Hi;
BOINC Version 6.4.5 is running on my 4-Core Q9550 with a GTX280 graphic-card.
I´m using Vista Ultimade 64 with 4 GB RAM.
In former days 4 WU were running plus 1 WU on the graphic-card (GPUGRID).
The GPUGRID WU is signed with 0,03CPUs, 1 CUDA.
I didn´t remember which change was made that now only 3 normal WU plus the GPUGRID WU are calculated.
Thanks in advance!

If you're referring to an option in the BOINC client, then it is the "on multiprocessor systems, use at most XX % of processors". If you have a Quad-core and don't select 100%, then this will limit you to 3 CPU-bound tasks (i.e., if you select 99%, you will see this 3+1 state). If you change it back to 100%, you should see 4+1.

It's up to you as numerous people have reporting varying levels of impact to GPU processing times when using 3+1 versus 4+1.

I´m on 100% CPU use, but as I discribed..... 8-(

I´m on 100% CPU use, but as I discribed..... 8-(

That might be an issue with BOINC 6.4.5. You might want to try client 6.5.0 to see if it helps. It might not be a bad thing, as dedicating a whole CPU to GPUGrid generally results in slightly better GPUGrid performance (with a small sacrifice in cpu-bound project performance).

I've run into this problem several times and found that if I suspend GPU Grid and then restart it, that it will sometimes go ahead and run 5 tasks. Otherwise, if you shut down the Boinc client and restart, it will usually start up with five clients. This is a little irritating, though since it requires more monitoring than I really want to do.

Running 4+1 with BOINC 6.4.5 is possible by writing a file called "cc_config.xml" with an editor. Type the following content:

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Place the file into the BOINC data directory and click at "Read config file" in the BOINC Manager Menu "Advanced".

The problem is solved by installing and using version 6.6.2

Running 4+1 with BOINC 6.4.5 is possible by writing a file called "cc_config.xml" with an editor. Type the following content:

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Place the file into the BOINC data directory and click at "Read config file" in the BOINC Manager Menu "Advanced".

Thank you for this suggestion. I tried this on one computer and tried the 6.6.2 option on another. Now the machine with 6.6.2 won't stop downloading new work units. It seems I've introduced a new problem:)

Running 4+1 with BOINC 6.4.5 is possible by writing a file called "cc_config.xml" with an editor. Type the following content:

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Place the file into the BOINC data directory and click at "Read config file" in the BOINC Manager Menu "Advanced".

Thank you for this suggestion. I tried this on one computer and tried the 6.6.2 option on another. Now the machine with 6.6.2 won't stop downloading new work units. It seems I've introduced a new problem:)

THis is a known issue with 6.6.2 ... opposite to mine where I could not get any work on the systems using 6.6.2

---

Running 4+1 with BOINC 6.4.5 is possible by writing a file called "cc_config.xml" with an editor. Type the following content:

<cc_config>
<options>
<ncpus>5</ncpus>
</options>
</cc_config>

Place the file into the BOINC data directory and click at "Read config file" in the BOINC Manager Menu "Advanced".

Thank you for this suggestion. I tried this on one computer and tried the 6.6.2 option on another. Now the machine with 6.6.2 won't stop downloading new work units. It seems I've introduced a new problem:)

THis is a known issue with 6.6.2 ... opposite to mine where I could not get any work on the systems using 6.6.2

On the positive side, I don't have any problems with GPUGrid failing to download new work units anymore.

---

Currently running:
2 x Intel Core 2 Quad CPU Q6700 @ 2.66GHz
PS3PF

@Adam
I also had the problem of overfilling the queue with WUs with 6.6.2. So I re-installed 6.4.5.
Yesterday I upgraded to BOINC 6.6.3. Now all works well and I don't need the <ncpus>5</ncpus>-tag any more.

Thanks for the info. I'll give that a shot

@Adam
I also had the problem of overfilling the queue with WUs with 6.6.2. So I re-installed 6.4.5.
Yesterday I upgraded to BOINC 6.6.3. Now all works well and I don't need the <ncpus>5</ncpus>-tag any more.

You do not need that tag with 6.5.0 either. There have been a few problems reported with 6.6.3, 6.6.4, and 6.6.5 ... but, if it works for you thats cool ... if not, you do not have to fall all the way back to 6.4.5 ... 6.5.0 is what I have been running on almost all my systems ...

I have tried all the suggestions here and I am still at the original problem. Boinc runs 4 WCG WUs and 0 GPUGrid or 3 WCGs and 1 GPUGrid. When I tried the <ncpus>5</ncpus> it ran 5 WCG WUs.

WCG is my main project so losing 1 core of 4 is not good. Any other ideas appreciated.

Thanks

Check that:
- In Preferences, the option "processor useage - use at most" (both of them) are set to 100%
- In Preferences, the option "processor useage - use gpu whilst computer in use" is checked in the box.

Close BAM (after disconnecting host in preferences) Remove any references to ncpu or gpu that maybe in the config file, then restart BAM.

Then come back with whats happening, and your BOINC version number, and card type

The rationale behind the above is not knowing what state your various options are in or the BOINC version, by doing the above we are at a "steady state" - whatever that state is ..... then can move forward one change at a time.

The reason for the latter is there are many BOINC versions out there, a lot of them are in fact not meant for Production - they are development versions with their own known problems (each will have different problems). Changes to config at same time or after a version change can introduce unexpected variables in terms of issues.

I empathise with your problem - at one time I was having CPDN tasks pre-empting GPU tasks and couldnt get the beast to behave itself :) I took Paul's advice and went to 6.5.0, cleaned out the config file, then went one step at a time. Got there in the end, no problem since, currently running 6.6.31

Regards
Zy

Thanks for reply.

I did get things sorted out and got all working nicely. It seemed a number of things and not just one.

BUT all 7 WUs had compute errors. So I detached from project.

First details:
Phenom 2 810 OCd 3gz. Win7 64bit. GTX 260 192, 186.18 drivers. Boinc 6.6.36.

Have run OOCT, CUDAMemtest, Fur, several times no problems. Ran F@H for 2 months and as long as I used 182.50 drivers no errors. Any others and errors 20-30% of the time. Have run WCG on this machine since Feb with no problems.

I got tired of F@H problems and failed WUs on the GTX 260 and the 4850. Because of heat, was hitting 95 degrees here, I shut down 4850. Wanted to use new nVidia drivers for gaming but F@H errors out so thought I would try GPUgrid. Really do not want to keep changing drivers to find one that will work. I lost alot of WCG work upgrading Boinc, my fault but I still do not like that.

If there is something else or a driver that is known to be good for the masses will give it a go but kinda getting tired of trying to get GPU computing to work. Have been with WCG since 2004 and it is my focus. This does look like a nice project. May try again later.

Thanks!!!

186.18 is known not to work for some (see here). Pretty much any 181, 182 or 185 series should do, preferralby non-beta. On Thursday a new app is scheduled which should work with 186 series.

MrS

---

Scanning for our furry friends since Jan 2002