Did you try something like that?

MrS

---

Scanning for our furry friends since Jan 2002

IDK about PG, but I did not. How would I do that?

@ETA yes

@Otis: either in your preferences on the home page or locally under extras / settings / network usage.

MrS

---

Scanning for our furry friends since Jan 2002

I wasn't sure how to do the resource share. I have cache down to one day which is as low as i'm willing to do since it's also running WCG and internet goes out regularly for up to 12-16 hours.

So at one day it still DLs 6.

On the homepage of each project you can go into "your account" and in the settings section the 3rd point should be called something like "ressource share and graphics". It's the top most option, by default set to 100.

If you have 2 projects, one with 100 and the other with 200 then BOINc will try to allocate 100/(100+200) = 1/3 of the computing time to the first project and 200/(100+200) = 2/3 to the second. The caches for individual projects are adjusted accordingly (in theory).

MrS

---

Scanning for our furry friends since Jan 2002

It is looking like I'm getting something like this problem...The 295 is running at around 1 WU every 7.5 hours per side but grid is down loading 8 at a time so the last 2 will fail the 24 hour since download timelimit and from that point on so will the rest this is however day1 so maybe things will automatically adjust??

Ihave no preferences set for grid on this rig this is just what I have been given automatically....boinc messagas gives max 8 wu/day

Addendum: I've moved my 1 day of work cached to 0.8days in boinc manager to see if this has an impact...Will report tomorrow

If it fetches 8 at a 1 day setting, 0.8 probably won't be low enough. This also depends on your BOINC client, though. Clients prior to 6.6 have been notoriously unaccurate regarding caches and ressource share in the presence of GPUs.

MrS

---

Scanning for our furry friends since Jan 2002

Ok, thanks for that.......there was no change at 0.8 but I've just received credit for a Wu that took >24hours from download to completion and it seems to have not effected the points bonus ? (not complaining :))

further observations will be reported

Now I'm a little embarrased....where did I get that 24hrs limit from???

Made me go searching.....

from the FAQ

So now, we apply a multiplier of 2.0x for all workunits, but give and additional 25% for WUs returned within two days. This is useful for us to reduce latency of the results while keeping us in line with a correct value for credits as in SETI.

I got 10 on one machine. It has BOINC version 6.10.3 and a GTX295. Resource share is 50% for the project and cache is set to 1 day. They have been taking around 8 and a half hours so I would have considered 6 to be sufficent for the day (ie 3 per GPU)... if it had a 100% resource share.

---

BOINC blog

*Nice* to hear that 6.1.0.3 is not much better in this regard. For 6.5.0 there's a reason why I wrote "I've got GPU-Grid set to 1/4th the combined ressource share of the cpu projects and have a cache setting of a few hours (so BOINC caches enough work for about 1 day)". I have no clue which logic tells it that a hand full of hours = 1 day, but that's the way it works. Lower your settings and see what you get.
Edit: with 6.5.0 I need the ressource share to be about (number of GPUs / number of CPUs) * combined cpu ressource share. Otherwise the long term debts get out of whack much quicker and at some point one of the ressources runs dry. Could be that 6.6. behaves better in this regard due to the distinction between "debts" and "CUDA debts".

MrS

---

Scanning for our furry friends since Jan 2002

I put my GTX 260 216 back on GPUgrid after having it on F@H for more than a week and when I come back I again get 6 WUs. I have the Resource Share at 80% WCG and 20% GPUGrid. I have changed my Addition Work BUffer to different numbers and still get 6 WUs.

Is there no solution to this?

Vista 64, Boinc 6.6.36, 190.62

I put my GTX 260 216 back on GPUgrid after having it on F@H for more than a week and when I come back I again get 6 WUs. I have the Resource Share at 80% WCG and 20% GPUGrid. I have changed my Addition Work BUffer to different numbers and still get 6 WUs.

Is there no solution to this?

Vista 64, Boinc 6.6.36, 190.62

I have a Quad with 2x 8800GT and 1x 9500GT. When starting GPUGrid after a fresh install of Windows, I get 10 WU's immediately. I had BOINC set to get 0.25 days of work (to try to limit that). However, it appears that after the initial 10, if I set it to a higher number (like 1 day of work), it will keep it around 6 WU's at a time.

The only thing I can think of is that initially you get WU = 2 * # of gpu's + 1 * # of processors.

In my case, that is 10. For your case, that is 6. I haven't been able to get less than 10 initially, but it corrects after a while by not downloading new WU's. Maybe give it a couple days to see if you don't download anymore and get closer to your desired # of WU's in queue.

I was able to decrease the number of workunits in the queue on my machine by decreasing

(estimated time between internet connections + amount to leave in the queue)/average time each workunit takes

to less than 1.

Now only one workunit fetched at a time, but just one or no workunits not started yet normally in the queue.

I have a Quad with 2x 8800GT and 1x 9500GT. When starting GPUGrid after a fresh install of Windows, I get 10 WU's immediately. I had BOINC set to get 0.25 days of work (to try to limit that). However, it appears that after the initial 10, if I set it to a higher number (like 1 day of work), it will keep it around 6 WU's at a time.

Don't forget that this project is still vastly under-estimating the compute times of the tasks it sends out. Once BOINC has completed a task, it can calculate a DCF to correct the faulty estimates and fetch a more sensible number next time: but even after the doubling of the estimates last time I complained about this, I'm still seeing a DCF of 8.825382 on my best-adapted machine.

With a fresh installation of Windows, and presumably a fresh installation of BOINC, you will start with a DCF of 1.000000 - so no wonder you got eight times as much work as you really need.

Time for another nudge on those <rsc_fpops_est>, boys? - It's 40 days since the last one.

Yes, we are increasing it gradually.

Hi all, I use this thread ...

This WU was calculated within 48 hours (from 9 Sep 2009 12:48:58 UTC to 10 Sep 2009 14:32:46 UTC) and received no bonus 25%!
What went wrong?

Have a nice weekend :-)

---

CNT-Cimrman is the first team computer operated by a non-profit organisation Czech National Team, o. s..

I have resorted to using the "Allow New tasks" "Do Not allow new tasks" or whatever button. When I get down to one I allow and get 5 more. Loosing bonus points on 2 of 6 WUs. It is the science getting done that is important but our team is coming up thru the ranks fast and I hate to not get the points for the team.