I've read that it's possible (and potentially yield higher credits) to run multiple WU's per GPU.
After browsing the web I saw to create an app_config.xml.
However, I have not seen this working!
Contents of file:

<app_config>
<app>
<name>Poem@Home</name>
<max_concurrent>2</max_concurrent>
<gpu_versions>
<gpu_usage>.500</gpu_usage>
<cpu_usage>1</cpu_usage>
</gpu_versions>
</app>
</app_config>

What have I done wrong here? Please do go easy on me (idiot instructions) as I've never edited this before, I've always run Boinc as standard.
Thanks in advance.

The name of the app is not Poem@Home, that's the project's name.
You can check in client_state.xml what the actual names are that Poem@Home has for its applications and use that.

---

Superficially, that looks OK (*). Two questions:

Did you put the file in the right place (the POEM@Home project sub-folder of your BOINC data folder)?
Did you issue the instruction 'Read config files' from the Options menu in BOINC Manager (Advanced view) afterwards?

If so, there should be useful information in the Event Log.

* I'm not familiar with POEM, but that <name>Poem@Home</name> looks dodgy. It needs an application name, not a project name.

Be prepared to refer to Application configuration for full documentation, if you didn't get your original inspiration from there.

That solved it - thanks.
Knew it would be me lol.

Was a bit of both - app name, not project and wrong location!

Now I'm going to a pain!
Is there any way to do this on just 1 of 2 GPU's?
I have a second PC with 1 x nvidia and 1 x AMD GPU - would like to just run 2 WU's on the AMD.

In the POEM account preferences you can specify to run on only AMD or NV GPUs.

If you want to run 2 tasks per GPU then there is an option in app_config.xml for that.

ngpus
the number of GPU instances (possibly fractional) used by the app version.

0.5 would be the # for 2 per GPU

https://boinc.berkeley.edu/wiki/Client_configuration

Thanks, but what I want to do is run just 1 task on Nvidia and 2 tasks on AMD.
The Nvidia card is getting on a bit now and probably wouldn't cope very well!

You can use the plan_class distinction.

<app_config>
   <app>
       <name>poemcl</name>
       <gpu_versions>
       <gpu_usage>0.5</gpu_usage>
       <cpu_usage>0.04</cpu_usage>
       </gpu_versions>
   </app>
   <app_version>
       <app_name>poemcl</app_name>
       <plan_class>opencl_ati_101</plan_class>
       <avg_ncpus>0.04</avg_ncpus>
       <ngpus>0.5</ngpus>
   </app_version>
   <app_version>
       <app_name>poemcl</app_name>
       <plan_class>opencl_nvidia_101</plan_class>
       <avg_ncpus>0.04</avg_ncpus>
       <ngpus>1</ngpus>
   </app_version>
</app_config>

The above will make the distinction between the ATI and NVIDIA GPUs, by plan_class as defined by the project. I used the information from this post to get to the application and plan_class names. The above should work. If it doesn't, run it by the project developers on their forums.

You do require a minimum BOINC version of 7.2.39 for the above to work.

---

That is fantastic!
Working like a charm - thank you very much