I thought it was the 290 series drivers that were bad.

Oh I'm not saying the 300 drivers are bad... if that were the case the NC forum would be on fire! Not to mention that none of the Top Computers would be using them:)

It could be a coincidence (sample size is too small to say for sure) but it looks like any problem card that isn't a 560Ti, is running the 300 series.

The graphics card which comes with my motherboard is the one I chose to have among certain different vendors of such cards.

Because I am not too familar with how the other types of cards goes with particular motherboards/processors and their respective vendors, my best knowledge goes with what I am currently having.

Is it possible to create some kind of statistics of how many inconclusives goes with specific types of graphics cards (including drivers being used) against the number of tasks which have been returned from one or more host(s) in total in order to be able to make up any comparison?

In this way, the graphics cards showing the best performance should end up at the top of such a comparison list and the graphics card which are having the poorest performance (most inconclusives) should end up at the bottom of such a list (or maybe the opposite way).

Or will this only become prejudice?

I have a ... relatively low number of inconclusives, no invalids, and only timeout errors.

And of course, no sooner do I say that than one of my inconclusives comes up as invalid. But why is my stderr_txt blank?

---

David
Sitting on my butt while others boldly go,
Waiting for a message from a small furry creature from Alpha Centauri.

...
But why is my stderr_txt blank?

It happens, and not only here. Last January Matt Arsenault of MilkyWay@Home started a thread on the boinc_dev mailing list stderr returned unreliable. At the end of the thread he proposed a potential fix, but Dr. Anderson didn't accept the guess at what was causing the problem so didn't accept the fix.

The apps you're using already have a similar but different change by Jason. That's part of the

boinc_exit(): requesting safe worker shutdown ->
boinc_exit(): received safe worker shutdown acknowledge ->

seen at the end when stderr does get transferred properly. I'm pretty sure it helps, but as you've seen not always.

                                                                    Joe