So, I've recently installed BOINC on a new system I built for my wife, system has plenty more CPU power than she'll ever need. (Computer link: http://milkyway.cs.rpi.edu/milkyway/show_host_detail.php?hostid=207756)

This machine is just running on the IGP, no Dedicated GPU on it. As such, it can only crunch on the CPU and it takes like 10 hours to complete a work unit.

I noticed that both the GPU and CPU crunch the exact same work units but of course the GPUs are able to complete them much faster.

My question is, are there any WUs that are strictly for CPUs? For example Folding@Home has some WUs that can only be crunched on CPU because they are not well suited to the vector-processing that GPUs specialise in.

If there are no CPU-Only work units, I think my contribution of compute power might be better utilized with another project that is more suited to CPUs.

Currently the Nbody simulation only runs on the CPU, so it's not exactly true that the CPUs and GPUs work the same workunits. The other application does the same thing on the CPU and GPU though.

Try the optimized 64bit cpu app, I crunched cpu tasks on my p4 3.6 long ago and I think it took less than 6 hrs to finish one.

http://www.arkayn.us/milkyway/

How do you get it to run both optimized CPU and GPU apps on the same machine?

(It would be a lot easier if the default app were SSEx-aware.)

Here's a sample of my app_info. Download a cpu opt app then replace the nbody things.

<app_info>
<app>
<name>milkyway</name>
</app>
<file_info>
<name>astronomy_0.23_ATI_x64.exe</name>
<executable/>
</file_info>
<file_info>
<name>brook64.dll</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway</app_name>
<version_num>23</version_num>
<plan_class>ati13ati</plan_class>
<avg_ncpus>0.05</avg_ncpus>
<max_ncpus>1</max_ncpus>
<flops>1.0e11</flops>
<coproc>
<type>ATI</type>
<count>0.5</count>
</coproc>
<cmdline>b-1</cmdline>
<file_ref>
<file_name>astronomy_0.23_ATI_x64.exe</file_name>
<main_program/>
</file_ref>
<file_ref>
<file_name>brook64.dll</file_name>
</file_ref>
</app_version>

<app>
<name>milkyway_nbody</name>
</app>
<file_info>
<name>milkyway_nbody_0.03_windows_x86_64.exe</name>
<executable/>
</file_info>

<app_version>
<app_name>milkyway_nbody</app_name>
<version_num>3</version_num>
<file_ref>
<file_name>milkyway_nbody_0.03_windows_x86_64.exe</file_name>
<main_program/>
</file_ref>
</app_version>

</app_info>

It's always worth to crunch for a Project, why else would you do
it? Credits, well they give an amount of work done, that;s what they're for,IMO.

Is it possible to run the nBody app.(?) on CPU+GPU?
Think I posted this question before? Sorry and feel free to remove it :)

Still run the GPU + CPU, Optimized Application, which still works like a charm, is there a Scientific Reason for this change?

Or a result of Moore's Law?

---

Knight Who says Ni

Is it possible to run the nBody app.(?) on CPU+GPU?
Think I posted this question before? Sorry and feel free to remove it :)

Someday, but not now.

Hi

I have a question

I run using the SSE3 then sometimes I get this message 05/10/2010 20:22:17 Milkyway@home Message from server: Your app_info.xml file doesn't have a version of MilkyWay@Home N-Body Simulation.
so the only way to get it is to detach and then reattach the project then the N body runs but the SSE3 disappears from the app-info.xml Then the next one is not an N-body file so to stop it running for 24 hours you abort it and download the SSE3 again and this removes the N_body for the next one.
is there a way to change something so it can automatically tell which one to run without this work around

i just have CPU on my computer

thanks Paul

You need to add the nbody app to your app_info.xml
There are examples in different threads (including this one).

Stop BOINC

Download the nbody app you need and save it in your milkyway folder:
http://milkyway.cs.rpi.edu/milkyway/downloadhttp://milkyway.cs.rpi.edu/milkyway/download/milkyway_nbody_0.07_windows_intelx86__sse2.exe

Add this to your app_info.xml (in your milkyway folder):
<app>
<name>milkyway_nbody</name>
<user_friendly_name>MilkyWay@Home nbody Simulation</user_friendly_name>
</app>
<file_info>
<name>milkyway_nbody_0.07_windows_intelx86__sse2.exe</name>
<executable/>
</file_info>
<app_version>
<app_name>milkyway_nbody</app_name>
<version_num>7</version_num>
<file_ref>
<file_name>milkyway_nbody_0.07_windows_intelx86__sse2.exe</file_name>
<main_program/>
</file_ref>
</app_version>

Start BOINC again

With 5 second cpu calculations, I would say yes!


213029069 208590 5 Oct 2010 23:49:14 UTC 5 Oct 2010 23:50:22 UTC Completed and validated 4.14 4.90 0.02 239.78 MilkyWay@Home v0.19
213031327 51715 5 Oct 2010 23:51:57 UTC 6 Oct 2010 0:35:33 UTC Completed, marked as invalid 434.40 10.16 0.05 0.00 Anonymous platform
213068385 193607 6 Oct 2010 0:36:23 UTC 6 Oct 2010 0:37:27 UTC Completed and validated 5.15 2.31 0.02 239.78 MilkyWay@Home v0.19

---

Personally I like to run GPU apps solely on the GPU and save the CPU for apps that don't run on GPU. I feel this optimizes my productivity.

Personally I like to run GPU apps solely on the GPU and save the CPU for apps that don't run on GPU.

This isn't actually possible. The application runs on the CPU; it manages sending tasks to the GPU and processing results from it. While most of the load is delegated to the GPU, there's no such thing as a pure GPU program.

Hi
my latest 2 are showing the elapsed time going up and the "to completion" going down as usual but the progress is 0-00% according to Boinc

should I abort them?