I would like to set my system up to run certain apps on GPU and others on CPU. Right now I am running the same apps on both. I want to run the Separation apps on CPU only and keep doing the Milky Way apps on GPU only. Is there any way to do that?

Do you mean you want to split the 4 MW@H apps,
MilkyWay@Home
MilkyWay@Home N-Body Simulation
Milkyway@Home Separation
Milkyway@Home Separation (Modified Fit)



so, these two run on the CPU only,
Milkyway@Home Separation
Milkyway@Home Separation (Modified Fit)


and only this one runs on the GPU
MilkyWay@Home

Correct. I don't do the N-Body because they always screw up in some way. Invalid (bay have a bug) or validation inconclusive. Something always seems to go wrong with them.

This is my current app config file-
<name>milkyway_nbody</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>milkyway</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>milkyway_separation__modified_fit</name>
<gpu_versions>
<gpu_usage>.5</gpu_usage>
<cpu_usage>.5</cpu_usage>
</gpu_versions>
</app>
<app>
<name>milkyway_nbody</name>
<max_concurrent>1</max_concurrent>
</app>
<app_version>
<app_name>milkyway_nbody</app_name>
<plan_class>mt</plan_class>
<avg_ncpus>1</avg_ncpus>
<cmdline>--nthreads 1</cmdline>
</app_version>
</app_config>

I could actually pull the N-Body stuff out since I won;t do them anymore.

I suspect that you are going down the right path with the XML configuration ... if a solution exists or not.

Do you see much difference in the time to complete 2 jobs in parallel versus 2 sequentially? GPU-Z indicates that my GTX970 is pretty loaded down with MH@H. Some other projects I have set to 0.5 GPU, but not MW.

I have found my card will run the most work at 2 tasks at a time. It takes 240 seconds to run 2 tasks at once, so 120 per task. If I do one at a time it takes 135 seconds. If I go up the 3 tasks it drops the output to even less than one at a time, so 2 is the sweet spot.

Sorry for the necro-thread, but I can't seem to find any other close references using site search or other docs that discuss project-specific app_config issues.
I'm trying to adjust my app_config.xml to control the max number of concurrent instances while trying to differentiate between CPU and GPU work.

I suspect the issue is that I need to define the plan_class for CPU as well as for GPU, and while I know from client_state.xml that my GPU work plan class is opencl_nvidia_101, I can't seem to locate a plan_class defined for CPU work, and without it the parameters seem to be ignored. (That assumes this is the right way to address the issue.)
What I'm trying to accomplish is something like this:

<app_config>
  <app>
    <name>milkyway</name>
    <gpu_versions>
      <gpu_usage>.5</gpu_usage>
      <cpu_usage>.05</cpu_usage>
    </gpu_versions>
  </app>

  <app_version>
    <app_name>milkyway</app_name>
    <plan_class> ???? </plan_class>            <!-- what is the p_c for CPU? -->
    <max_concurrent>1</max_concurrent>
  </app_version>

  <app_version>
    <app_name>milkyway</app_name>
    <plan_class>opencl_nvidia_101</plan_class>
    <max_concurrent>2</max_concurrent>
  </app_version>

</app_config>

Would appreciate it if anybody could get my mind right on this, or point me in the right direction!!
Thx, Jim ...

---

Yeah, the above pretty much isn't going to happen, at least this way. While max_concurrent works within <app></app>, apparently if you try to nest it deeper, it's just ignored.
Now, if the GPU and CPU versions of "milkyway" had different app names, as I've seen at other projects, this is a piece of cake.
Seems like using the same app name for two different versions of an app is a pretty bad idea.

---

Yeah, the above pretty much isn't going to happen, at least this way. While max_concurrent works within <app></app>, apparently if you try to nest it deeper, it's just ignored.
Now, if the GPU and CPU versions of "milkyway" had different app names, as I've seen at other projects, this is a piece of cake.
Seems like using the same app name for two different versions of an app is a pretty bad idea.

You've got two different pc's just put them in different venues and on one run cpu wu's and on the other one run gpu wu's. Gpu and cpu wu's take vastly different times to run and your cache will never be right especially with the 10 minute timeout to get gpu wu's here at MilkyWay. Unfortunately MilkyWay is pretty happy with the way they do things so any changes, ie names of the apps or Server side timeout settings, aren't going to happen anytime soon.

You've got two different pc's just put them in different venues and on one run cpu wu's and on the other one run gpu wu's. Gpu and cpu wu's take vastly different times to run and your cache will never be right especially with the 10 minute timeout to get gpu wu's here at MilkyWay. Unfortunately MilkyWay is pretty happy with the way they do things so any changes, ie names of the apps or Server side timeout settings, aren't going to happen anytime soon.

Might want to rethink your suggestion there. As stated, I'm not sure what that would accomplish, except to leave CPUs idle on one machine and GPUs idle on the other.
As you noted, nothing new about the unwillingness in academia to see and adjust to reality. Don't get me started.:)
Some of this is really a BOINC issue, as well. It may have taken a PhD to write BOINC, but shouldn't require one to operate it.
I was hoping that somewhere on this site there was a listing of apps with app names and the like that I could use as a reference.
Bottom line: Giving two different apps the same name is just plain silly, even if they are manipulating the same data task.

---

You've got two different pc's just put them in different venues and on one run cpu wu's and on the other one run gpu wu's. Gpu and cpu wu's take vastly different times to run and your cache will never be right especially with the 10 minute timeout to get gpu wu's here at MilkyWay. Unfortunately MilkyWay is pretty happy with the way they do things so any changes, ie names of the apps or Server side timeout settings, aren't going to happen anytime soon.

Might want to rethink your suggestion there. As stated, I'm not sure what that would accomplish, except to leave CPUs idle on one machine and GPUs idle on the other.
As you noted, nothing new about the unwillingness in academia to see and adjust to reality. Don't get me started.:)
Some of this is really a BOINC issue, as well. It may have taken a PhD to write BOINC, but shouldn't require one to operate it.
I was hoping that somewhere on this site there was a listing of apps with app names and the like that I could use as a reference.
Bottom line: Giving two different apps the same name is just plain silly, even if they are manipulating the same data task.

There are more than a dozen other Boinc Projects still running that would LOVE to have your help, so it's not necessary for them to be idle!! I have 15 pc's running and NONE of them are running the same Project on the cpu's and gpu's in the same machine. In the past I have had the cpu's on one machine on Project A and the gpu's in another machine also on Project A but I no longer do that because of the inefficiency of the cpu when comparing them against each other. ie a gpu can do 10 times the work is less time than a cpu can. So now I run my cpu's on Projects that do not have gpu apps and everything runs fairly smoothly.

As for the Apps look above under Computing and the you will see Applications. MW is NOT the easiest Project to get help from, yes there are some very good Users here but the Admins seem to be limited in either their ability or their allowability to make changes. In short THEY don't see a problem so there's no reason to fix anything, I'm not complaining just stating what I see.

The problem comes when you need the actual binary app name to write configuration files. The friendly name does not suffice and having the friendly name for two different binaries is unacceptable.
Any help here is going to come from the volunteers as the administrators (in general, some have been helpful but no longer around) over the past have proved incapable of simple server configuration changes. Or ascribe to simple indifference to the volunteers.

---

In short THEY don't see a problem so there's no reason to fix anything, I'm not complaining just stating what I see.

Yeah, I looked at the list of apps (Thanks for that, Mikey) and Good Grief, the version and names and numbers are all the same regardless of platform or what ever.
To the degree that projects compete for the resources of volunteers it's right that one can easily move on to another, but that badly misses the point.
Whatever ...

---

In short THEY don't see a problem so there's no reason to fix anything, I'm not complaining just stating what I see.

Yeah, I looked at the list of apps (Thanks for that, Mikey) and Good Grief, the version and names and numbers are all the same regardless of platform or what ever.
To the degree that projects compete for the resources of volunteers it's right that one can easily move on to another, but that badly misses the point.
Whatever ...

Sometimes you just have to pick your battles MW is NOT going to change the way they do things until they have some SERIOUS turnover, so you either learn to accept it as it is and make it work for you or you walk away and give your resources to a more appreciative place that will at least listen and explain why it's the way it is. They just don't seem to care when people leave!!

hi all made vid on youtube for multiple instances instruction's and at full load on a Radeon VII
RADEON VII GIGABYTE// 3 Instances_ Milkyway@home WUs BOINC_ 3_instances
https://www.youtube.com/watch?v=4xKy9wGKmz4
all the best and welcome to earth

---