Avoiding the 511pt GPU WU's

Is there any way to avoid these work units? According to this page those projects aren't exactly the best for hitting big points. It looks like over 1000ppd difference between 511's and 477 or 548 point projects with my Radeon card. My PPD is meager right now, with the 4890 pulling about 2500 on 511pt projects. About 3 minutes a frame.

In addition, this new 8800gt card would be best with the smaller points it seems like. Can I just do -configonly and change the project size to small to hit these?

You would be doing nothing more than hurting the science results by trying to avoid a certain type of WU's. Stanford will only assign the WU's they have a demand for.

I hardly get any 511's anymore. Weird???

Ok I was under the impression that PPD was related to efficiency in processing work orders and that for some reason if the work orders were worth fewer points, that meant they were smaller and the GPU was able to process them quicker for whatever reason.

So what does the project size config correlate to?

Porvalsh said:

Ok I was under the impression that PPD was related to efficiency in processing work orders and that for some reason if the work orders were worth fewer points, that meant they were smaller and the GPU was able to process them quicker for whatever reason.

So what does the project size config correlate to?

Click to expand...

The project size relates more to the download size and the memory requirements.

Well, at last count there were only 8 511's left. Yes they are slow, 3.5K on a OC'd 9800GT as opposed to 5.5K for a 353 pointer. I had my share of them and they're on their way out, at least I hope so, for the science part.

All GPU WUs are small downloads with small memory requirements so the configuration options have no effect on assignment. The benchmark machine (HD3850) folds all the WUs at the same production level, but differences in hardware make very large differences in points production for the donor. We just have to fold them and can't be allowed to cherry pick the best WUs. Imagine the difficulty the project would face if the donors decided what work to do instead of the researchers.