPU crunching when watching films, never needed to touch CPU crunching there.
Yea the cost of building them will have nothing to do with what you could sell them for, that reminds me, I've got a Q9550, Asus mbdr & 8GB RAM to sell, I'll count myself lucky if I get £50 for that lot lol. Oh, as for low balling, that's what ebay reserves are for :p (but of course ebay has quite significant commission these days!). Never been scammed for selling, to date! But yea donation is a good plan :).
 

borandi

Member
Feb 27, 2011
138
117
116
Picked up a couple of SETI WUs today, they look like they are ones that timed out on other machines
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
Yea resends. I discovered when looking at my tasks list on the SETI website that I supposedly have 90 tasks in progress!
When I tried to update the optimised app on my 2nd rig a few weeks ago, all the tasks just vanished! :( I went back to the original app but the tasks never came back (& luckily I got more WUs so I didn't worry about it then), so I believe it's those & I'd forgotten about them.
I did ask on the SETI forum if I could recover them, but the only way I could is if I have a completed task ready to send, which I don't. Pity, as the tasks don't expire until well into May!
 
Last edited:

Endgame124

Senior member
Feb 11, 2008
435
269
136
I'm down to my last 3 SETI jobs, which will be complete in like 3 hours. Its been grand SETI!
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
A fond farewell then ;).

I'm sure they'll be back someday with SETI v2, afterall we've only searched a small fraction of the galaxy for a small fraction of time, over a small (?) portion of the radio bands!
If we ever get to the point where we're searching in all directions of our galaxy (at least), at all useable radio bands, continuously for at least a few decades & still find nothing, we could then reasonably assume that ET doesn't use radio. Then we'd need to do the same in the optical wave bands too!
If nothing from either of those 2, then maybe we are alone! (which to me seems unrealistic), or they have some other form of comms, or are much more stealthy than we are! ;)
The above work is probably decades worth of work, even if we could process all that data in real time, which we can't, although maybe quantum computing could help a lot??

Thoughts?
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
I heard on the SETI forums that a bunch of WUs had got through (resends), not got any myself, but then BOINC isn't asking for more WUs atm as I significantly cut the work queue (due to Rosetta's much tighter deadlines. I set it from 2+2 to 0.5+0.5! The queue is still shrinking atm).
 

Endgame124

Senior member
Feb 11, 2008
435
269
136
When I setup boinc, I only had 1 GPU in my host (the integrated card with the APU / A10-7870K). I added another 2 weeks ago, but Seti doesn’t seem to use it... is there a way to tell boinc / Seti to check for new hardware?
 

borandi

Member
Feb 27, 2011
138
117
116
Check the project page in case there's zero work to send.
Check the program applications page to make sure there's even a GPU application to begin with.

Anyone still getting SETI units now is just lucky.
 
  • Like
Reactions: Assimilator1

Endgame124

Senior member
Feb 11, 2008
435
269
136
I have 124 Seti GPU jobs in my queue. I would LOVE to be able to use both the integrated GPU as well as my r250X to process the queue more quickly, but I'm not sure how to get it to put load on the r250 as well as the APU GPU.
 

Assimilator1

Elite Member
Nov 4, 1999
23,569
224
106
Hmm, I think that requires editing the config file, I know I've asked about this before as I was going to have my HD 7970 as a 2nd GPU.
I'll see if I can find the relevant thread......
Yep, this one, think it's still relevant, but it is a few years old!
[edit2]That might not be much use to you after all, no talk of editing config files there as I thought!
Maybe as borandi was implying their is no GPU app for your iGPU, you'd need to look on the SETI apps page for that.
 
Last edited:

Endgame124

Senior member
Feb 11, 2008
435
269
136
Hmm, I think that requires editing the config file, I know I've asked about this before as I was going to have my HD 7970 as a 2nd GPU.
I'll see if I can find the relevant thread......
Yep, this one, think it's still relevant, but it is a few years old!
[edit2]That might not be much use to you after all, no talk of editing config files there as I thought!
Maybe as borandi was implying their is no GPU app for your iGPU, you'd need to look on the SETI apps page for that.
I don't think I explained the behavior well. To be clear:
iGPU - getting work and processing SETI data.
r250x card - idle, not picking up SETI jobs.
 

Endgame124

Senior member
Feb 11, 2008
435
269
136
Ah ok, maybe their are no AMD GPU WUs? Did you look at the task server page?
Yes - I have 124 tasks that use .26 (or some similarly small number) CPU + 1 AMD / ATI GPU.

The iGPU is AMD (built into the A10-7870K) and is listed on the rosetta page as "AMD AMD Radeon(TM) R7 Graphics (7208MB) OpenCL: 2.0".

Unfortunately, the r250x does not appear to identified. Restarting Boinc / rebooting does not pickup the r250x (but it works fine for folding). Can boinc use multiple GPUs, or only 1?
 

ASK THE COMMUNITY