Thanks Jacob!

Well! Just got two Gerards for one of my PCs. Queue remains
empty...

I am sure there are many, like me, who would welcome a status report from the project team. The silence is deafening.

I'm not in the project team, but I'm keeping one eye on the project's webpages.

Well! Just got two Gerards for one of my PCs. Queue remains empty...

I see on the webpages (and on my hosts) what you've just experienced:
The long queue is continuously (couple of times in a day) filled with work, without significant reserve.
I guess that the results are needed as soon as possible for the ongoing research, so it's practical to produce short batches with no reserve, as if there were twice as many batches in progress at the same time, it would take twice as much time to complete all of them.

I guess that the results are needed as soon as possible for the ongoing research, so it's practical to produce short batches with no reserve

You're guessing, I'm guessing, EVERYONE is guessing!

Pretty please, project team, tell us what's happening...

As Retvari suggested, the amount of WU we have now in circulation allows us to have a fast turnover which is ideal for our adaptive protocol, that analyses new returned simulations as soon as they arrive and generates new ones from the results obtained.

Anyway, I will start preparing some extra WU.

I apologize for not answering right away, but as you can imagine sometimes academic life doesn't give you time for even checking the forums...

Would it to not be beneficial to the project to have a policy of only 1 WU for each available and free GPU. You would save a lot of time as some users cache WU's for hours that could be running on a machine that has no work.

As Retvari suggested, the amount of WU we have now in circulation allows us to have a fast turnover which is ideal for our adaptive protocol, that analyses new returned simulations as soon as they arrive and generates new ones from the results obtained.

Could be the "very long" workunits more appropriate for this method? Perhaps a bit better prepared than last time :)
There should be a 3rd queue for this kind of workunits, which would send work only for GTX 780, 780 Ti, 970, 980, 980 Ti, Titan, Titan X cards?

Bit of a short WU drought....

I would like to keep my two GTX 650Ti GPUs busy with GPUGrid WUs. There are in short supply: any idea when we will have more?

I am using the two GTX 660Ti GPUs for long WUs and hope there will be enough available for this long weekend in Canada - long and short WUs welcome.

Happy Thanksgiving, Valencia Day and any other special day to all wherever you may be.

JCM

Sure would love more work units. Have 2 hungry GPUs waiting for work.

Thanks.

Sure would love more work units. Have 2 hungry GPUs waiting for work.

Thanks.

Agreed the process is taking too long, or people are too busy to keep up with it. Lack of units is forcing people into backup projects, which then means you have to be extra lucky to ask for new units just at the right time when there are some, or people will stay on the backup projects forever. At some point alot of people will just stop asking, which then means those who stay could get some work, but will it be worthwhile doing at that point? I hope this does not become a downward spiral for GpuGrid!!

Not sure how I have setup my backup project which is poem@home. But when there's work available for gpugrid my client downloads the WU and my backup project stops and the GPUgrid WU starts. Works well so far

Gerard has alread explained why WU's are short so stop whining.

Lots of software changes have been happening in the last three weeks (essentially all analysis/simulation building programs in Matlab have been recoded almost from scratch to Python) and most of the lab has been dedicated to this endeavour. This left me as the only net contributor to GPUGRID and, as I explained elsewhere (https://www.gpugrid.net/forum_thread.php?id=3570), the strategy I took is to send the exact number of Workunits to keep the whole the network busy and allowing a fast turnover following our adaptive scheme (for which each simulation generates a new one after analysis of the results of previous ones). This is the cause for the slight WU availability drop.

Right now im coming back with a (overwinter, homeheating) 780TI and no work, noo good start *wait wait waitiinnggg* :)

---

---------- 24/7 Crunching since 2011 -----------
DSKAG Austria: http://www.dskag.at

I'll be sending some heavy long WU later this afternoon

EDIT: batch name is CXCL12_DIM_3GG_c36; takes around 12.2 hours in my tesla K20c (according to prediction).

I'll be sending some heavy long WU later this afternoon (batch *3GG_DIM_TRIM*)

♫♫ Give them to us, we're worth them! ♫♫

---

From skimming over this thread I understand the shortage of WUs comes from the benefit of doing faster/better science? If I am wrong please correct me.

If not then the tone and the attitude of some of you guys is like that of entitled spoiled brats.

Indeed you are correct. Takes not much more work to send double the number of WU. However, cueing WU slows the overall thoughput.

I normally set my "Resource share" to 0% anyway for this project, in order to obtain a zero-buffer size. That of course maximizes the time available for crunching and allows me to get the 24-hour bonus with some cards.

Maybe you could set up a separate queue for that purpose? Some people may like it, and others may not.

Out of WU today ?

There should be plenty of WU in the next hours.