It could be the flops setting in the app-info file that affects this

This is the value in the app-info for for the 6950 running duel tasks

<flops>1157115231469.729200</flops>

Any suggestion what it should be?

After 6 days I'm still receiving normal (192s) on the 6950 & 6850.
Could it be something in the app-info file causing this.

It could be the flops setting in the app-info file that affects this for anonymous platform users. However, scheduler will still measure the actual speed and adjust things if the flops provided to it from your BOINC Client/app-info is strange so it could be hard to convince it otherwise even with tweaking that value.

I've witnessed that the normal units are been given excessively to user so it seems scheduler is preferring them and this doesn't seem to change even after stats have been gathered. I do plan to look into fixing this soonish as it would be better to get huge ones out more since that would also help with server/traffic load.

-w

I am getting the same on my HD5850, and have been for days. The normal WUs (192s, etc) take around 10 minutes, whereas the huge units took 38 to 40 minutes.

With no app_info my 5770 has been getting the normal 192 (16 minutes) tasks. It used to get the 768 tasks (59 minutes).

Could the server be confused? It's been a week and the server hasn't adjusted yet.

Rick

With no app_info my 5770 has been getting the normal 192 (16 minutes) tasks. It used to get the 768 tasks (59 minutes).

Could the server be confused? It's been a week and the server hasn't adjusted yet.

Rick

My 5770 machines are doing the same thing. I just returned some units in the 8 minute range, dual 5770's on the machine though.

Could the server be confused? It's been a week and the server hasn't adjusted yet.

It seems to be. I'll try to take a look and hopefully fix it this weekend.

I started a new day job at the start of Feb and that's eating away my free time. :(

-w

Not sure I understand, but this app_info.xml file is for running two WUs on 1 GPU, so you need to have it at .5 versus "1" which will probably cause the crash you're talking about?

---

It seems to be. I'll try to take a look and hopefully fix it this weekend.

I started a new day job at the start of Feb and that's eating away my free time. :(

-w

Thanks Teemu

Received the first "Huge" WUs

I was getting the same crashes.

1)Drivers
2) Power
3) Overclock isn't Stable

Okay I have a question for you app_info specialists...
I want to put 2 mismatched gpu's in one machine, perhaps an Nvidia and an AMD, perhaps 2 AMD's. In either case they will NOT be the same model, my question is...if I use an app_info to say use gpu 0 on Moo and gpu 1 on project B do I need all the stuff telling each project what file to use etc? ORRR can I just make a small file saying exclude gpu1 from this project and exclude gpu0 for that projects. I am trying to consolidate machines and don't want to move 5 gpu's if I don't have to. I DO have extra pci-e(x8 or x16) slots in some machines but not all, hence the question.

I want to put 2 mismatched gpu's in one machine, perhaps an Nvidia and an AMD, perhaps 2 AMD's. In either case they will NOT be the same model

I believe the only way to accomplish this task is to use the cc_config.xml file placed in the main BOINC folder. I have the same issue on one of my machines.

cc_config.xml:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
<exclude_gpu>
<url>http://http:/moowrap.net/</url>
<device_num>0</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>1</device_num>
</exclude_gpu>
<report_results_immediately>1</report_results_immediately>
</options>
</cc_config>

The only caveat is you have to use BOINC 7.x.x or the exclude statements are not recognized.

In the above example, I'm telling BOINC not to use device 0 for moowrap and not to use device 1 for collatz. Hope this helps.

I want to put 2 mismatched gpu's in one machine, perhaps an Nvidia and an AMD, perhaps 2 AMD's. In either case they will NOT be the same model

I believe the only way to accomplish this task is to use the cc_config.xml file placed in the main BOINC folder. I have the same issue on one of my machines.

cc_config.xml:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
<exclude_gpu>
<url>http://http:/moowrap.net/</url>
<device_num>0</device_num>
</exclude_gpu>
<exclude_gpu>
<url>http://boinc.thesonntags.com/collatz/</url>
<device_num>1</device_num>
</exclude_gpu>
<report_results_immediately>1</report_results_immediately>
</options>
</cc_config>

The only caveat is you have to use BOINC 7.x.x or the exclude statements are not recognized.

In the above example, I'm telling BOINC not to use device 0 for moowrap and not to use device 1 for collatz. Hope this helps.

That IS EXACTLY what I want, but I use Boinc version 6! I may have to move those multi gpu machines over to version 7, hmmm. Thank you!!

That IS EXACTLY what I want, but I use Boinc version 6! I may have to move those multi gpu machines over to version 7, hmmm. Thank you!!

No problem. I use BOINC 7.0.18 with the below cc_config file and have no issues with scheduling, receiving or sending completed tasks, and using the 2 different GPUs. I discoverd on one of my dual GPU machines I had to have the use_all_gpus statement to have BOINC use both GPUs. Before I found this, BOINC 7.x.x would recognize both GPUs but only use 1 of them. I don't know if it's 7.0.18 or the reports statement that keeps the cache full, but 7.0.8 didn't (I didn't have the cc_config file then).

You can suspend all tasks and set the projects to no new tasks and let the running WUs finish to upgrade. When you go from 6.x.x to 7.x.x, it will not destroy your current WUs on hand. Going from 7.x.x to 6.x.x is a different story. Good luck.

That IS EXACTLY what I want, but I use Boinc version 6! I may have to move those multi gpu machines over to version 7, hmmm. Thank you!!

No problem. I use BOINC 7.0.18 with the below cc_config file and have no issues with scheduling, receiving or sending completed tasks, and using the 2 different GPUs. I discoverd on one of my dual GPU machines I had to have the use_all_gpus statement to have BOINC use both GPUs. Before I found this, BOINC 7.x.x would recognize both GPUs but only use 1 of them. I don't know if it's 7.0.18 or the reports statement that keeps the cache full, but 7.0.8 didn't (I didn't have the cc_config file then).

You can suspend all tasks and set the projects to no new tasks and let the running WUs finish to upgrade. When you go from 6.x.x to 7.x.x, it will not destroy your current WUs on hand. Going from 7.x.x to 6.x.x is a different story. Good luck.

I have already passed the info to TWO other people so far, thru forum threads so countless others have read it too. THANKS!!!

Hi,
I'm new here (1 day) and read through all the posts but cannot find
an .xml that will allow me to run a task apiece on gpu's.
I have a machine with two 5870 gpu's and Moo! uses both cards for the same task.
I would like to run one task per gpu, i.e. one on each of the cards.
Using both cards for one task seem to take forever on this machine.
Any help will be appreciated.

Hi,
I'm new here (1 day) and read through all the posts but cannot find
an .xml that will allow me to run a task apiece on gpu's.
I have a machine with two 5870 gpu's and Moo! uses both cards for the same task.
I would like to run one task per gpu, i.e. one on each of the cards.
Using both cards for one task seem to take forever on this machine.
Any help will be appreciated.

Message #2799 has the details in it but you MUST upgrade to version 7 of Boinc, which can be VERY problematic for some Boinc projects!

BTW you have been here ONE DAY and have almost 1 MILLION credits, I would NOT be complaining about that, your pc's are SCREAMING fast!!

That IS EXACTLY what I want, but I use Boinc version 6! I may have to move those multi gpu machines over to version 7, hmmm. Thank you!!

No problem. I use BOINC 7.0.18 with the below cc_config file and have no issues with scheduling, receiving or sending completed tasks, and using the 2 different GPUs. I discoverd on one of my dual GPU machines I had to have the use_all_gpus statement to have BOINC use both GPUs. Before I found this, BOINC 7.x.x would recognize both GPUs but only use 1 of them. I don't know if it's 7.0.18 or the reports statement that keeps the cache full, but 7.0.8 didn't (I didn't have the cc_config file then).

You can suspend all tasks and set the projects to no new tasks and let the running WUs finish to upgrade. When you go from 6.x.x to 7.x.x, it will not destroy your current WUs on hand. Going from 7.x.x to 6.x.x is a different story. Good luck.

This no longer works for me!! I can make one gpu run Poem but I can't make the other gpu run Moo! I can also, with no cc_config file make 2 Poem units run at once, but I cannot make a Moo unit run at all!! I am using version 7.0.23. I have reset Moo, dumped a dozen units, got new units etc, etc, ETC!!! Something is wrong somewhere! Here is a copy of my cc_config.xml file:

<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>

<exclude_gpu>
<url>http://moowrap.net/</url>
<device_num>0</device_num>
<app>dnetc</app>
</exclude_gpu>

<exclude_gpu>
<url>http://boinc.fzk.de/poem/</url>
<device_num>1</device_num>
<app>poemcl</app>
</exclude_gpu>

<report_results_immediately>1</report_results_immediately>
</options>
</cc_config>

Anyone have any ideas?

Have you tried Prime/M@H? Changing your cc_ of course.

---

Have you tried Prime/M@H? Changing your cc_ of course.

No I have not, I have plenty of credits at Prime right now but not enough at Poem, I have some goals that I am working on.

Have you tried Prime/M@H? Changing your cc_ of course.

No I have not, I have plenty of credits at Prime right now but not enough at Poem, I have some goals that I am working on.

But I might try that today, maybe it IS the combination of projects not the cc_config file! Maybe it is just Moo that is the problem!!

Have you tried Prime/M@H? Changing your cc_ of course.

No I have not, I have plenty of credits at Prime right now but not enough at Poem, I have some goals that I am working on.

But I might try that today, maybe it IS the combination of projects not the cc_config file! Maybe it is just Moo that is the problem!!

Could it be(not like you have not thought of this your self) your BM? Did you put the right Poem tag?(<url>http://boinc.fzk.de/poem/</url>)
Could you try and switch http tags around?
Please keep me/us informed..Very interesting I think.

---