SETI Is Online! 15 Feb 2011

[eaglescouter]

The project is back online.

Expect some significant traffic for the next day or so, thus it may be a while before all your machines can download a full que of work or upload the all the completed work units.

RAC shall begin climbing again!!!

[Bakkoda]

Oh man my poor RAC.

[eaglescouter]

Yeah, the RAC graph is looking pretty ugly right now, but a few weeks from now this project glitch will be a dim memory. But overall I think the SETI team at Berkeley does a pretty good job running an international distributed computing project on a shoestring budget.

[4GHZ_or_bust]

They may not be 100% full due to heavy demand for new WU and uploading completed WUs. I'm getting lots of HTTP failure but I did get some WUs on my main rig. Oddly, only 1 CPU WU, the rest are GPU only.

Edit: spoke too soon. Out of WU again for today.

[eaglescouter]

Since the splitters have been disabled while the project was offline that means that no new work units were available. Thus the project is now creating work units and handing them out as fast as they are generated. This will take a few days to normalize as the initial demand will be substantial.

[4GHZ_or_bust]

I got a few

Ran out again

Got a few more...

ran out again...

Seems that I get about 8 or 10 per hour, all only for GPU. GPU eats all of them and spit them out in less than 10 minutes.

[Pete Church]

I got a few

Ran out again

Got a few more...

ran out again...

Seems that I get about 8 or 10 per hour, all only for GPU. GPU eats all of them and spit them out in less than 10 minutes.

I got 1 GPU WU, more of a tease than anything... :). I believe by tomorrow this time we'll all a small cache built up again.

[Pete Church]

Btw, if some of you hadn't heard about it, you can use a Rescheduler (http://www.efmer.eu/download/boinc/scheduler/boinc_rescheduler_2_4.zip) to move GPU units to CPU and vice versa. Helps when WUs are low and you want to keep thing optimal. It also can be used to make sure VHAR and VLAR units never go to your GPU, if you have trouble with those types of units. You can also set it to run automatically and periodically (for large 10 day caches, don't run it more than once a day, it gets pretty slow).

EDIT: Sorry that link above is for the installer itself. A more descriptive post is here (http://www.efmer.eu/forum_tt/index.php?topic=428.0) for what the rescheduler is etc.

[Careface]

Woohoo!

Now if only I could get <exclusive_gpu_app> to work as intended, I would be able to just set it and forget it! :D Someone here may have some experience with it - I've asked on the SETI forums, but I should check here just in case too :)

Hey all :) Just wondering if anyone has any experience using this feature, and whether or not they've gotten it to work properly.

The idea behind it is that BOINC will stop running applications such as S@H whenever the program that you specify in <exclusive_gpu_app> is running... In theory.

What I've noticed instead however, is that the S@H application will only stop -after- the app I've specified in <EGA> is closed. Here's what I mean:

<cc_config>
<options>
<exclusive_gpu_app>mpc-hc.exe</exclusive_gpu_app>
</options>
</cc_config>


which gives me

14/02/2011 8:30:50 a.m. Starting BOINC client version 6.10.58 for windows_intelx86
14/02/2011 8:30:50 a.m. Config: don't use GPUs while mpc-hc.exe is running


So, in theory, the GPU should stop working whenever mpc-hc.exe is running. However, I start up S@H, mid way through crunching, I start up mpc-hc.exe (my media player - crunching on GPU lags my video playback something shocking), S@H still runs. I leave it for a minute or so to make sure it's not just waiting to close or some such, no dice. I go to close mpc-hc.exe, and lo and behold, S@H stops crunching. But, since mpc-hc.exe is no longer running, it starts S@H back up again o_O so essentially, all that is happening is S@H restarts whenever mpc-hc.exe is closed.. Which isn't -quite- what I'm after..

Any ideas? :)

EDIT: So, the <exclusive_app> (note: not gpu_app) function seems to work as intended - in which case my problem is solved. But for future reference, rather than halting all computing when a program is run, is there a way to halt just the GPU (i.e. get <exclusive_gpu_app> to work), so CPU crunching can continue? :)

[Pete Church]

I have not played with these options at all, but found this in one of the threads

<exclusive_app> is for CPU only. It can take anywhere from immediate to 10 seconds to respond.
<exclusive_gpu_app> is for GPU only. It can take anywhere from 10 seconds to a minute to respond (depending on the throughput speed of the videocard and the slot it's in).

So maybe you just have a wait even a bit longer than a minute for it to kick in?

[Saaby900t]

I hate it when WU get stuck in the "transfer Phase".

[4GHZ_or_bust]

Finally I have extra WUs waiting to be processed on my AMD rig. No more idling in between slow downloads.

[HankB]

Finally I have extra WUs waiting to be processed on my AMD rig. No more idling in between slow downloads.You're ahead of me. I did get a few work units, but I think I screwed some up trying to get issues sorted with the enhanced application from lunatics. I think I got that sorted and processed several regular work units this morning. :) And then I wound up with two astropulses stuck about half way downloaded. :( So I wound up with one SETI process. So I opened the spigot for Rosetta again. I only got two units so now I'm running 3 of four cores. At least my laptop has a couple SETI and Rosetta WUs ready to go.

[Careface]

I have not played with these options at all, but found this in one of the threads

<exclusive_app> is for CPU only. It can take anywhere from immediate to 10 seconds to respond.
<exclusive_gpu_app> is for GPU only. It can take anywhere from 10 seconds to a minute to respond (depending on the throughput speed of the videocard and the slot it's in).

So maybe you just have a wait even a bit longer than a minute for it to kick in?

Thanks for the info mate :) I looked a bit further into it, and apparently it's a known bug.

client: fix bug that caused GPU apps to not be suspended or resumed immediately after exclusive GPU app transition

Rather than running the risk of trashing the WU I have now, I'll just let my cache run dry then give it a go :) New rig will be up and running next week anywho :)