......... Pick another project & set the resource share to 0 in Boinc Manager - that way work will only be requested when projects with a greater than 0 resource share are unavailable.

I have Einstein set to resources share of 0 all the time - so when GPUGrid developed this problem my system started requesting a single Einstein GPU task at a time -a resource share of 0 does not download more than one task at a time.

Make sure you edit your Einstien preferences to only use your nVidia GPU - or you may find CPU tasks appearing ........

That's right -- To use a "Backup project", make sure it's Resource Share is 0. See, that's a special case -- "0 Resource Share" means "Only ask this project for work if all other projects couldn't give me any, and even then, just give me enough work to keep busy right now, don't build a cache of work."

So, I have Einstein/SETI/Albert/SETIBeta all set to 0 resource share as backup projects, and all allowed to download GPU tasks.

I have tons of CPU projects (so I don't have much chance of running out of work on CPU), but I only have GPUGrid/POEM for GPU projects. And so, basically, I end up never asking those backup projects for work, except when GPUGrid has an issue. And when that happens, my setup works flawlessly to keep GPUs busy with backup project GPU tasks, all without user intervention.

Pick another project & set the resource share to 0 in Boinc Manager

OK. I have three GPUGrid uploads @ 100%, and one GPUGrid and one Einstein GPU task running.

I've tried but I cannot find how to set the resource share by project in BOINC Manager. Help!!

"Resource Share" is a project setting.
http://boinc.berkeley.edu/wiki/preferences

It can be set either at:
- the project's website (and then issue an Update command in BOINC), or
- if you use an account manager (like BOINCStats), you can set it there (and then issue an Update command in BOINC).

It can even be set "per venue/location", but I don't use venues/locations to do my settings.

"Resource Share" is a project setting.

Ah! It's not a BOINC Manager setting.

I have GPUGrid @ 100% and Einstein @ 1% (it does not like 0%), so I guess I'm OK for the eventual return to normal of GPUGrid!

...and Einstein @ 1% (it does not like 0%)

Can you please explain?.

You should be able to put 0, click "Update preferences", and then it'll show "---" meaning "it is now set as a backup project".

...and Einstein @ 1% (it does not like 0%)

Can you please explain?.

You should be able to put 0, click "Update preferences", and then it'll show "---" meaning "it is now set as a backup project".

The first time I tried it it would not let me enter a "0". Just tried it again and it did!! And it now shows ---

Yes its definate running with 0%, its working here too. I would prefer to run POEM as backup but they have two Problems.

1.) Not enough workunits all the time for a 100% working backupproject, because they need a backupproject itself :D
2.) does not like dual gpu machines, yes i know exclude blabla, but then point 1. is again for one of the GPUs ^^

So its only Einstein for science reason you can use as backup thats really working as it should. Im sure they are happy about some additional hardcore gpus witch are running on this project :D

---

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

Hmm

I was hoping with the beginning of the work week the disk space issues would be confirmed by a project message and some information regarding a lasting solution shared with people here.

I am still hopeful we'll something along these lines.

I too would revert to POEM GPU's -- but they are exceedingly rare at the moment as they work through application issues.

Instead, I'm running work from Prime Grid or Collatz or Moo. My definitely preference with Nvidia cards is GPU Grid which has been running well over the past several months.

So here's hoping we get some message from the project letting folks know of a projected time frame for resolution.

---

So its only Einstein for science reason you can use as backup thats really working as it should.

I'm not so sure. I'm running my last GPUGrid and one Einstein. I set Einstein to "No CPU" and did an "update". So what's this enormous 500meg download?:



There's much more below...

Seems your getting very much workunits ;) did you set it to 0% and update your client before you accept new work?

---

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

OK -- I see I missed an earlier message from Toni describing the problem -- my apologies.

I am wondering if it is possible to connect an external drive not on the array and to use it to offload files for archive purposes and thus free up space. It is possible that the server hardware doesn't directly support an external USB or ESATA drive though -- in which case, it would require downing the server (never any fun) and installing a card to provide the interface. Just a thought since the message suggests this is going to be an ongoing issue.

Hmm

I was hoping with the beginning of the work week the disk space issues would be confirmed by a project message and some information regarding a lasting solution shared with people here.

I am still hopeful we'll something along these lines.

I too would revert to POEM GPU's -- but they are exceedingly rare at the moment as they work through application issues.

Instead, I'm running work from Prime Grid or Collatz or Moo. My definitely preference with Nvidia cards is GPU Grid which has been running well over the past several months.

So here's hoping we get some message from the project letting folks know of a projected time frame for resolution.

@Jacob: Thanks, Einstein works now perfekt as a backup-project.



But as with Tomba, there are a lot of WU's now. And the GPU load varies between 0 and 50%, CPU-application is running as well, although I have GPU-only applications clicked in the settings. Einstein does not work the way I'm used to from GPUGrid. Hmm..

---

Regards, Josef

Seems your getting very much workunits ;) did you set it to 0% and update your client before you accept new work?

Yep. I set it to 0% and did an update.

Those 50+ files are downloaded and they are NOT CPU WUs. Not sure what they are but I'm currently running my final GPUGrid WU and one Einstein.

@Jacob: Thanks, Einstein works now perfekt as a backup-project.



But as with Tomba, there are a lot of WU's now. And the GPU load varies between 0 and 50%, CPU-application is running as well, although I have GPU-only applications clicked in the settings. Einstein does not work the way I'm used to from GPUGrid. Hmm..

If you accidentally got tons of work units without having the Resource Share set to 0, but now have it correctly set to 0 Resource Share, you can feel free to abort any unstarted work units, without fear of wasting any work. You can abort started ones too, but it's preferable not to do that, since you'd be throwing away some work that has already been done.

Regarding GPU Usage, yeah, if it's an "OpenCL" task, those only use the GPU at certain times. So, you'll see usage fluctuate often, and maybe even be 0 most of the time, but at least you're trying to keep it busy. That's just the nature of their OpenCL app, which is indeed quite different than a GPUGrid CUDA app.

Just keep the GPUs busy, or at least try :)

you can feel free to abort any unstarted work units, without fear of wasting any work.

Done :)

Edit:
Got a much better GPU-utilization with the following app_config.xml in the Einstein-directory:

<app_config>

<app>
	<name>einsteinbinary_BRP4G</name>
	<gpu_versions>
		<gpu_usage>0.5</gpu_usage>
		<cpu_usage>0.2</cpu_usage>
	</gpu_versions>
</app>

<app>
	<name>hsgamma_FGRP3</name>
	<gpu_versions>
		<gpu_usage>0.50</gpu_usage>
		<cpu_usage>0.2</cpu_usage>
	</gpu_versions>
</app>

</app_config>

---

Regards, Josef

Seems to be working.

Edit:
Got a much better GPU-utilization with the following app_config.xml in the Einstein-directory:

<app_config>

<app>
	<name>einsteinbinary_BRP4G</name>
	<gpu_versions>
		<gpu_usage>0.5</gpu_usage>
		<cpu_usage>0.2</cpu_usage>
	</gpu_versions>
</app>



<app>
	<name>hsgamma_FGRP3</name>
	<gpu_versions>
		<gpu_usage>0.50</gpu_usage>
		<cpu_usage>0.2</cpu_usage>
	</gpu_versions>
</app>

</app_config>

Generally for most tasks, E@H GPU utilization for the most part is not as efficient as GPUGRID. If you want a simpler option, other than the above, which is very useful and may help to address "CPU suffocation" issues that E@H GPU tasks may occur on certain computers, if not video cards, by changing the "cpu usage" value, then follow these simple steps:

1) Open up your "Einstein@Home preferences" page;

2) Scroll down to the bottom of the "default/home/work/school" subsection where you see:
"GPU utilization factor of [Name of] apps DANGEROUS! Only touch this if you are absolutely sure of what you are doing! Wrong setting might even damage your computer! Use solely on your own risk! Min: -1.0 / Max: 1.0 / Default: 1.0, negative values will disable GPU tasks of this type"

3) The number "1" is your default setting of 1 GPU task per 1 GPU card. If you want a higher amount, then change the setting as follows:

0.5 means 2 GPU tasks per GPU card;
0.25 means 4 GPU tasks per GPU card, etc..

Please remember to save the changes, and CPU default usage values for E@H remain unaffected using this method.

Hope this helps and good luck!

So its only Einstein for science reason you can use as backup thats really working as it should. Im sure they are happy about some additional hardcore gpus witch are running on this project :D

Or MilkyWay@home, there multiple GPU's work great too.

---

Greetings from TJ

Everything runs fine again. Thanks to all of the problem solving involved people!

---

Regards, Josef

So its only Einstein for science reason you can use as backup thats really working as it should. Im sure they are happy about some additional hardcore gpus witch are running on this project :D

Or MilkyWay@home, there multiple GPU's work great too.

Im really not using nvidia gpus where DP Performance is needed ;) MW has a 7950 24/7 running from me that must be enough until poem has enough ati workunits ^^

but yes everything is running fine now again, great! :)

---

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