Would it be possible to only give out Nbody for CPUs and not seperation? Seems daft to do seperation on CPUs when the GPUs are so much faster. There's not currently a way to choose Nbody only if you have a CPU and GPU in the same system.

Perhaps something like the Einstein settings, so Seperation for GPU and CPU are seperate options.

In https://milkyway.cs.rpi.edu/milkyway/prefs.php?subset=project (Project>Preference) you can:
- set if you want to compute on CPU, GPU ATI (AMD) and/or GPU NVIDIA.
- choose what app you want crunch for (Milkyway@home N-Body Simulation or Milkyway@home Separation).

However, and unlike Einstein settings, both setting are currently coarse settings. So you can't choose to crunch N-Body on both CPU and GPU and crunch Separation on GPU only. (Or I didn't find the relevant setting)

That why I disabled CPU for Milkyway.

By the way, I just had a job that was canceled by the server, here it is https://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=877811669
then I looked at this task why it was canceled and there were two executions from the video card and one from the processor which was a little less than 2 minutes earlier and I thought it could happen that I would have completed this task earlier than the processor and those resources processors would simply fly away into the empty?

---

By the way, I just had a job that was canceled by the server, here it is https://milkyway.cs.rpi.edu/milkyway/workunit.php?wuid=877811669
then I looked at this task why it was canceled and there were two executions from the video card and one from the processor which was a little less than 2 minutes earlier and I thought it could happen that I would have completed this task earlier than the processor and those resources processors would simply fly away into the empty?

If you look at this 30 Apr 2023, 8:34:10 UTC and this 30 Apr 2023, 8:35:54 UTC you can see that your task was cancelled mere seconds after the 3rd result was returned and confirmed to be valid, therefore you task was no longer needed so it was cancelled. Now IF you had returned the task prior to the 3rd pc then their task would have been cancelled and yours would have been the one getting the credits. In the long run it's better to have a task cancelled by the Server than for us to crunch it and get zero credits for it because they didn't need it. Your task was probably still 'in the queue' of tasks on your pc waiting to be crunched when it was cancelled.

Would it be possible to only give out Nbody for CPUs and not seperation?

Maybe...

In project preferences:
1. Enable "Use CPU" and "Use ATI GPU" or "Use NVIDIA GPU" (or both types of GPUs if you wish).
2. In "Run only the selected applications" set "Milkyway@home N-Body Simulation: yes" and "Milkyway@home Separation: no"
3. Set "If no work for selected applications is available, accept work from other applications?" to "yes".
4. See what the server sends you.

I think you still will get some separation for CPUs, but it should be less than with both applications set to "yes". Unfortunately Milkyway does not offer the setting "Run CPU versions of applications for which GPU versions are available", it seems to have been disabled when they configured their server, that would be the 100% solution.

---

I thought it could happen that I would have completed this task earlier than the processor and those resources processors would simply fly away into the empty?

Well, yes, it's the fault of the owner of that PC, who has simply too large cache. Turnaround time for CPU apps is close to 10 days, they are basically driving it at the "won't finish in time" limit.

Now IF you had returned the task prior to the 3rd pc then their task would have been cancelled and yours would have been the one getting the credits.

Nope, the task was running already, otherwise it would have been canceled by the client and send back to the server as "Not started by deadline - canceled". That computer has few such tasks in the list.

---

Would it be possible to only give out Nbody for CPUs and not seperation?

Maybe...

In project preferences:
1. Enable "Use CPU" and "Use ATI GPU" or "Use NVIDIA GPU" (or both types of GPUs if you wish).
2. In "Run only the selected applications" set "Milkyway@home N-Body Simulation: yes" and "Milkyway@home Separation: no"
3. Set "If no work for selected applications is available, accept work from other applications?" to "yes".
4. See what the server sends you.

I think you still will get some separation for CPUs, but it should be less than with both applications set to "yes". Unfortunately Milkyway does not offer the setting "Run CPU versions of applications for which GPU versions are available", it seems to have been disabled when they configured their server, that would be the 100% solution.

Tried it before, I get just as many. I assume the Nbodys are sometimes not quite ready to send so I get seperations. Either that or the computer asks for work for GPU and CPU at the same time, the server decides it can't forfill my nbody preference because of the GPU, and sends separation, for both.

So sometimes I stick the CPUs on another project, sometimes I cancel CPU seperations when I see them, so most of the time it does Nbodys. Since I have double precision AMDs, the speed difference between those and CPUs is phenominal, and therefore pointless to run them on my CPUs.

I found another double precision project, WCG. Although their tasks are not very common. They're sending them out now, but slowly, I think they tried full speed the other day and the server started dozing off again.

Would it be possible to only give out Nbody for CPUs and not seperation?

Maybe...

In project preferences:
1. Enable "Use CPU" and "Use ATI GPU" or "Use NVIDIA GPU" (or both types of GPUs if you wish).
2. In "Run only the selected applications" set "Milkyway@home N-Body Simulation: yes" and "Milkyway@home Separation: no"
3. Set "If no work for selected applications is available, accept work from other applications?" to "yes".
4. See what the server sends you.

I think you still will get some separation for CPUs, but it should be less than with both applications set to "yes". Unfortunately Milkyway does not offer the setting "Run CPU versions of applications for which GPU versions are available", it seems to have been disabled when they configured their server, that would be the 100% solution.

Tried it before, I get just as many. I assume the Nbodys are sometimes not quite ready to send so I get seperations. Either that or the computer asks for work for GPU and CPU at the same time, the server decides it can't forfill my nbody preference because of the GPU, and sends separation, for both.

So sometimes I stick the CPUs on another project, sometimes I cancel CPU seperations when I see them, so most of the time it does Nbodys. Since I have double precision AMDs, the speed difference between those and CPUs is phenominal, and therefore pointless to run them on my CPUs.

I found another double precision project, WCG. Although their tasks are not very common. They're sending them out now, but slowly, I think they tried full speed the other day and the server started dozing off again.

Why not just uncheck the box for cpu tasks and just always run the Separation tasks on the gpu and don't do the N-Body tasks at all? There are plenty of other cpu projects out there and as you said WCG is now sending some out again as well. If you have much more than a zero cache size it will work better as well when you run two different Projects like that, one on the cpu and a different one on the gpu.

Why not just uncheck the box for cpu tasks and just always run the Separation tasks on the gpu and don't do the N-Body tasks at all? There are plenty of other cpu projects out there and as you said WCG is now sending some out again as well. If you have much more than a zero cache size it will work better as well when you run two different Projects like that, one on the cpu and a different one on the gpu.

Because I like Nbodys.

Because I like Nbodys.

Well, you can run one BOINC client for GPU work and one for CPU work.

---

Would it be possible to only give out Nbody for CPUs and not seperation? Seems daft to do seperation on CPUs when the GPUs are so much faster. There's not currently a way to choose Nbody only if you have a CPU and GPU in the same system.

Perhaps something like the Einstein settings, so Seperation for GPU and CPU are seperate options.

I think you got your wish Mr P Hucker!!

I think you got your wish Mr P Hucker!!

Just be careful how you pronounce that! Yes that is my real name, and no my mother didn't think of that when she named me. I guess that makes her a mother P Hucker.

---

The above was double spaced between sentences, I apologise for the forum software ruining my post.