Thanks. I've validated well over 11 ATM tasks. Oh well, it is what it is. I just hate to see all this GPU horsepower I have doing nothing.

Bonjour ,
j'aimerais savoir s'il est possible d'attribuer une carte graphique a un calcul en particulier car je possede une gtx 1650 4gb mais GPUGRID lui donne toujours du quantum chemistry a calculer.
Avec 4 gb seulement tout part en erreur contrairement a ma rtx 4060.
Je voudrais que ma gtx 1650 calcule pour atm car cela ne pose pas de probleme.
Le calcul réussit a chaque fois.
d'apres ce que je sais gpugrid ne gere pas de fichier appconfig.xml.
Je voudrais que ma gtx 1650 calcule pour ATM et ma rtx 4060 pour Quantum chemistry.
Est ce possible?

Good morning,
I would like to know if it is possible to assign a graphics card to a particular calculation because I have a gtx 1650 4gb but GPUGRID still gives it quantum chemistry to calculate.
With only 4 gb everything goes wrong unlike my rtx 4060.
I would like my gtx 1650 to calculate for atm as this is not a problem.
The calculation succeeds every time.
from what I know gpugrid does not handle appconfig.xml file.
I would like my gtx 1650 to calculate for ATM and my rtx 4060 for Quantum chemistry.
Is that possible?

---

Keith Myers wrote:

Once you've validated 11 ATM tasks, the DCF for that application should come down and the estimated times to completion should come down to something reasonable.

none of my 4 hosts (with total of 7 GPUs) is showing this behaviour, though.

est ce que gpugrid arretera d'envoyer des unites quantum chemistry sur ma gtx 1650 si elles partent toutes en erreur a cause de la quantité de ram et mettra automatiquement ma gtx 1650 sur ATM qui devrait fonctionner coorectement?


Will gpugrid stop sending quantum chemistry units on my gtx 1650 if they all go wrong because of the amount of ram and will automatically put my gtx 1650 on ATM which should work coorectly?

---

est ce que gpugrid arretera d'envoyer des unites quantum chemistry sur ma gtx 1650 si elles partent toutes en erreur a cause de la quantité de ram et mettra automatiquement ma gtx 1650 sur ATM qui devrait fonctionner coorectement?


Will gpugrid stop sending quantum chemistry units on my gtx 1650 if they all go wrong because of the amount of ram and will automatically put my gtx 1650 on ATM which should work coorectly?

go into your project preferences and disable running test applications and uncheck the Quantum Chemistry project.

---

Good morning,
I would like to know if it is possible to assign a graphics card to a particular calculation because I have a gtx 1650 4gb but GPUGRID still gives it quantum chemistry to calculate.
With only 4 gb everything goes wrong unlike my rtx 4060.
I would like my gtx 1650 to calculate for atm as this is not a problem.
The calculation succeeds every time.
from what I know gpugrid does not handle appconfig.xml file.
I would like my gtx 1650 to calculate for ATM and my rtx 4060 for Quantum chemistry.
Is that possible?

Yes, use an exclude_gpu statement in your cc_config.xml file.
https://boinc.berkeley.edu/wiki/Client_configuration#Application_configuration


Something like this in the Options section.

<exclude_gpu>
<url>http://www.gpugrid.net/</url>
<device_num>1</device_num>
<type>NVIDIA</type>
<app>PYSCFbeta</app>
</exclude_gpu>

The 1650 should be enumerated as gpu#1 in Boinc in relation to the 4060. But check first in the Event Log which card is #0 and #1 in Boinc's thinking.

merci mais cela ne fonctionne pas avec gpugrid car ce dernier ne gere pas de fichier appconfig.xml.
je viens d'essayer mais c'est sans effet.

thank you but this does not work with gpugrid because it does not manage appconfig.xml file.
I just tried, but it doesn’t work.

---

merci mais cela ne fonctionne pas avec gpugrid car ce dernier ne gere pas de fichier appconfig.xml.
je viens d'essayer mais c'est sans effet.

thank you but this does not work with gpugrid because it does not manage appconfig.xml file.
I just tried, but it doesn’t work.

you refer to appconfig.xml. However, the addition suggested by Keith needs to be done in the cc_config.xml (located in the main BOINC folder). Further, after making all this in the cc_config.xml, stop and restart the BOINC manager. Only then the changes will work.

You asked for a solution for excluding your 1650 from Quantum Chemistry tasks but still able to run both ATM and QC tasks on your 4060.

That is the solution I presented. The exclude statement goes into your cc_config.xml file, NOT your app_config.xml file.

You don't even need a app_config.xml file for this project and can control everything just by your Compute Preferences on the project via your browser.

Look through the client configuration doc I linked for you.

merci j'ai bien mis dans cc config.xml et je vais attendre des taches atm pour ma gtx 1650.
vous etes sympathique et merci encore.

thank you I put well in cc config.xml and I will wait for atm spots for my gtx 1650.
you are friendly and thank you again.

---

Thanks. I've validated well over 11 ATM tasks. Oh well, it is what it is. I just hate to see all this GPU horsepower I have doing nothing.

You can try to reset your DCF to a low value, which may fix the issue at least temporarily. If you're mixing pyscf & ATM WU's on your machine, this will not work for long but with ATM-only it might have lasting effect.

- Stop BOINC client
- edit client_state.xml (in your main BOINC folder)
- look for the <project> section of gpugrid
- look inside this <project> section for <duration_correction_factor>
- set the value to 0.01 (lower than that is ignored by BOINC and will be updated to 0.01 after the first WU)

<duration_correction_factor>0.010000</duration_correction_factor>

- save and close
- Start BOINC client

je vous conseille de passer sous linux .
il y a toujours du travail.c'est moins évident a mettre en ouevre que windows mais vous aurez du quantum chemistry a calculer en permanence.
Perso j'envisage de racheter des gpu rtx a 2000 ou rtx 4000 sff expres pour gpugrid.
Gpugrid étant le seul projet boinc sur gpu qui vaille la peine de se donner du mal.
Les projets sur les chiffres ou l'espace,pour moi,c'est bof bof.

I advise you to switch to linux .
there is always work. it is less obvious to implement than windows but you will have quantum chemistry to calculate constantly.
Personally I plan to buy GPUs rtx a 2000 or rtx 4000 sff expres for gpugrid.
Gpugrid being the only boinc project on gpu that is worth the trouble.
Projects on numbers or space, for me, is bof bof.

---

Thanks...all kind of over my head. I've been running boinc stuff since 2003 and have never had to edit something. I'm just an old guy now.

after the ATMs have been running well within the past few weeks, almost all ATMs that my hosts received this afternoon failed after less than a minute with error

13:14:21 (5332): app exit status: 0xc0000135
13:14:21 (5332): called boinc_finish(195)

why all of a sudden?

P.S. they failed on other volunteers' hosts, too. So the problem is not with my hosts.

I have two windows computers running the ATM tasks and I have not even received a task to run since 3/22/24

If you look at the server status page https://www.gpugrid.net/server_status.php
you can see that there are only Quantum Chemistry tasks in large supply.
The ATM tasks have been very spotty for the last month.

je vous conseille de passer sous Linux .
il ya toujours du travail.c'est moins évident a mettre en œuvre que windows mais vous aurez du quantum chemistry a calculer en permanence.
Perso j'envisage de racheter des gpu rtx a 2000 ou rtx 4000 sff express pour gpugrid.
Gpugrid étant le seul projet boinc sur GPU qui vaille la peine de se donner du mal.
Les projets sur les chiffres ou l'espace,pour moi,c'est bof bof.

Je vous conseille de passer à Linux.
il y a toujours du travail. c'est moins évident à mettre en oeuvre que windows mais vous aurez de la chimie quantique à calculer en permanence.
Personnellement, je prévois d'acheter des GPU rtx a 2000 ou rtx 4000 sff express pour gpugrid.
Gpugrid étant le seul projet boinc sur GPU qui en vaut la peine.
Les projets sur les chiffres ou l'espace, pour moi, c'est bof bof.

I have two windows computers running the ATM tasks and I have not even received a task to run since 3/22/24

Some ATM batches have been launched in late April 🎉🙂

what is this? Seems new:

https://www.gpugrid.net/result.php?resultid=35068097

stderr: exceeded elapsed time limit 17162.78 (250000000.00G/14180.09G)</message>

any explanation?

The work unit generator has an incorrect value for estimated time to complete in the task profile.