Greetings. I still don't understand how to set up app_config.xml for the GPU.
What I set and what I get don't make sense.

Are these "suggestions" rather than fixed settings?

I have one Radeon HD 7750 that is crunching and running my video.
It has 2 GB of ram in it.
My CPU has 8 cores and I run it at 87.5% - 7 of the 8 - just to let other programs run.

Here is the cc_config.xml

<cc_config>
    <log_flags>
        <file_xfer>1</file_xfer>
        <sched_ops>1</sched_ops>
        <task>1</task>
    </log_flags>
    <options>
        <ncpus>-1</ncpus>
        <save_stats_days>14</save_stats_days>
        <use_all_gpus>1</use_all_gpus>
    </options>
</cc_config>

Here is the app_config.xml

<app_config>
  <app>
  <name>hcc1</name>
  <max_concurrent>2</max_concurrent>`
    <gpu_versions>
      <gpu_usage>0.5</gpu_usage>
      <cpu_usage>0.5</cpu_usage>
    </gpu_version>
  </app>
  <app>
    <name>milkyway</name>
    <max_concurrent>1</max_concurrent>
    <gpu_versions>
    <gpu_usage>1.0</gpu_usage>
    <cpu_usage>1.0</cpu_usage>
    </gpu_versions>
  </app>
</app_config>

When the WCG GPU task runs, the BOINC task page shows:

World Community Grid ---- Running (1 CPUs + 1 ATI GPU ) (( and the normal, elapsed times and Application and Name))

When the MilkyWay GPU task runs, the BOINC task page shows:

Milkyway@Home  ---------- Running (0.05 CPUs +1 ATI GPU)   (( and the normal, elapsed times and Application and Name))

ONLY one runs at a time.

Question 1) On the Milkyway status, where did the 0.05 CPU come from?

Question 2) I'm trying to get the card to do more than one wu at a time on the WCG settings - but BOINC runs only one and says there is only one.

Question 3) Does the CPUtask (that is GPU related) task feed only tasks of ONE project? Or if I put one CPU for each project, will BOINC end up using 2 cpu cores to feed the one GPU card?

Question 4) What is the scope of the "max_concurrent" setting?
Does it refer to total WU an All GPU cards?
Or does it mean a partial sum of WU running the defined task -
In my case 1 Milkyway WU and 2 WCG WU on the one GPU card?

Every time I read the Wiki of a post about this, I get confused at a higher level. :-)

I hope this can also save someone else some headaches.

THANKS in advance!!!
Jay

Greetings. I still don't understand how to set up app_config.xml for the GPU.
What I set and what I get don't make sense.

Are these "suggestions" rather than fixed settings?

I have one Radeon HD 7750 that is crunching and running my video.
It has 2 GB of ram in it.
My CPU has 8 cores and I run it at 87.5% - 7 of the 8 - just to let other programs run.

Here is the cc_config.xml

<cc_config>
    <log_flags>
        <file_xfer>1</file_xfer>
        <sched_ops>1</sched_ops>
        <task>1</task>
    </log_flags>
    <options>
        <ncpus>-1</ncpus>
        <save_stats_days>14</save_stats_days>
        <use_all_gpus>1</use_all_gpus>
    </options>
</cc_config>

Here is the app_config.xml

<app_config>
  <app>
  <name>hcc1</name>
  <max_concurrent>2</max_concurrent>`
    <gpu_versions>
      <gpu_usage>0.5</gpu_usage>
      <cpu_usage>0.5</cpu_usage>
    </gpu_version>
  </app>
  <app>
    <name>milkyway</name>
    <max_concurrent>1</max_concurrent>
    <gpu_versions>
    <gpu_usage>1.0</gpu_usage>
    <cpu_usage>1.0</cpu_usage>
    </gpu_versions>
  </app>
</app_config>

When the WCG GPU task runs, the BOINC task page shows:

World Community Grid ---- Running (1 CPUs + 1 ATI GPU ) (( and the normal, elapsed times and Application and Name))

When the MilkyWay GPU task runs, the BOINC task page shows:

Milkyway@Home  ---------- Running (0.05 CPUs +1 ATI GPU)   (( and the normal, elapsed times and Application and Name))

ONLY one runs at a time.

Question 1) On the Milkyway status, where did the 0.05 CPU come from?

Question 2) I'm trying to get the card to do more than one wu at a time on the WCG settings - but BOINC runs only one and says there is only one.

Question 3) Does the CPUtask (that is GPU related) task feed only tasks of ONE project? Or if I put one CPU for each project, will BOINC end up using 2 cpu cores to feed the one GPU card?

Question 4) What is the scope of the "max_concurrent" setting?
Does it refer to total WU an All GPU cards?
Or does it mean a partial sum of WU running the defined task -
In my case 1 Milkyway WU and 2 WCG WU on the one GPU card?

Every time I read the Wiki of a post about this, I get confused at a higher level. :-)

I hope this can also save someone else some headaches.

THANKS in advance!!!
Jay

Question 2) I'm trying to get the card to do more than one wu at a time on the WCG settings - but BOINC runs only one and says there is only one.

Change the cc_config.xml file to ONLY have the lines:
<cc_config>
<options>
<use_all_gpus>1</use_all_gpus>
</options>
</cc_config>

You forgot the <options> line in the beginning for a major thing.
The rest I don't know what it does but the lines above will get your pc to see both gpu's. I am Windows guy and the other lines could have to do with Linux, I just don't know.

Question 1) On the Milkyway status, where did the 0.05 CPU come from?

This is an editable line, usually it says 1.0 instead of 0.5, so the pc can use as much of the cpu as it needs.

Question 3) Does the CPUtask (that is GPU related) task feed only tasks of ONE project? Or if I put one CPU for each project, will BOINC end up using 2 cpu cores to feed the one GPU card?

If you also change the setting in the app_config.xml file to 1.0 it will only use more than 1 cpu per gpu if it needs to. Normally it is 1 cpu per gpu. Although on some projects, poem for example, it can use more thn 1 cpu per gpu.

Question 4) What is the scope of the "max_concurrent" setting?
Does it refer to total WU an All GPU cards?
Or does it mean a partial sum of WU running the defined task -
In my case 1 Milkyway WU and 2 WCG WU on the one GPU card?

I am only guessing but it seems like it would be total, not per project. You CANNOT run more than one PROJECT at the same time on a a single cpu core or a gpu. Boinc is NOT that kind of a multi-tasker. Boinc WILL swap units in and out of the cpu or gpu per your settings, but it is best to try and set this so it can finish a task prior to swapping to the next project, OR swap just after starting a new unit. Boinc treats the gpu like one big cpu core.

Hi Mikey!!

Thanks for the reply!!
a) Had already replaced cc_config.xml with much shorter version that had ncpus defined.
Replaced file with exact copy of yours.
Will wait for current WU in WCG to finish and will restart, reboot,
turn it upside-down and shake it over my head.

b) The 0.05 value caught my eye since I had set the app-config.xml value to 1.0
and, yet, Boinc reports it as 0.05 in the expert-view task list.

c) Thanks for the explanation of running multiple projects concurrently on a GPU.
That makes perfect sense since - considering the complexity of loading and unloading data to/from the GPU.

Other forum-posts have led me to believe that a single GPU card could hold or process more than one gpu WU at a time. The posts mentioned that, at least, the card could be crunching on one WU while uploading/downloading another WU.

Other forum posts talk about setting up multiple GPU cards, each one fed by a (possible different) BOINC project.

I have a niggling thought to draw up some pictures and submit them to the boinc-wiki for clarification on the subject.

I'm trying small tests to see what works with different projects.
When I get stumped (often), I ask in the forums.

Thanks again!!
Jay

--Edit: PS
Try Linux some time.
a dual boot gives an easy way to try it -
or a live version that will boot and run from a usb-stick.
If I get frustrated, I go watch you-tubes of Gallagher from 1981
Enjoy!

Hi Mikey!!

Thanks for the reply!!
a) Had already replaced cc_config.xml with much shorter version that had ncpus defined.
Replaced file with exact copy of yours.
Will wait for current WU in WCG to finish and will restart, reboot,
turn it upside-down and shake it over my head.

b) The 0.05 value caught my eye since I had set the app-config.xml value to 1.0
and, yet, Boinc reports it as 0.05 in the expert-view task list.

c) Thanks for the explanation of running multiple projects concurrently on a GPU.
That makes perfect sense since - considering the complexity of loading and unloading data to/from the GPU.

Other forum-posts have led me to believe that a single GPU card could hold or process more than one gpu WU at a time. The posts mentioned that, at least, the card could be crunching on one WU while uploading/downloading another WU.

Other forum posts talk about setting up multiple GPU cards, each one fed by a (possible different) BOINC project.

I have a niggling thought to draw up some pictures and submit them to the boinc-wiki for clarification on the subject.

I'm trying small tests to see what works with different projects.
When I get stumped (often), I ask in the forums.

Thanks again!!
Jay

--Edit: PS
Try Linux some time.
a dual boot gives an easy way to try it -
or a live version that will boot and run from a usb-stick.
If I get frustrated, I go watch you-tubes of Gallagher from 1981
Enjoy!

Yes you can SOMETIMES run more then one unit at a time on a gpu, it all depends on how much memory each unit requires. IF the unit requires less then half of what the gpu has built in, no add-in memory, like laptops try to do, allowed! I have, for instance, an AMD 5770 gpu with 1gb of onboard memory, it was doing a unit in about 1.5 hours, at another project. I removed the 5770 and put in an AMD 7970 with 3gb of ram and now units finish in 10 minutes. Each unit takes ALL the available memory, but since the new card has more memory the unit crunches faster. In that case trying to run more then one unit at a time would just slow things down. I had an AMD 5870 here, it would start out running two units at once but as the units progressed it would back off and only one unit at a time would run. Each unit was taking more then half of the 1gb of onboard memory! An easy program to use is gpu-z and you can get it here:
http://downloads.guru3d.com/CPU-Z-1.6.1-download-2868.html
That site has ALOT of helpful stuff to make crunching better.

The 0.5 is what the pc is actually using, the 1.0 just tells it 'you can use all you want up to 1 full cpu core, but NO MORE THEN THAT'. I am not sure the setting allows for more then 1.0 though. Poem for instance has their software setup so it can use more then one full cpu core to run SOME of its gpu units if you have them available.

I HAVE used Linux in the past but since it is so rarely I just haven't taken the time to learn it. I have a corporate license for Windows 7 so setting up new machines is not an issue for me. Long story about where I used to work, them switching to Apples, MS not giving refunds, etc, etc.

My warehouses, having burned down,
I have a clear vision of the moon.

Mikey,
Thanks again!
After days of trials and false starts,
your explanation gives a clear vision of the moon. (and BOINC)
Jay

My warehouses, having burned down,
I have a clear vision of the moon.

Mikey,
Thanks again!
After days of trials and false starts,
your explanation gives a clear vision of the moon. (and BOINC)
Jay

No problem, I am glad it helped.
Keep on crunching!