BRP6 work not working properly

I don't believe it did. As I recall (my memory is hazy because I wasn't using GPUs at the time) the app_config.xml mechanism came in around late 2012 with BOINC 7.0.42 or something like that. I think that prior to that, people had to use the very user unfriendly app_info.xml mechanism if they wanted to schedule more than 1 concurrent task per GPU. Bernd avoided this pain and suffering by developing the GPU utilization factor as a project configuration option, easily settable by the user. As I understand it, the success of that feature led to the BOINC Devs, some time later, adopting the idea and providing the functionality through app_config.xml. I may be wrong and I'm sure someone will correct me if I am. I don't remember exactly when the GPU utilization factor came in but my understanding is that it was unique to Einstein when first introduced.

It's really quite easy to use compared to app_info.xml. Sure it's not quite as convenient as editing a single number in the GPU utilization preference box but once you have a basic file installed, editing, followed by a click on 'reread config files' is quick, simple and convenient - and part of the user interface :-).

As an example of its usefulness, I wanted to tweak a dual core host with a GTX650 that had been running with a GPU utilization factor of 0.5. When using that factor, the server supplied cpu_usage of 0.2 and gpu_usage of 0.5 would allow 2 CPU tasks and 2 concurrent GPU tasks to run. I wanted to try 3 GPU tasks and then experiment with CPU tasks. A GPU factor of 0.33 would give the three tasks but would also always allow 2 CPU tasks, which I didn't want. Sure, I could reduce the allowed cores to 50% which would do the job but at the expense of always having a free core even if there were no GPU tasks running at the time. It was just easier to set cpu_usage to 0.35 and gpu_usage to 0.33 through app_config.xml to get the 3+1 task mixture I wanted to test.

For me, however, the biggest disadvantage of GPU utilization factor is not the above but rather that it is tied to venues. Every other host you have in the same venue gets the same change in settings. Of course this is of no concern to people with less than 4 computers, since there are 4 venues available to house the different preference sets. For me it was a real problem until I bit the bullet and deployed app_config.xml on every single host with a GPU.

We're talking about different things. I was aiming at the server including something like

    
        ATI
        0.330000
    

in the application info. I'd call that a GPU utilization factor and it existed before it was adjustable by the user. Let's ignore app_info.xml. Then Bernd added the option to modify this code on the Einstein server side and even later a way to override it at the client side was added to BOINC. That's how I understand it.

I'm not aware of any major changes at Einstein.

You did a very slow GPU task. Now BOINC thinks your whole computer is slow. It will adjust, but probably not to a point where estimates for GPU and CPU are correct at the same time.

In this case it's not about running different projects, it's about doing different types of work at one project.

POGS doesn't have a GPU application. You might want to opt-out from CPU applications at Einstein, then do GPU work for Einstein and CPU work for POGS. There should be little or no interference.